Content uploaded by Dorothy L Espelage
Author content
All content in this area was uploaded by Dorothy L Espelage on Jan 03, 2014
Content may be subject to copyright.
A Meta-Analysis of School-Based Bullying Prevention
Programs’ Effects on Bystander Intervention Behavior
Joshua R. Polanin
Loyola University Chicago
Dorothy L. Espelage
University of Illinois Urbana—Champaign
Therese D. Pigott
Loyola University Chicago
Abstract. This meta-analysis synthesized bullying prevention programs’ effec-
tiveness at increasing bystander intervention in bullying situations. Evidence
from 12 school-based programs, involving 12,874 students, indicated that overall
the programs were successful (Hedges’s g⫽.20, 95% confidence interval [CI] ⫽
.11 to .29, p⬍.001), with larger effects for high school (HS) samples compared
to kindergarten through eighth-grade (K-8) student samples (HS effect size [ES]
⫽0.43, K-8 ES ⫽0.14; p⬍.05). A secondary synthesis from eight of the studies
that reported empathy for the victim revealed treatment effectiveness that was
positive but not significantly different from zero (g⫽.05, 95% CI ⫽⫺.07 to .17,
p⫽.45). Nevertheless, this meta-analysis indicated that programs increased
bystander intervention both on a practical and statistically significant level. These
results suggest that researchers and school administrators should consider imple-
menting programs that focus on bystander intervention behavior supplementary to
bullying prevention programs.
Bullying perpetration often occurs when
bystanders are present (Hawkins, Pepler, &
Craig, 2001; Lagerspetz, Bjorkqvist, Bertz, &
King, 1982). In fact, some research has indi-
cated that more than 80% of the time an ob-
server witnesses victimization (O’Connell,
Pepler, & Craig, 1999). Despite the presence
of witnesses and bystanders, nearly 1 in 3
children report victimization by a bully in the
past 2 months (Frey, Hirschstein, Edstrom, &
Snell, 2009; Nansel et al., 2001; Wang, Ian-
notti, & Nansel, 2009). Consequentially, bul-
lying occurs with an audience of members
who play multiple roles (Salmavalli, Lager-
spetz, Bjorkqvist, Osterman, & Kaukianen,
1996) and often fail to intervene on behalf of
the victim with regularity. These bullying in-
cidents have lasting negative effects on the
bully, victim, and bystanders (Olweus, 2002;
Swearer, Espelage, Villancourt, & Hymel,
2010; Sweeting, Young, West, & Der, 2006;
Stevens, Oost, & Bourdeaudhuij, 2004).
Correspondence regarding this article should be addressed to Joshua R. Polanin, Loyola University
Chicago, 820 N. Michigan Ave., Suite 1022C, Chicago, IL 60611; e-mail: jpolanin@luc.edu
Copyright 2012 by the National Association of School Psychologists, ISSN 0279-6015
School Psychology Review,
2012, Volume 41, No. 1, pp. 47–65
47
The past 20 years have seen a burgeon-
ing of bullying prevention programs (Fergu-
son, Miguel, Kilburn, & Sanchez, 2007; Ryan
& Smith, 2009; Ttofi & Farrington, 2009,
2011). Researchers, school administrators, and
teachers have used myriad designs, theories,
and techniques in an attempt to mitigate the
prevalence of bullying (Astor, Meyer, Ben-
benishty, Marachi, & Rosemond, 2005). Ttofi
and Farrington’s (2011) recent large-scale
meta-analysis of over 90 studies found that the
majority of these programs have been success-
ful at slowing the rate of bullying.
Although successful bullying programs
remain important accomplishments, Ttofi and
Farrington (2011) found that few programs
specifically target the behavior of bystanders
(i.e., an individual who witnesses bullying).
As such, prevention programs deemphasize a
population that constitutes between 60% and
70% of primary or secondary school students
(Glew, Fan, Katon, Rivara, & Kerntic, 2005;
Rivers, Poteat, Noret, & Ashurst, 2009). This
program oversight is unfortunate because ob-
servational research has found that when by-
standers intervene on behalf of the victim,
they successfully abate victimization more
than 50% of the time (Craig, Pepler, & Atlas,
2000; O’Connell et al., 1999).
Supported by the knowledge that by-
standers can successfully intervene on behalf
of the victim, a small amount of literature has
focused recently on increasing this behavior.
These programs explicitly emphasize the im-
portance of bystander intervention behavior
and measure this construct. Given these con-
ditions, the purpose of this meta-analysis is to
synthesize school-based bullying prevention
programs’ effectiveness to change bystander
intervention behavior. We also aggregated the
program’s influence on empathy for the victim
as a secondary synthesis because it has re-
ceived recent investigation (Gini, Albiero,
Benelli, & Altoe`, 2007). The following sum-
marizes the relevant literature, provides a
comprehensive examination of the synthesis
process and quantitative analysis outcomes,
and elucidates moderator analysis and publi-
cation bias results. Suggestions for future re-
search and policy are also provided.
Bullying in the Schools
Definition
Olweus (1973) first described bullying
as “mobbing” where a group or individual
teases or harasses another individual. As such,
early research focused solely on the physical
aspects of school environment (e.g., teacher–
student ratio), but found little connection to
perpetration or victimization (Swearer et al.,
2010). Recently, Frey et al. (2009) described
bullying as a social construct that disrupts
social connections among students. Ross and
Horner (2009) summarized the plethora of
definitions:
Common definitions of bullying involve re-
peated acts of aggression, intimidations, or
coercion against a victim who is weaker in
terms of physical size, psychological or so-
cial power, or other factors that result in a
notable power differential. (p. 748)
The bully construct has received much
research and hosts of definitions remain.
Taken together, the bully generally involves
an individual or group who incites physical or
emotional abuse on another individual or
group. Although other research exists on In-
ternet and workplace bullying (Mishna, Cook,
Saini, Wu, & MacFadden, 2010), this review
focuses on school bullying.
