Journal of Adolescent Health 41 (2007) S14 S21

Original article

Prevalence and Predictors of Internet Bullying

Kirk R. Williams, Ph.D.a,* and Nancy G. Guerra, Ed.D.b
b

a
Presley Center for Crime and Justice Studies, University of California at Riverside, Riverside, California
Southern California Center of Excellence on Youth Violence Prevention, University of California at Riverside, Riverside, California
Manuscript received June 19, 2007; manuscript accepted August 28, 2007

Abstract

Purpose: With the Internet quickly becoming a new arena for social interaction, it has also become
a growing venue for bullying among youth. The purpose of the present study was to contrast the
prevalence of Internet bullying with physical and verbal bullying among elementary, middle, and
high school boys and girls, and to examine whether key predictors of physical and verbal bullying
also predicted Internet bullying.
Methods: As part of an ongoing, statewide bullying prevention initiative in Colorado, 3,339 youth
in Grades 5, 8, and 11 completed questionnaires in 78 school sites during the fall of 2005, and
another 2,293 youth in that original sample participated in a follow-up survey in 65 school sites in
the spring of 2006. Questionnaires included measures of bullying perpetration and victimization,
normative beliefs about bullying, perceptions of peer social support, and perceptions of school
climate.
Results: The highest prevalence rates were found for verbal, followed by physical, and then by
Internet bullying. Physical and Internet bullying peaked in middle school and declined in high
school. Verbal bullying peaked in middle school and remained relatively high during high school.
Males were more likely to report physical bullying than females, but no gender differences were
found for Internet and verbal bullying. All three types of bullying were significantly related to
normative beliefs approving of bullying, negative school climate, and negative peer support.
Conclusions: Preventive interventions that target school bullying by changing norms about bullying and school context may also impact Internet bullying, given the shared predictors. 2007
Society for Adolescent Medicine. All rights reserved.

Keywords:

Internet bullying; Prevention; Normative beliefs; School context

Bullying is a form of aggression involving intentional

and harmful behavior marked by repeated engagement and
an asymmetric physical or psychological power relationship
[13]. The specific type of harmful behavior can vary considerably; previous studies of childrens bullying have considered a range of diverse behaviors including name calling,
saying mean things, destruction or taking of property, demanding money, social exclusion, hitting, and kicking [4].
What brands these behaviors as bullying rather than aggression is that they occur repeatedly, typically involving a

*Address correspondence to Kirk R. Williams, Ph.D., Presley Center

for Crime and Justice Studies, 110 College Building South, University of
California at Riverside, Riverside, CA 92521.
E-mail address: [email protected]

weaker victim within the context of an ongoing social interaction [5]. As such, research examining the prevalence,
predictors, and prevention of bullying largely has examined
this behavior as it unfolds within a specific social context.
For children, schools have been the primary context for
studying bullying behavior [2,6 9].
Over the last decade, interest in understanding and preventing bullying among school children in the U.S. and
internationally has surged [6,10]. Such interest coincides
with a growing awareness of the detrimental consequences
of being bullied on childrens well-being as well as the
recognition that bullying is a significant problem in schools
[1]. Still, prevalence rates vary as a function of how bullying is measured, what type of bullying (e.g., physical vs.
verbal) is assessed, the age of respondents, and the country

1054-139X/07/$ see front matter 2007 Society for Adolescent Medicine. All rights reserved.
doi:10.1016/j.jadohealth.2007.08.018

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

where the study takes place. Reported perpetration and

victimization rates typically range from 10% of students
reporting physical bullying or victimization to more than
50% reporting indirect bullying or victimization involving
teasing, name calling, spreading rumors, or verbal aggression [11,12].
Because bullying has been framed as a schoolyard issue,
the focus of research has largely been on bullying in
schoolswhether in the classroom, locker room, hallway,
or bathroom based on the assumption that personal contact is a prerequisite to bullying. Yet, in recent years, technology has transformed the landscape of childrens social
lives. With an estimated 45 million children between the
ages of 10 and 17 in the U.S. alone using the Internet every
day, social interactions have increasingly moved from personal contact in the school room to virtual contact in the
chat room, and Internet bullying has emerged as a new and
growing form of social cruelty. Such bullying is clearly not
physical in nature and has more in common with verbal
bullying. Intimidation is quickly being augmented by humiliation, destructive messages, gossip, slander, and other
virtual taunts communicated through e-mail, instant messaging, chat rooms, and blogs [13]. The Internet has become
a new arena for social interactions, allowing children and
youth to say and do things with a certain degree of anonymity and limited oversight by adult monitors. Internet
bullying, as defined here, refers to the willful use of the
Internet as a technological medium through which harm or
discomfort is intentionally and repeatedly inflicted through
indirect aggression that targets a specific person or group of
persons.
The purpose of the present study is to contrast the prevalence of Internet bullying with physical and verbal bullying
among 5th, 8th, and 11th grade boys and girls and to examine
whether key predictors of physical and verbal bullying also
predict Internet bullying. The focus is on youth involvement
in such bullying as perpetrators, not victims. Perhaps because this type of bullying has unfolded on the fringes of the
adult radar screen, it has only recently been considered in
studies of bullying among children and youth. Relatively
little is known about the prevalence of Internet bullying and
how this compares with other types of bullying for boys and
girls across different ages. Little is also known about
whether Internet bullying is predicted (and potentially prevented) by the same types of factors that have been linked
to physical and verbal bullying in schools [13]. To date,
information on the prevalence of Internet bullying comes
primarily from anecdotal reports and a limited number of
youth surveys. Findings suggest that Internet bullying and
victimization rates are around 25% [13,14]. This is higher
than physical bullying rates but lower than indirect bullying
rates from most school-based prevalence studies. Very few
studies have examined age and gender differences for Internet bullying, although overall bullying appears to peak
during early adolescence, with verbal bullying remaining

S15

high throughout the adolescent years. Furthermore boys are

more likely to engage in physical bullying than girls [1].
Even fewer studies have examined the correlates and
predictors of Internet bullying and whether these are similar
or distinct from factors linked to bullying in schools. Overall, predictors of physical and verbal bullying are quite
similar to predictors of aggression more generally. However
studies of the etiology and prevention of bullying have
emphasized a smaller set of predictors reflecting the social
and normative context of bullying within peer networks and
school settings [10]. Two important predictors linked to
bullying in both prediction and prevention studies are student perceptions of the acceptability (or moral approval) of
bullying and student perceptions that school is an unsupportive context in which peers and adults cannot be trusted.
Bullying is behavior that is harmful to others and falls
within the domain of behaviors with moral consequences:
endorsement of the acceptability of bullying is akin to moral
approval because harm to others is considered to be a key
element of moral reasoning [15]. A consistent finding in
both the aggression and bullying literature is that children
who endorse normative beliefs supporting such behavior are
more likely to be perpetrators [2,1517]. A peer and school
culture that supports bullying is more likely to have individuals who view this behavior as acceptable, further increasing normative support for bullying. Indeed a primary
emphasis of many school-based bullying prevention programs is to change the normative climate so that bullying is
seen as unacceptable [6,10].
Besides normative climate, several studies have examined the impact of various contextual characteristics of
schools on bullying perpetration and victimization. For example, increases in student bullying over time are more
likely in high-conflict, disorganized schools than in lowconflict, harmonious schools [18]. Low levels of supervision within school settings have also been associated with
higher rates of bullying [19]. Disciplinary harshness, safety
problems, and negative peer interactions have also been
linked to behavior problems and bullying [20]. The influence of peers is particularly noteworthy. Within the school
setting, peers can escalate bullying through encouragement
and validation [21]. However, just as peers can enable bullying, they can also provide a supportive social context that
encourages acceptance, belonging, and trust; many effective
bullying prevention programs encourage students helping
other students to form positive peer support systems [2,6].
If students believe that bullying is acceptable and if they
feel disconnected and unsupported at school and by peers,
they should be more likely to engage in all types of bullying
behavior, including Internet bullying. However because
both verbal and Internet bullying can occur behind the
victims back with a greater degree of anonymity than with
physical bullying, both the prevalence and predictors of
Internet bullying are expected to be more similar to verbal
bullying than to physical bullying.