Prevalence and Negative Effects
School bullying is not a problem local to
the United States; rather, it is recognized
worldwide (Espelage & Swearer, 2003). Re-
ports from Europe to North America have
indicated that anywhere from 1% to 50% of
students had been bullied or victimized within
the last 2 months (Wang et al., 2009). Some
observations have shown that as many as 30%
of students were involved in bullying as either
the bully or victim (Frey et al., 2009). Hymel
and Swearer (2010) recently reported that 35%
of students indicated being bullied at least
once in the last 2 months with as many as 11%
of those sampled reported being bullied more
than 2 or 3 times in the last 2 months. More-
over, some research has found that bullying
roles remain relatively stable across time. In a
School Psychology Review, 2012, Volume 41, No. 1
48
sample of 516 middle school students, Espel-
age, Bosworth, and Simon (2001) found that
individuals who perpetrated bullying contin-
ued to across multiple years. This finding has
been replicated from other settings and popu-
lations (McDougal, Hymel, & Vailliancourt,
2009; Scholte, Engels, Overbeek, Kemp, &
Haselager, 2007; Sourander, Helstela, Hele-
nius, & Piha, 2000).
Although bullying’s pervasiveness is
cause for concern, the negative consequences
related to bullying remain tantamount. Re-
searchers have indicated that bullying has
been linked to anger and misconduct (Bos-
worth, Espelage, & Simon, 1999), criminal or
delinquent behavior (Olweus, 2002), and sui-
cidal ideations (Kaltiala-Heino, Rimpela,
Marttunen, Rimpela, & Rantanen, 1999). Vic-
timization, on the other hand, has been linked
to poor physical health (Rigby, 1999), low
self-esteem (Rigby & Slee, 1993), depression
(Sweeting et al., 2006), anxiety (Craig, 1994),
and school avoidance (Kochenderfer & Ladd,
1996). Indeed, the negative effects of bullying
and being bullied are persistent and
problematic.
Bystander
Definition, Roles, and Negative Effects
A significant proportion of individuals
within school systems are considered individ-
uals who are bystanders of bullying (Glew et
al., 2005). Twemlow, Fonagy, and Sacco
(2004) defined a bystander as an individual
who lacks participation in bullying scenarios
as either the bully or victim. The bystander
may actively intervene to stop the bully, en-
courage the bully to continue, or view bullying
passively; bystanders can be either boys or
girls (Cowie, 2000; Smith, Twemlow, &
Hoover, 1998).
There are specific roles that the by-
stander can demonstrate. Some authors refer
to the bystander as a passerby, observer, wit-
ness, or participant (Salmivalli, Kaukianinen,
& Voeten, 2005; Twemlow et al., 2004); oth-
ers described roles in relation to sustaining or
preventing the bullying behavior such as rein-
forcer (e.g., laughing or seeing what is hap-
pening), assistant (e.g., follower of the bully),
defender (e.g., being supportive of the victim),
or outsider (e.g., remaining away from the
bullying situation; Salmivalli et al., 1996). For
the purposes of this review, we defined the
bystander generally as any student who wit-
nessed a bullying episode, with the operation-
alizing characteristic being the witnessing
presence of bullying, regardless of other
characteristics.
Despite these literature discordances,
few disagree about the adverse effects wit-
nessing bullying can have on the bystander.
Bystanders felt significantly more uncomfort-
able in bullying situations compared to bullies
(Stevens et al., 2004), and reported feelings of
anxiety and insecurity (Rigby & Slee, 1993).
This anxiety due to witnessing bullying has
been linked to aggressive retaliation (Musher-
Eizenman et al., 2004), and the fear of being
bullied often prevented bystanders from seek-
ing adult help (Unnever & Cornell, 2003). A
recent large-scale study conducted in the
United Kingdom found that compared to per-
petrators, bystanders were at elevated risk for
nonclinical outcomes (i.e., interpersonal sen-
sitivity), and compared to victims bystanders
were more likely to have elevated levels of
substance abuse (Rivers et al., 2009).
Evidence of the Bystanders’ Effects on
Bullying
Individuals who are bystanders remain
present more than 80% of bullying situations
(O’Connell et al., 1999), and therefore some
research has focused on a social-ecological
model of bullying prevention and intervention
(Frey et al., 2009; Swearer & Espelage, 2004).
The social presence and pervasiveness of the
bystander fosters myriad opportunities to in-
tervene. For example, bystanders supported
victims by reporting bullies to adults when
participating in a setting specifically designed
to change bullying behavior patterns through
bystanders (Sharp, Sellors, & Cowie, 1994).
Ross and Horner (2009) recently implemented
a school-wide bullying intervention program
that resulted in a decrease in reinforcing by-
stander behavior and bullying perpetration
A Meta-Analysis of School-Based Bullying
49
overall. Moreover, interventions that focused
on dealing with conflict through peers instead
of direct interventions with adults led to pos-
itive effects (Cowie & Hutson, 2006), and an
individual’s willingness to intervene in bully-
ing situations was inversely related to the
amount of peer-group bullying perpetration
(Espelage, Green, & Polanin, 2011).
Bystander Intervention Program
Characteristics
To date, best practice guidelines to pro-
mote effective bystander intervention behav-
iors remain undefined because research find-
ings varied widely with regard to their imple-
mentation focuses and approaches. Several
mediums for interventions have been studied
to teach children about bystander behavior,
including classroom-based drama (Merrell,
2004), media such as videotaped reenactments
(McLaughlin, 2009; Schumacher, 2007), and
individualized computer-adaptive software to
track students’ progress within social scenar-
ios and provided feedback on effective by-
stander behavior (Evers, Prochaska, Van Mar-
ter, Johnson, & Prochaska, 2007). However,
all of these programs focus on bystander be-
havior perhaps, because there seems to be
some support for targeting peer-group behav-
iors to mitigate individual bullying (Salmivalli
et al., 1996). Peer-group interventions often
encourage bystander intervention (Andreou,
Didaskalou, & Vlachou, 2008; Frey et al.,
2009; Stevens et al., 2000) or enhance by-
stander empathy for the victim (Gini et al.,
2007; Nickerson, Mele, & Princiotta, 2008).
However, few studies have examined the ef-
fects of peer-group interventions on previctim
empathy (Merrell, Gueldner, Ross, & Isava,
2008) and there are resulting guidelines to
promote this behavior.
Previous Meta-Analyses on Bullying
Prevention Programs
A number of recent quantitative meta-
analyses (Ferguson et al., 2007; Ttofi & Far-
rington, 2009, 2011; Merrell et al., 2008;
Smith, Schneider, Smith, & Ananiadou, 2004)
and qualitative systematic reviews (Ryan &
Smith, 2009) have been conducted regarding
bullying and victimization intervention and
prevention programs. However, none of these
meta-analyses focused specifically on by-
stander intervention constructs. Merrell et al.’s
(2008) review included three studies that mea-
sured “intervene to stop bullying behavior,”
which resulted in a mean effect size of 0.17,
but this was a secondary analysis of a small
number of studies. Therefore, the goal of the
present study was to conduct a meta-analysis
that would directly address bystander inter-
vention behavior and empathy attitudes.
Given this goal, two primary research
questions are addressed:
1. What is the average treatment effect,
across the current literature, of bullying
prevention programs on bystander in-
tervention behavior?
2. What study characteristics produced the
largest treatment effect?
A secondary research question ad-
dressed bystander empathy for the victim:
3. What is the average treatment effect,
across the current literature, of bullying
prevention programs on bystander em-
pathy for the victim?
Method
Search Strategy
We used a comprehensive search to re-
trieve articles from the international research
literature within the last 30 years (1980 –
2010). We searched primarily five online da-
tabases: Dissertation Abstracts International,
Education Resources Information Center
(ERIC), PsycINFO, Medline, and Science Di-
rect. Combinations of the following terms
were used: “bystander or participant or de-
fender or other, ” “bully or victim, ” “school,
school program, or program, ” “prevention or
intervention, ” “aggression, ” and “not higher
education or not cyber-bully.” To ensure that
the identified studies focused on bystander
behavior as the primary goal, these terms were
searched in the abstract of the study. In addi-
School Psychology Review, 2012, Volume 41, No. 1
50
tion, we searched the bibliographies of all
articles selected for relevant studies.
The search retrieved 360 total articles,
but only 83 were unique and compared to the
criteria listed in the next section. Of those 83,
53 were deemed irrelevant, 13 did not address
the intervene construct, 6 failed to include a
control group, and 1 was a repeat of a previous
study. Finally, we corresponded with a num-
ber of experts in the field to ensure inclusion
of all relevant articles. This correspondence
produced 1 relevant study. Hence, we re-
viewed and included 11 studies total.
Criteria for Considering Studies for
Review
The present study focused on school-
based interventions that emphasized changing
the bystander’s intervention behavior. To as-
sess the effects of these programs, we col-
lected peer-reviewed studies published or con-
ducted from 1980 to 2010, based solely within
a school system and intended purposefully to
modify bystander intervention behavior. Sub-
sequently, we excluded studies that focused on
changing bullying behaviors primarily and
collected a bystander measure only as a sec-
ondary procedure. The review included inter-
ventions from the United States and Europe,
but we limited inclusion to English-written
studies.
We reviewed studies that included par-
ticipants from the kindergarten through 12th-
grade population, but interventions with
school-aged children based outside the school
setting were excluded. In addition, we at-
tempted to collect studies that included “at-
risk” students and the general population, but
none of the studies distinguished between
these populations. It should be mentioned that
one study attempted to deconstruct the by-
stander into several types of bystanders to
observe treatment effects (Evers et al., 2007).
Although the deconstruction was informative,
it was the only study to implement such a
procedure. As such, for that study we used the
average intervention effects across all by-
stander types.
Furthermore, we included only studies
that used a treatment-control research design.
These designs included true experimental ran-
domly assigned groups, nonrandom quasi-ex-
perimental designs, and nonrandomly as-
signed matched group. We also included all
control group types; these included wait list,
treatment-as-usual, and “straw-man.” How-
ever, single-group pre/post-test (e.g., gain
scores) and cohort designs were excluded.
Outcomes
Studies must have included a bystander
intervention measure. We operationalized this
outcome as a measure that assessed the con-
tribution of the bystander to a bullying situa-
tion (Frey et al., 2005, 2009). Therefore, we
included studies that measured intention to
intervene, intention to stop bullying, direct
intervention, or conversely, difficulty in re-
sponding assertively to a bullying situation.
For example, Andreou et al. (2008) included
items that assessed students’ intention to in-
tervene on a 5-point Likert scale. We included
items that concerned students’ intention to
“seek teacher’s help,” “react against bullying,”
and “support the victims of bullying” (p. 241).
Table 1 provides the measures used, the
study’s stated construct, and the number of
items combined to create the measure.
In addition to the intervention outcome,
we collected results of the program’s effects
on changes in attitudes of empathy toward the
victim. We operationalized the empathy out-
come as a measure that indicated empathy for
the victim. For instance, Stevens et al. (2000)
used an empathy measure that included “feel-
ing sad about students who are bullied” and
“unpleasantness when another student is being
bullied” (p. 26). Measures that included items
similar to this construct constituted an empa-
thy scale. We should also mention that this
outcome constituted a secondary outcome
measure and thus was not a criterion for syn-
thesis inclusion. Studies that included a by-
stander intervention behavior measure but
failed to include a measure of empathy were
included.