S16

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

Methods
Data for the present research were collected as part of a
larger study evaluating a statewide initiative in Colorado to
strengthen the skills and willingness of youth and adults to
intervene in bullying situations. The Bullying Prevention
Initiative (BPI) is a 3-year, $8.6-million initiative funded by
The Colorado Trust, a private grant-making foundation in
Denver, Colorado. The grantees funded by this initiative
represent school districts, individual schools, or communitybased organizations, evenly split between rural and urban
areas of the state and responsible for implementing bullying
prevention programming in 78 schools across 40 of Colorados 64 counties.
The larger BPI evaluation will provide an empiricallybased understanding of bullying and bystander behavior
among youth, including an increased awareness of this
behavior, social cognitive processes involved in the prevention of bullying, the social context surrounding bullying
incidents, and the involvement of adults and youth in preventing such incidents. These issues are being addressed by
collecting survey data from youth and adults in schools,
collecting data from grantees concerning program implementation, and conducting a supplemental qualitative study
seeking to acquire in-depth information from adults about
challenges and successes in program implementation and
from youth in terms of their awareness of the bullying
prevention programs and their perceived effectiveness. A
prepost survey design collects data from youth in the fall
and the spring of 3 academic years (20052006, 2006
2007, and 20072008) within the 78 schools. This design
allows the assessment of single year changes in individual
youth and contextual (school level) changes over the full 3
years of the BPI. All instruments developed to collect data
from youth were piloted in the summer of 2005 before full
implementation in the fall of that year, with all indices having
acceptable reliabilities (i.e., alpha coefficients .70).
Participants in the present study
The first year of the BPI was a start-up period for both
the prevention programming as well as the evaluation study
in refining its data collection instruments and procedures.
During this year, 3,339 youth completed questionnaires in
the 78 school sites during the fall of 2005, and another 2,293
youth in that original sample participated in a follow-up
survey in 65 school sites in the spring of 2006. Data were
collected in 5th, 8th, and 11th grades, representing transition
years in elementary, middle, and high schools. All data
collection was conducted in compliance with the protocol
approved by the human subjects review board, including
acquiring informed parental consent and youth ascent. To
ensure the quality of data and that school samples are
representative, a subsample of these first-year participants
was selected for the present analysis based on two criteria:
(1) schools must have successfully completed both fall and

Table 1
Demographic characteristics of the total and subsample
Demographic characteristic

Subsample

Total sample

Rural
Urban
5th Grade
8th Grade
11th Grade
Male
Female
White, Non-Latino
Latino
African-American
Asian-American
Native-American
Reduced-cost or free lunch

61.0%
39.0%
43.5%
48.2%
08.3%
45.0%
55.0%
62.0%
26.9%
06.4%
03.5%
01.2%
49.0%

46.0%
54.0%
31.3%
49.7%
19.0%
46.0%
54.0%
60.5%
28.1%
05.9%
04.0%
01.5%
53.1%

Note: Qualification for reduced-cost or free lunch was used as an

indicator of socioeconomic status.