A Meta-Analysis of School-Based Bullying
51
Table 1
Characteristics of the Studies used in the Meta-Analysis
Study (DoP) Type Location
N
(%Male)
Grade
Setting Program Title
Program
Characteristics
Facilitator;
(Length)
Intervention
Measurement Design
IN ES
(95% CI)
EM ES
(95% CI)
Andreou
(2008)
J Greece 418
(60)
4
th
-6
th
Urban
Curriculum-
Based Anti-
Bullying
Awareness-building
(Awareness); Self-
reflection; Behavior
modification
(BehMod)
Teacher (1) Bystander
intention to
intervene
Q⫺.01
(⫺.20, .19)
⫺.19
(⫺.38, .01)
Evers (2007) J Multiple
US States
710
(41)
9
th
-12
th
Mixed
Build Respect BehMod;
Individualized
computer software;
Parent component
(Parent)
Computer
Based (2)
Passive
bystanding
Q .46
(.27, .64)
N.M.
Fonagy
(2009)
J Kansas 578
(46)
3
rd
-5
th
Rural
CAPSLE Psychiatric
consultation;
Awareness; Parent
Teacher (12) Helpful
bystanding
E .05
(⫺.11, .22)
⫺.23
(⫺.40, ⫺.07)
Frey (2005) J Washington 913
(51)
3
rd
-6
th
Suburban
Steps to
Respect
Awareness; Social-
emotional skill-
building; Parent
Teacher (6) Bystander
responsibility
E .11
(⫺.02, .24)
.18
(.05, .31)
Karna (2011) J Finland 8166
(50)
4
th
-6
th
Mixed
KiVa Awareness; Role-
playing; Modeling;
Parent
Teacher (12) Bystander
defending
E .14
(.10, .19)
.15
(.10, .19)
McLaughlin
(2009)
D Ohio 41
(41)
6
th
Urban Effective Bully
Prevention
Awareness; Modeling
with media
(ModMed)
Researcher
(1)
Bystander traits Q .21
(⫺.42, .83)
⫺.17
(⫺.81, .46)
Menesini
(2003)
J Italy 293
(53)
6
th
-8
th
Urban
Befriending
intervention
Awareness; Enhanced
capacity;
Responsibility-
training
Not Reported
(12)
Defender traits Q .03
(⫺.21, .26)
Not
Measured
(Table 1 continues)
School Psychology Review, 2012, Volume 41, No. 1
52
Table 1 Continued
Study (DoP) Type Location
N
(%Male)
Grade
Setting Program Title
Program
Characteristics
Facilitator;
(Length)
Intervention
Measurement Design
IN ES
(95% CI)
EM ES
(95% CI)
Merrell
(2004)
D New York 56
(30)
9
th
Urban
5 W’s
Approach to
Bullying
Awareness; ModMed Researcher
(2)
Bystander
intention to
intervene
E .60
(⫺1.17, 2.36)
Not
Measured
Schumacher
(2007)
M Pennsylvania 825
(42)
9
th
-12
th
Mixed
Bullying Video
Program
Awareness; ModMed Researcher
(1)
Bystander
intention to
intervene
E .43
(.29, .57)
.29
(.15, .43)
Stevens “A”
(2000)
J Belgium 301
(50*)
4
th
-8
th
Not
Reported
Anti-bullying
Intervention
Social-cognitive
training; ModMed
Teacher (1) Bystander
intention to
intervene
E .06
(⫺.17, .29)
⫺.05
(⫺.28, .18)
Stevens “B”
(2000)
J Belgium 401
(50*)
9
th
-11
th
Not
Reported
Anti-bullying
Intervention
Social-cognitive
training; ModMed
Teacher (1) Bystander
intention to
intervene
E .39
(.19, .59)
.14
(⫺.06, .33)
Whitaker
(2004)
B Texas 1763
(50)
5
th
Mixed
Expect Respect Awareness; Psycho-
education; Parent
Teacher (12) Bystander
intention to
intervene
E 0.25
(.15, .34)
Not
Measured
Notes: DoP ⫽Date of Publication; J ⫽Journal Article; D ⫽Dissertation; M ⫽Master’s Thesis; B ⫽Book Chapter; * ⫽imputed percent male; Parent ⫽Parent Training; Length ⫽
Time to Posttest in months; Q ⫽Quasi-experimental; E ⫽Experimental; T ⫽Treatment groups; C ⫽Control groups; IN ES ⫽Intervention Effect Size; EM ES ⫽Empathy Effect
Size.
A Meta-Analysis of School-Based Bullying
53
As shown in Table 1, 10 of the 83
unique articles met the inclusion criteria and
were included in the meta-analysis. We dis-
continued the literature search on May 20,
2010, but added 1 article brought to our atten-
tion 3 months later, which brought the total
number to 11 studies.
Coding
Study details, appropriate program, and
sample information were coded directly into
an EXCEL (2010) database. This included
publication year, publication type, funding
provided, country of origin and publication,
program location, treatment and control sam-
ple characteristics (e.g., age, gender, race,
SES, disabilities), program characteristics
(e.g., length of time, intervention details), and
program facilitator. In addition, intervention
and empathy outcome measures were tran-
scribed. By coding directly into the EXCEL
database, we eliminated errors that might have
occurred during the normal transcription phase
(Lipsey & Wilson, 2001).
The first author coded all 12 studies, but
one independent rater coded a randomly se-
lected portion of studies (5) for reliability pur-
poses. The two raters agreed 92% of the time.
The coders came to an agreement for all dis-
crepancies prior to completion.
Analysis
Quantitative synthesis, or meta-analysis,
is a statistical technique that combines related
research studies to estimate an overall treat-
ment effect (Cooper, Hedges, & Valentine,
2009; Glass, 1976; Hedges & Olkin, 1985).
Often, and in the case of the present review,
meta-analysis aggregates treatment effect
sizes to assess an intervention’s effectiveness.
The purpose of a meta-analysis, then, is to
generalize findings across multiple treatment
and setting types, participants, and times (Matt
& Cook, 2009). We conducted analyses using
SPSS (2010) and Comprehensive Meta-Anal-
ysis (Borenstein, Hedges, Higgins, & Roth-
stein, 2005) software.