spring data collection, and (2) consent and completion rates

at those schools must be 50% or greater. Applying these
selection criteria yielded a subsample of 1,519 youth participating in both the fall and spring data collection, representing 46 school sites. As shown in Table 1, the subsample
has a greater percentage of rural participants, compared with
the total first-year sample, and a greater percentage of 5th
grade but a lower percentage of 11th grade participants,
compared with the total first-year sample. The percentages
for the remainder of the demographic characteristics are
very similar between the subsample and the total first-year
sample.
Procedures
Data were collected using two different electronic methods, with the choice of methods negotiated with schools in
terms of what was deemed best for their students. However
paper questionnaires were used by a small percentage of
youth absent the day for which data collection was scheduled (4% of the total first-year sample and 1.6% of the
subsample analyzed here). First, data collectors used a liquid crystal display projector to present questionnaire items
in classrooms of approximately 30 students or less (used by
61.7% of the students in this subsample). After the data
collectors read each question aloud, youth used a wireless
response pad to enter their answers, which were automatically recorded in an electronic database and linked to the
student identification code. The questionnaire was administered in English or Spanish as needed, using standard backtranslation methods. Second, the questionnaire was adapted
to a web-based format linked to the electronic database
(36.7% of students in this subsample used this method). The
youth web-based questionnaire was administered in school
computer laboratories. Data collectors assisted youth in
logging on to the password-protected questionnaire and
were available for assistance as youth answered questions at

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

their own pace. No evidence was found that these different

data collection procedures influenced responding.
Measures
Bullying perpetration. Items bearing on the perpetration
(not victimization) of different types of bullying were
adapted from Espelage et al [22]. Youth were asked to
respond to the following four items: I pushed, shoved,
tripped, or picked fights with students I know are weaker
than me; I teased or said mean things to certain students;
I spread rumors about some students; and I told lies
about some students through e-mail or instant messaging.
Numeric coding (in parentheses) and response options included (1) never, (2) one or two times, (3) several times, and
(4) a lot. The first item was used to measure physical
bullying perpetration, and the second and third items were
combined to measure verbal bullying perpetration. Internet
bullying perpetration was measured by the fourth item.
Even thought these items were used separately to identify
perpetrators of physical, verbal, and Internet bullying, they
form a reliable measure when combined as a summary
index (alpha coefficient .73).
As mentioned below, the appropriate temporal ordering
between the predictors examined in this empirical analysis
and these types of bullying perpetration requires using data
collected in the fall of 2005 for the predictors and data
gathered in the spring of 2006 for bullying perpetration.
Therefore, the reference period for the four bullying items
was since the school year began. The distributions for
these variables are highly skewed, with a high concentration
of participants scoring one (i.e., never) and a precipitous
decline in the distribution of participants across the higher
scores. Hence these variables were dichotomized, with comparisons made between those participants who reported
never engaging in these types of bullying (scoring 1) with
those who reported doing so one or more times during the
school year (scoring 2 4).
Moral approval of bullying. Moral beliefs about bullying
perpetration and bystander involvement in bullying situations were assessed by asking participants to evaluate six
different items on a four point Likert-type scale ranging
from really wrong to perfectly ok. These items were
taken from the Normative Beliefs About Aggression Scale
[15] and were modified slightly to refer to bullying instead
of aggression. The items included bullying perpetration as
mentioned above (including Internet bullying), in addition
to items pertaining to negative bystander involvement, such
as encouraging others to fight smaller students or spread lies
and rumors about them. These six items were summed to
construct a summary additive index (alpha coefficient
.93). The distribution on the index was positively skewed,
but it had a clustering of cases at the high end of the
continuum of scores. The distribution of this measure was
adjusted by sub-dividing it into six categories having ap-