Independent findings. A paramount
meta-analytic assumption is independence of
findings. Cooper (2010) discussed several oc-
currences that constitute nonindependence and
their effects on subsequent findings. Thus, we
conducted several common procedures to en-
sure independent findings. To ensure that only
one effect size was derived from each inter-
vention, we used only the first treatment out-
come reported for studies that reported multi-
ple post-treatment outcomes (see Andreou et
al., 2008). If studies implemented interven-
tions with two groups but only one control
group, we synthesized the treatment effects
prior to calculating the study effect size (see
McLaughlin, 2009). Finally, if one author im-
plemented an intervention and published mul-
tiple articles on the same sample, then we
reviewed only the first article published (see
Frey et al., 2005, 2009).
Effect size metrics. The majority of
effect sizes calculated used a continuous scale.
As such, the appropriate effect size metric was
the standardized mean difference (Equation
1):
d⫽XG1⫺XG2
Sp
(1)
where the numerator is the mean differ-
ence between treatment and control group
posttests, and the denominator is the pooled
standard deviation for the intervention and
comparison groups. Further, all dmetrics were
bias corrected using Hedges’s (1981) small
sample correction (g). This correction as well
as the sampling variance is represented by
Equations 2 and 3:
g⫽
冋
1⫺
冉
3
4N⫺9
冊册
*d(2)
Varg⫽
冑
nG1⫹nG2
nG1nG2
⫹g2
2共nG1⫹nG2兲(3)
where Nis the total sample, dis the
original standardized mean difference, and n
G1
School Psychology Review, 2012, Volume 41, No. 1
54
and n
G2
represent the treatment and control
group sample sizes, respectively.
In addition, we calculated logged odds
ratio effect sizes for two studies that used a
categorical outcome measure (Evers et al.,
2007; Merrell, 2004). Both measures were ob-
servations of treatment and control partici-
pants during free period times. The authors
observed how many times treatment children
intervened (or intended to intervene) in a bul-
lying situation compared to children in the
control group. Standard odds ratio calculations
were first used (Sanchez-Meca et al., 2003).
We then converted the logged odds ratio into
a standardized mean difference as outlined by
Lipsey & Wilson (2001) (Equations 4 and 5):
d⫽
冑
3*ln共OR兲
(4)
SE ⫽
冑
3*SEln共OR兲
2
3(5)
where ln(OR) represents the original
logged odds ratio and SE
ln(OR)
represents the
original sampling variance.
Missing data. Only one of the studies
failed to provide appropriate descriptive sta-
tistics. Schumacher (2007) provided only a t
statistic, as well as sample sizes, that com-
pared the treatment and control groups. Lipsey
and Wilson (2001) provided an appropriate
conversion (Equation 6):
ESsm ⫽t
冑
n1⫹n2
n1n2
(6)
where trepresents the tstatistic the
study provided, n
1
represents the sample size
of the treatment group, and n
2
represents the
sample size of the control group.
Random effects model. We assumed
that the treatments were derived from a ran-
dom sample of the literature but lacked a com-
mon effect size. Given this assumption, Bo-
renstein, Hedges, Higgins, and Rothstein
(2010) posited that the random effects model
was most appropriate. We further assumed
that an underlying distribution of effect sizes
was plausible; thus our goal, given the random
effects framework, was to estimate the distri-
bution’s mean and confidence interval.
To estimate a random effects mean and
confidence interval, we calculated the
weighted treatment effect of each study. The
weighted effect estimation synthesized both
within-study error variance and common be-
tween-study variance. We represented this
weight calculation as Equation 7:
Wi⫽1
Vi⫹T2(7)
where Wrepresents the ith study weight,
V
i
indicates the within-study error variance,
and T
2
represents the between-study variance.
We used these weights then to estimate the
combined treatment effect. This can be repre-
sented by Equation 8:
M⫽
冘
Wi*gi
冘
Wi
(8)
where Mrepresents the combined effect,
W
i
represents the ith study weight, and g
i
indicates Hedges’s effect size gfor the ith
study. Further, we calculated confidence inter-
vals and pvalues by taking the square root of
the inverse of the sum of the weights.
Moderator analysis. A critical next
step to the investigation of effect size distri-
bution is moderator analysis. We started by
calculating the homogeneity statistic Q. This
statistic provided information about the distri-
bution of effect sizes, and a large test statistic
(i.e., rejecting the null hypothesis of study
homogeneity) indicated that moderator analy-
ses were appropriate (Raudenbush, 2009).
Given this statistical confirmation, we used
procedures analogous to analysis of variance
(ANOVA), where one attempts to model ef-
fect-size heterogeneity associated with cate-
gorical study-level variables. Further, because
of the small number of studies that constituted
the review, we calculated the variance compo-
A Meta-Analysis of School-Based Bullying
55
nent across all groups rather than within
groups as is generally conducted (Hedges &
Vevea, 1998).
Lipsey (2009) discussed three types of
independent variables common to meta-ana-
lytic practice: extrinsic variables, method vari-
ables, and substantive variables. Extrinsic and
method variables relate to the study’s dissem-
ination (i.e., published or unpublished) or
methodological constraints (i.e., randomized
or nonrandomized). Substantive variables, on
the other hand, should be regarded as variables
of interest and generally include characteris-
tics of the population or treatment. For this
review, we coded substantive independent
variables to reflect the participant’s age, length
of treatment, and treatment type (e.g., individ-
ual or group).
Finally, we conducted a moderator anal-
ysis procedure analogous to regression,
weighted meta-regression. This statistical pro-
cedure allows for the simultaneous estimation
of study-level effects, but shares the problems
of typical regression (Cooper, 2010). For this
review, we modeled two independent vari-
ables simultaneously, treatment population
(categorical variable) and the percent of males
in the treatment group.
Sensitivity analysis. Two of the re-
viewed studies contributed a measure that
used an odds ratio. To ensure that study find-
ings were not biased by including these mea-
sures, we conducted a sensitivity analysis. The
analysis consisted of removing the 2 studies
(Evers et al., 2007; Merrell, 2004) and recal-
culating the weighted effect size. We hypoth-
esized that no difference would be found be-
tween the two types of measurements.