S17

proximately equal distributions. High scorers indicate participants who approve bullying perpetration and negative
bystander behavior in general, not behavior-specific approval. Participants scoring low on this index disapprove of
bullying and negative bystander behavior in general. This
measure is appropriate given the emphasis on capturing the
normative orientation of students about bullying overall, not
behavior-specific moral beliefs.
School climate. Student perceptions of school climate were
assessed using the California School Climate Scale [23].
This measure contains nine items about teachers, school
staff and administrators, school policy, and a students perceived personal connection to the school. For example,
participants were asked whether they disagree or agree with
statements like My teachers respect me, My teachers are
fair, or Teachers at my school are nice people. Other
items addressed whether the principal in their school listens
to the ideas of students, whether students who break school
rules are treated fairly, and whether teachers and staff are
doing the right things to prevent bullying in general, not
specific forms of bullying in the school. The nine items were
summed to form an additive index (alpha coefficient .84),
with scores ranging from 9 to 36, given the response categories of the individual items. Respondents with high scores
perceived a positive school climate, whereas those with low
scores perceived a more negative school climate.
Perceived peer support. This four-item scale focused on
positive and negative qualities of peers as a source of social
support. It was adapted from the Generalized Perception of
Peers Scale [24]. Regardless of social context (e.g.,
schools), participants were asked to assess whether students their age care about what happens to them, will help
them in time of need, can be trusted, and are sensitive to
their feelings. Response options range from no, not at all
to yes, completely. The four items were summed to form
an additive index, with scores ranging from a low of 4 to a
high of 16, given the response categories for each of the
individual items. High scores indicated higher perceptions
of peers as supportive, whereas low scores indicated the
opposite view. The alpha coefficient for this scale was .79.
Results
The results presented below are arranged according to
the two primary research objectives of this analysis: (1) to
determine the prevalence of Internet compared with verbal
and physical bullying perpetration in this sample of youth,
and (2) to determine whether predictors of Internet bullying
perpetration are similar to predictors of verbal and physical
bullying perpetration. The first objective is addressed simply by tabulating the distributions of the three forms of
bullying and examining whether these distributions vary by
gender and grade level (5th, 8th, and 11th grades). The
empirical examination by gender and grade was done by

S18

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

estimating their effects on each form of bullying perpetration through logistic regression, given the dichotomous bullying measures used.
The second objective is addressed by estimating the
bivariate relations between each of the three predictors and
the three forms of bullying perpetration. As noted above,
these behavioral measures are dichotomized, differentiating
between participants who report never perpetrating such
behavior and those reporting they did so one or more times
during the school year. Hence, the empirical relations were
estimated using logistic regression, similar to the estimation
of gender and grade level effects. The results of these
analyses are presented in the text and in line graphs, where
the predicted probabilities of each bivariate equation are
plotted against the predictor variable respective to each
equation [25]. Statistical tests for the equality of estimated
coefficients across equations are also conducted. Such tests
determine whether the estimated effects for Internet bullying
perpetration are significantly different from those of physical
and verbal bullying perpetration (i.e., are such effects similar or
different for the three forms of bullying). The following formula was applied for these statistical tests [26]:
Z

b1 b2

SEb12 SEb22

Prevalence of bullying
Table 1 shows the distribution of Internet, physical, and
verbal bullying perpetration for the total sample of youth.
Verbal bullying is clearly most prevalent for the total sample, followed by physical bullying and then bullying via
e-mail or instant messaging. In short, Internet bullying was
a part of the behavioral repertoire of only a minority of
youth in this sample during the last school year, but its
prevalence is nontrivial. These three types of bullying perpetration are clearly interrelated, with ordinal associations
(gamma coefficients) ranging from .66 for the relation between Internet and physical bullying to .87 for the relation
between Internet and verbal bullying. However, distinctions
remain, as suggested by 24.8% of the sample refraining
from any type of bullying, 37.9% engaging in only one type,
30.7% perpetrating two types, and only 6.6% self-reporting
involvement in all three types.
Logistic regression was used to estimate the effects of
gender and grade on each of the three forms of bullying
perpetration. No gender differences were found for Internet
and verbal bullying, although such differences were pronounced for physical bullying, with males being more than
twice as likely as females to report perpetrating such behavior (b .79, odds ratio [OR] 2.21, p .00). Grade
was significantly related to all forms of bullying perpetration, with the estimated effect being literally identical between Internet and verbal bullying (b .20, OR 1.22,

Table 2
Percentage of youth self-reporting verbal, physical, and Internet bullying
perpetration, by grade
Type of
bullying

5th Grade

8th Grade

11th Grade

Total sample

Verbal
Physical
Internet

32.6%
34.8%
04.5%

78.5%
44.6%
12.9%

72.3%
37.8%
09.9%

70.7%
40.3%
09.4%

p .00), but substantially smaller for physical bullying (b

.08, OR 1.09, p .00).
Given these significant grade-level effects, the distribution by grade for verbal, physical, and Internet bullying
perpetration is also displayed in Table 2. About one third of
5th-graders reported engaging in verbal bullying in the past
school year, with the prevalence rates then peaking in 8th
grade and dropping off only slightly by 11th grade. A similar
pattern holds for physical bullying as well, although the
prevalence rates are lower than verbal bullying, especially
among 5th and 11th-graders. A relatively small percentage of
5th-graders reported engaging in Internet bullying, with the
distribution peaking in 8th grade and again declining slightly
among 11th-graders.
Predictors of bullying
Moral approval of bullying. Figure 1 presents the results of
the logistic regressions in which the empirical relations
between moral approval of bullying and verbal, physical,
and Internet bullying perpetration were estimated. The results show that this predictor is significantly and positively
related to all three forms of bullying as expected; that is,
beliefs endorsing bullying and negative bystander behavior
are associated with self-reported involvement in verbal,
physical, and Internet bullying. Specifically, an increase in
one of the six ordinal categories of the moral beliefs measure (see description of measure above) is associated with a
43% increase in the odds of verbal bullying (b .36, OR
1.43, p .00), a 27% increase in the odds of physical
bullying (b .24, OR 1.27, p .00), and a 24% increase
in the odds of Internet bullying (b .26, OR 1.24, p
.00). However, the estimated effects for verbal bullying are
significantly greater than those of both physical and Internet
bullying (Z 2.40 and Z 2.34, p .00, respectively), but
the difference between the estimated effects for physical
and Internet bullying is not statistically significant. In short,
although moral beliefs approving of bullying and negative
bystander behavior are significantly and positively related to
all three forms of bullying perpetration, the estimated effects for Internet bullying appear to be significantly lower
than for verbal bullying.
Perceived school climate. The results of the logistic regressions for perceived school climate are shown in Figure 2.
Also as expected, the line graph shows that the more youth
perceive themselves as connected to their schools, with the

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

S19

Figure 1. Empirical relations between moral beliefs and the predicted probability of verbal, physical, and Internet bullying perpetration.

climate being trusting, fair, pleasant, etc. (i.e., a positive

school climate), the lower is their self-reported involvement
in verbal, physical, and Internet bullying perpetration. An
increase of a single unit on the 9 36-point perceived school
climate index (see description of measure above) is associated with a 7% decline in the odds of physical bullying (b
.07, OR .93, p .00), a 9% decline in the odds of
Internet bullying (b .10, OR .91, p .00), and a 10%

decline in the odds of verbal bullying (b .11, OR .90,

p .00). Moreover, the difference between the estimated
effects for Internet bullying and either physical bullying or
verbal bullying is not statistically significant.
Perceived peer support. As displayed in Figure 3, youth
perceptions that friends their age are trustworthy, caring,
and helpful are significantly associated with lower self-

Figure 2. Empirical relations between school climate and the predicted probability of verbal, physical, and Internet bullying perpetration.