Publication bias. Publication bias re-
mained an important consideration during the
literature search and analyses. Rosenthal
(1979) introduced the “file-drawer problem,”
which stated that studies with small or nonsig-
nificant effect sizes tended to remain unpub-
lished. To combat this problem, we included
unpublished works from three dissertations
and one master’s thesis (Rothstein, Sutton, &
Borenstein, 2005).
To avoid and interpret the overestima-
tion of the random-effects estimate, we used
the nonparametric trim and fill procedure (Du-
val & Tweedie, 2000) to assess the sensitivity
of results to publication bias. This procedure
estimates the number of publications theoret-
ically missing because of funnel plot asymme-
try, and then recalculates the random-effects
mean and confidence interval to include the
imputed missing studies. We also reported
Rosenthal’s fail-safe N (1979) and Egger,
Davey Smith, Schnieder, & Minder’s (1997)
regression coefficient.
Results
Meta-Analysis Literature
Table 1 provides characteristics for each
study included in the review. The studies re-
viewed include 7 published journal articles, 1
book chapter, and 3 unpublished papers (two
dissertations and one master’s thesis). Seven
of the 11 studies were conducted within the
United States and the other 4 were conducted
in Belgium, Finland, Greece, and Italy, respec-
tively. All studies were completed between
2000 and 2010.
One article contributed two effect sizes
because it included two mutually exclusive
interventions from two separate populations
(Stevens et al., 2000). Therefore, we synthe-
sized a total of 12 interventions. Each of
the 12 interventions included a treatment and
control group; 4 of the 12 programs used qua-
si-experimental design and the other 8 used a
randomized experimental design. A total
of 12,874 students participated in the 12
interventions.
Outcome Effect Sizes
Bystander intervention outcome. As
delineated in Lipsey and Wilson (2001), we
estimated the random effects weighted mean
by using Equation 8. The results revealed a
statistically significant positive weighted aver-
age (g⫽.20, p⬍.001, 95% CI ⫽.11 to .29).
In other words, the treatment increased by-
stander intervention behavior 20% of one stan-
dard deviation more than individuals in the
School Psychology Review, 2012, Volume 41, No. 1
56
control group. Table 2 provided a forest plot of
the random-effects model’s relevant statistics.
Empathy outcome. Of the 12 inter-
ventions used to calculate the bystander inter-
vention outcome weighted average, 8 included
a measure on victim empathy. As previously
conducted, we used a random effects model to
estimate the weighted treatment mean. The
results revealed a very small, nonstatistically
significant result (g⫽.05, p⫽.38) with a
confidence interval that included zero (95%
CI ⫽⫺.07 to .17). The small number of
studies that included a measure of empathy
may not provide enough power to detect a
small effect. Therefore, the results of this anal-
ysis should be considered inconclusive (see
Table 3).
Moderator Analysis
We categorized the effect sizes into sev-
eral relevant groups and conducted the ran-
dom-effects ANOVA-like analysis (Table 4).
A statistically significant Qvalue indicated
appropriate heterogeneity between studies that
Table 3
Summary Statistics for Empathy Effect Sizes
Study Name Hedges’ g
Standard
Error
Lower
Limit
Upper
Limit Z-Value p-Value
Andreou ⫺.19 .10 ⫺.38 .01 ⫺1.90 .06
Fonagy ⫺.23 .09 ⫺.40 ⫺.07 ⫺2.75 .01
Frey .18 .07 .05 .31 2.72 .01
Karna .15 .02 .10 .19 6.62 .01
McLaughlin ⫺.17 .32 ⫺.80 .45 ⫺.05 .59
Schumacher .29 .07 .15 .43 4.13 .01
Stevens “A” ⫺.05 .12 ⫺.28 .18 ⫺.40 .69
Stevens “B” .14 .10 ⫺.06 .33 1.35 .18
Overall .05 .06 ⫺.07 .17 .74 .46
Table 2
Summary Statistics for Bystander Intervention Effect Sizes
Study Name Hedges’ g
Standard
Error
Lower
Limit
Upper
Limit Z-Value p-Value
Andreou ⫺.01 .10 ⫺.20 .19 ⫺.04 .97
Evers .46 .09 .27 .64 4.88 .01
Fonagy .05 .08 ⫺.11 .22 .62 .54
Frey .11 .07 ⫺.02 .24 1.68 .09
Karna .14 .02 .10 .19 6.51 .01
McLaughlin .21 .32 ⫺.42 .83 .65 .52
Mensini .03 .12 ⫺.21 .26 .22 .82
Merrell .60 .90 ⫺1.17 2.24 .66 .51
Schumacher .43 .07 .29 .57 6.08 .01
Stevens “A” .06 .12 ⫺.17 .29 .53 .60
Stevens “B” .39 .10 .19 .59 3.89 .01
Whitaker .25 .05 .15 .34 5.16 .01
Overall .20 .04 .11 .28 4.54 .01
A Meta-Analysis of School-Based Bullying
57
measured the bystander intervention construct
(Q⫽39.81, df ⫽11, p⬍.001; I
2
⫽72.36).
However, because of the relatively small num-
ber of studies that included an empathy mea-
sure, we chose not to conduct moderator anal-
yses to protect against findings of chance.
The results of these analyses were de-
composed into substantive and methodologi-
cal characteristics. Samples that consisted of
high school students only generated a signifi-
cantly greater treatment effect (ES ⫽0.43,
CI ⫽.33 to .52) compared to samples of
primary schools only (ES ⫽0.14, CI ⫽.11 to
.18). We further assessed sample differences
by conducting an analysis to evaluate location
differences. The results revealed that pro-
grams located in the United States (US) did
not differ significantly from those located in
Europe (EU; US ES ⫽0.26, EU ES ⫽0.13,
p⫽.17).
Another substantive moderator, treat-
ment length, failed to produce significantly
greater treatment effects (1–2 months
ES ⫽0.31, 6 –12 months ES ⫽0.16; p⫽.09).
Similarly, treatments that included a parental
component (e.g., parent guides, parent training
sessions) failed to influence the treatment sig-
nificantly compared to programs without the
component (parent included ES ⫽0.19, parent
excluded ES ⫽0.20, p⫽.92).
We also grouped the studies by who
facilitated the treatment programs. The teach-
Table 4
Moderator & Meta-regression Analysis of Bystander Intervention
Effect Sizes
Moderator K Hedges’ g95% C.I. p-value
Population
3
rd
–8
th
Grade 8 .14 .11, .18 .04ⴱ
9
th
–12
th
Grade 4 .43 .33, .52
Location
United States 7 .26 .14, .38 .17
Europe 5 .13 ⫺.01, .26
Treatment length
1–2 months 5 .31 .16, .45 .09
6–12 months 7 .16 .06, .25
Parent Component
Yes 5 .19 .08, .31 .92
No 7 .20 .07, .34
Facilitator
a
Teacher 7 .15 .09, .22 .01ⴱ
Other 4 .43 .30, .56
Assignment
Non-Random 4 .17 ⫺.01, .35 .74
Random 8 .21 .10, .31
Publication type
Peer-review 8 .16 .06, .25 .07
Non-peer review 4 .32 .17, .48
Meta-Regression S.E. Z-score p-value
Percent Male ⫺.01 .009 1.04 .15
HS (1 ⫽Yes) .25 .13 1.98 .02ⴱ
ⴱp⬍.05; a-Menesini et al. (2003) did not indicate the facilitator; HS ⫽High School sample.
School Psychology Review, 2012, Volume 41, No. 1
58
ers implemented a significant portion of the
programs for 7 of the 12 programs. Four other
programs were facilitated by the researcher, a
counselor, or in one case (Evers et al., 2007),
with computer software. One study failed to
indicate who implemented the program (Me-
nesini, Codecasa, Benelli, & Cowie, 2003).
The results of this moderator analysis revealed
significantly greater treatment effects for pro-
grams that implemented the program with fa-
cilitators other than the teacher (teacher
ES ⫽0.15, other ES ⫽0.43, p⬍.01). How-
ever, serious caution should be given to this
finding because two of the four programs that
used researchers as facilitators had the small-
est sample sizes, and therefore this could be a
reflection of imprecision or biased effects be-
cause of small samples (Levine, Asada, &
Carpenter, 2009).
Finally, we conducted moderator analy-
ses with two methodological groupings. We
first observed mean group differences between
randomly assigned (RA) and nonrandomly as-
signed treatments (NRA) groups. The results
indicated that there were no statistically sig-
nificant differences (RA ES ⫽0.21, NRA
ES ⫽0.17, p⫽.74). We also estimated group
differences between peer-reviewed (PR) and
nonpeer-reviewed studies (NPR). The results
of this calculation revealed that nonpeer-re-
viewed studies did not produce a greater treat-
ment effect (PR ES ⫽0.16, NPR ES ⫽0.32,
p⫽.07).
In addition to the ANOVA-like model-
ing, we conducted a weighted regression anal-
yses. This analysis allowed us to estimate the
effects of several predictors simultaneously.
We hypothesized that the percentage of males
in the treatment interventions and programs
conducted in a high school would be signifi-
cantly related to the treatment effect (i.e.,
Hedges’s g). The results of this analysis again
revealed that, after controlling for the percent-
age of males in the treatment group, high
school samples produced a greater treatment
effect compared to middle or elementary
school interventions (⫽0.25, Z⫽1.98,
p⬍.05). These results bolstered the previous
findings.
To ensure that the overall effect size was
not upwardly biased by including different
measurements, we conducted a sensitivity
analysis removing studies that used a dichot-
omous outcome. The results of the analysis
found that the overall weighted effect size
decreased slightly (original ES ⫽0.20, mod-
ified ES ⫽0.18). The overall effect remained
statistically and practically significant.
Publication Bias
We applied Duval and Tweedie’s (2000)
trim and fill procedure to address publication
bias. This procedure revealed that one nega-
tive result was missing from the bystander
intervention outcomes. However, the imputed
missing values would only slightly change the
overall fixed effect size (ES ⫽0.20); more
important, it remained practically and statisti-
cally significant (see Figure 1). We also used
Rosenthal’s fail-safe N procedure; the results
of this calculations indicated that 236 null
studies would be required to result in a non-
significant finding. Egger’s regression inter-
cept coefficient calculation also produced non-
significant results (
0
⫽0.57, p⫽.26). Taken
together, we concluded that the review’s re-
sults were not affected significantly by publi-
cation bias.
Discussion
The purpose of this quantitative synthe-
sis was to examine the treatment effects of
bullying prevention programs on bystander in-
tervention behavior. Empathy for the victim
was also synthesized as a secondary outcome,
but was not of primary purpose for the current
review. In total, we reviewed 11 studies (12
effect sizes) from the United States and Eu-
rope that included 12,874 children.
Using meta-analytic techniques, the re-
sults revealed that the intervention behavior of
bystanders increased (i.e., bystanders indi-
cated greater intervention behavior in bullying
situations) compared to control groups (g
⫽.20). The results of a secondary analysis
revealed that intervention programs did not
have a similar effect on empathy for the victim
(g⫽.05), but this finding should be viewed as
A Meta-Analysis of School-Based Bullying
59
inconclusive because of the small number of
studies that reported this outcome and its sec-
ondary nature.
These overall results mirrored the find-
ings of a previously conducted small synthe-
sis. Merrell et al. (2008) synthesized bullying
prevention programs to investigate the effects
of the programs on bullying perpetration, but
also included several secondary measures
(e.g., bystander intervention behavior, empa-
thy). The authors reported small but signifi-
cant treatment effects for bystander interven-
tion behavior (k⫽3, g⫽.17) and nonsignif-
icant negative effects with regard to empathy
(k⫽3, g⫽⫺.10). Taken together, these
replications results provided evidence against
mono-operation bias and thus greater validity
(Shadish, Cook, & Campbell, 2002).