S20

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

Figure 3. Empirical relations between perceived peer support and the predicted probability of verbal, physical, and Internet bullying perpetration.

reported participation in verbal, physical, and Internet bullying. Again, these empirical relations were anticipated. A
single unit increase on the four- to 16-point index of perceived peer support (see description of measure above) is
associated with a 7% decline in the odds of physical bullying (b .08, OR .93, p .00), with the estimated
effects being greatest for Internet (b .14, OR .87, p
.00) and verbal (b .15, OR .86, p .00) bullying.
Similar to the empirical relations between perceived school
climate and these three forms of bullying perpetration, the
estimated effects for Internet bullying are not significantly
different from those of verbal or physical bullying.
Summary and conclusion
The findings on the prevalence of bullying perpetration
suggest that distributions vary by type, with verbal being
most prevalent, followed by physical and then by Internet
bullying. Physical and Internet bullying peaked in 8th grade
and declined in 11th grade, whereas verbal bullying peaked
in 8th grade and remained relatively high in the 11th grade.
Males were more likely than females to report physical
bullying perpetration. Consistent with the expectation that
Internet bullying and verbal bullying would share common
features, no gender differences were found for prevalence of
Internet and verbal bullying.
Three predictors of bullying were empirically examined.
One reflects an individuals normative orientation about the
moral acceptability of bullyingthat is, whether such behavior is right or wrong. However, the other two predictors

capture perceptions of either the context in which youth

regularly participate (i.e., schools) or the nature of peers
with whom youth regularly interact (i.e., students their age).
Regardless, Internet bullying was significantly related to all
three predictors, and these empirical relations were similar
to those of physical and verbal bullying, with the one exception of moral beliefs having a significantly greater positive estimated effect on verbal bullying than on Internet
bullying. These findings suggest that the causal pathways to
Internet bullying may not be unique; rather, it appears to
share common causal pathways with other forms of bullying, particularly verbal bullying.
Technological advances (e.g., Internet e-mail, web sites,
and blogs) merely provide yet another venue through which
bullying among youth can occur. Indeed a limitation of the
current study is that we did not consider a wider range of
electronic bullying methods such as picture cell phones and
text messages. However preliminary evidence from the second year of the larger study suggests that expanding the
scope of the Internet question does not make a major difference. Specifically, the second-year student survey added
cell phone text messaging and bullying on various websites.
The result was an Internet bullying prevalence rate of 13.6%
for the full pre-tested sample in year two, compared wtih
12.0% for the full pre-tested sample in year 1. These rates are
comparable to more recent estimates of Internet bullying
(13.1%) based on only 84 adolescents 1318 years of age [27].
The present study highlighted common predictors of
different types of bullying, including Internet bullying. The

K.R. Williams and N.G. Guerra / Journal of Adolescent Health 41 (2007) S14 S21

findings suggest that preventive interventions can impact

these diverse types of bullying by changing normative beliefs about the acceptability of bullying while simultaneously considering how to increase trust and support
among peers and within the school setting. For schools, this
suggests a whole school approach to bullying prevention
that facilitates changes in beliefs and behaviors toward
greater support, trust, and cohesion [6,10]. Furthermore, it is
important to consider additional potential predictors uniquely
linked to the anonymous nature of bullying via the Internet or
other communication technologies (e.g., cell phone cameras, text messaging) that may further enhance and understanding of this behavior and thus preventive efforts in this
domain. Once again, however, the findings reported here
underscore a critical point of this paper: Common social
forces influence the various ways in which bullying is expressedthrough physical aggression, verbal aggression, or
aggression perpetrated through new communication technologies.
References
[1] Nansel TR, Overpeck M, Pilla RS, et al. Bullying behaviors among
U.S. youth: Prevalence and association with psychological adjustment. JAMA 2001;285:2094 2100.
[2] Olweus D. Bully/victim problems among schoolchildren: Basic facts
and effects of a school based intervention program. In K Rubin, D
Pepler, eds. The development and treatment of childhood aggression.
Hillsdale, NJ: Lawrence Erlbaum; 1991:411 48.
[3] Smith PK, Junger-Tas J, Olweus D, Catalano R, Slee P, eds. The
nature of school bullying: A cross-national perspective. New York:
Routledge; 1999.
[4] Arora CMJ. Defining bullying: Toward a clearer and general understanding and more effective intervention strategies. School Psychol
Int 1995;17:31729.
[5] OConnel P, Pepler DJ, Craig W. Peer involvement in bullying:
Insights and challenges for intervention. J Adolesc 1999;22:43752.
[6] Olweus D, Limber SP. Bullying prevention program. In: Elliott DS,
ed. Blueprints for Violence Prevention: Book Nine. Boulder, CO:
University of Colorado; 1999.
[7] Orpinas P, Horne AM, Staniszewski D. School bullying: Changing the
problem by changing the school. School Psychol Rev 2003;23:431 44.
[8] Pelligrini AD. Bullying, victimization, and sexual harassment during
the transition to middle school. Educ Psychol 2002;37:151 63.
[9] Rodkin PC, Hodges EVE. Bullies and victims in the peer ecology:
Four questions for psychologists and school professionals. School
Psychol Rev 2003;2:384 400.

S21

[10] Espelage DL, Swearer S, eds. Bullying in American schools. Mahwah, NJ: Lawrence Erlbaum; 2004.
[11] Craig WM, Pepler DJ. Observations of bullying and victimization in
the schoolyard. Can J School Psychol. 1997;13:41 60.
[12] Demaray MK, Malecki CK. Perceptions of the frequency and importance of social support by students classified as victims, bullies, and
bully/victims in an urban middle school. School Psychol Rev 2003;
23:47190.
[13] Willard NE. Cyberbullying and cyberthreats. Eugene, OR: Center for
Safe and Responsible Internet Use; 2006.
[14] Kowalski R. et al. 2005. Electronic bullying among school-aged
children and youth. Paper presented at the annual meeting of the
American Psychological Association. Washington, DC: American
Psychological Association.
[15] Huesmann LR, Guerra NG. Social norms and childrens aggressive
behavior. J Pers Soc Psychol 1997;72:408 19.
[16] Bentley KM, Li AKF. Bully and victim problems in elementary
school and students beliefs about aggression. Can J Sch Psychol
1995;22:153 65.
[17] Espelage DL, Swearer S. Research on bullying and victimization:
What have we learned and where do we go from here? Sch Psychol
Rev 2003;23:365 85.
[18] Kasen S, Berenson K, Cohen P, Johnson JG. The effects of school
climate on changes in aggressive and other behaviors related to
bullying. In: Espelage DL, Swearer S, eds. Bullying in American
Schools. Mahwah, NJ: Lawrence Erlbaum; 2004:187210.
[19] Craig WM, Pepler DJ, Atlas R. Observations of bullying in the
playground and in the classroom. Sch Psychol Int 2000;2:2236.
[20] Brand S, Felner R, Shim M, et al. Middle school improvement and
reform: Development and validation of a school-level assessment of
climate, cultural pluralism, and school safety. J Educ Psychol 2003;
95:570 8.
[21] Salmivalli C, Huttunen A, Lagerspetz KM J. Peer networks and
bullying in schools. Scand J Psychol 1997;38:30512.
[22] Espelage DL, Holt M, Henkel RR. Examination of peer group contextual effects on aggression during early adolescence. Child Dev
2003;74:20520.
[23] Furlong MJ, Bates MP, Smith DC, Kingery P, eds. Appraisal and
Prediction of school violence: Methods, Issues and Contents. Hauppauge, NY: Nova Science; 2004.
[24] Salmivalli C, Ojanen T, Haanpaa J, Peets K. Im OK but Youre
Not and other peer relational schemas: Explaining individual differences in childrens social goals. Dev Psychol 2005;41:36375.
[25] Hamilton LC. Regression with Graphics. Pacific Grove, CA: Brooks/
Cole Publishing; 1992.
[26] Paternoster R, Brame R, Mazerolle P, Piquero A. Using the correct
statistical test for the equality of regression coefficients. Criminology
1998;36:859 66.
[27] Raskauskas J, Stolz AD. Involvement in traditional and electronic
bullying among adolescents. Dev Psychol 2007;43:564 75.