Moderator analyses also revealed sev-
eral findings of interest. Results of both
ANOVA-like and weighted-regression analy-
ses revealed that the treatment effects were
greater for high school only samples. This is
somewhat surprising because some scholars
have postulated that bullying prevention pro-
grams are more effective for middle school-
aged children (Williford et al., 2011). These
results may indicate that bystander interven-
tion behavior is a developmental process and
programs may not influence younger students
as intended.
The purpose of meta-analytic research is
to generalize findings across populations,
treatments, outcomes, and designs (Matt &
Cook, 2009). Although this synthesis aggre-
gated a smaller number of studies, its findings
rendered generalizability of bullying preven-
tion program’s effects on the bystander inter-
vention construct. A few factors bolster this
belief.
The populations assessed varied across
ages, locations, and treatments. The largest
effects were found for high school only sam-
ples; however, an overall significant treatment
effect was also found for the total sample.
With regard to location, no significant differ-
ences were found between U.S. and European
samples. In addition, the studies employed a
wide variety of treatment programs that
proved efficacious. Moderator analyses also
suggested that study design, publication type,
and parental components produced similar
-2.0 -1.5 -1.0 -0 .5 0.0 0.5 1.0 1.5 2.0
0.0
0.2
0.4
0.6
0.8
1.0
Standar d Error
Funnel Plot of Standard Error by Hedges’s g
Hedges’s g
Figure 1. Trim and fill funnel plot of intervention effect sizes. This figure
illustrates each effect size relative to its standard error; the shaded dot indicates
an imputed effect size.
School Psychology Review, 2012, Volume 41, No. 1
60
findings across studies. These results consti-
tuted a test of effects holding “across pre-
sumed irrelevancies” (Shadish et al., 2002, p.
455) and increases the findings’ external va-
lidity. Further investigation and evaluation is
certainly required, but these results should
cautiously suggest program generalizability.
Limitations
Several limitations should be noted.
First, this meta-analysis included only 11 stud-
ies and 12 effect sizes. Although we took
precautions to ensure unbiased effect sizes and
findings, a great deal of caution should be used
when interpreting the findings. The findings
from a small collection of studies, no matter
the statistical technique or number of students
surveyed, should not enact immediate policy
and practical changes. This becomes espe-
cially clear when one considers the effect of
studies across time. Recent research in the
field of meta-analysis publication bias has in-
dicated that as programs increase in size and
fidelity, effect sizes tend to decrease (Trikali-
nos & Ioannidis, 2005). Therefore, we plan to
update these results periodically to observe the
effect of time.
Second, because of the nature of meta-
analyses, causal inferences should be stated
cautiously. Quantitative meta-analysis, al-
though statistically sophisticated and impor-
tant, remains essentially an observational
study (Lipsey & Wilson, 2001; Cooper &
Hedges, 2009). On the other hand, the sample
of studies we synthesized contained only those
that used a treatment and control group, and
these research designs constitute the most ef-
ficient measure of treatment effect (Shadish et
al., 2002). Therefore, the synthesis of these
results should partially reflect the nature of the
primary studies.
Third, because of the low number of
studies per factor, the ANOVA-like moderator
analyses generated relatively low statistical
power and the results should be interpreted
with caution (Hedges & Pigott, 2004). More-
over, the low number of studies per group
could easily capitalize on chance, or the
groupings may reflect some other unforeseen
variable. This becomes especially clear with
regard to the studies grouped by treatment
length. As alluded to previously, a majority of
the studies grouped in the 1–2 month category
were the smallest of studies. Smaller studies
tend to produce larger and more unstable ef-
fect sizes (Levine et al., 2009), and therefore
the grouping could reflect this phenomenon.
Fourth, although we made efforts to col-
lect all primary studies that focused on by-
stander intervention behavior, it is quite pos-
sible that studies failed to be included. New
material published postsearch, misspecified
search terms, or simple human error could
all cause inadvertent omission of extant lit-
erature. As such, we must temper our infer-
ences with regard to the extrapolation of this
information.
Implications for Future Policy and
Practice
This meta-analysis should help cau-
tiously to shift the emphasis of policy and
practice. The results of this meta-analysis re-
vealed two implications for policy. First, state
and national bullying legislation should imple-
ment and evaluate programs that address bul-
lying behaviors as a group process. Prevention
frameworks and programs that attempt to
abate bullying within schools are increasingly
emphasizing changes in school climate that
desist reinforcing bystander behavior or bul-
lying perpetration (Cohen, 2006). The results
of this study support these efforts to raise
awareness about the participant roles, to en-
courage active and prosocial behavior, and to
provide opportunities to role-play and practice
bystander intervention in vivo.
Second, the results of this meta-analysis
revealed that bullying prevention programs
might be effective at encouraging presocial
bystander intervention when the framework,
program, and/or curriculum explicitly target
bystander attitudes and behaviors. It is simply
not sufficient to only define prosocial by-
stander behaviors, such as walk away, get
help, or stand up to those engaged in bullying.
Policy must encourage the adoption of pro-
grams and interventions that shift attitudes
A Meta-Analysis of School-Based Bullying
61
supportive of intervention (willingness to in-
tervene) and behaviors through a consistent
message about intervention and ample support
from adults and administrators.
Future Research
This research suggests important future
projects. First, primary research should focus
on designing programs, implementing change,
and measuring the bystander construct. As
mentioned previously, researchers should fo-
cus on changing the behavior of the bystander.
Second, further work is required to evaluate
the effects of bystander behavior on bullying.
This meta-analysis merely demonstrated that
explicitly stated bystander programs have the
ability to increase bystander intervention be-
havior. However, future research must con-
tinue to assess how bystanders implement
these processes and the direct effects on active
bullying. Third, the results from the empathy
review revealed inconclusive findings. Future
research is required to elucidate the effects of
prevention programs on this outcome. Finally,
continued quantitative syntheses that focus on
bystander behavior are required. Certainly
only 11 studies from the last 10 years cannot
accurately describe the scope of this issue.
Future meta-analyses should incorporate,
hopefully, new relevant literature.
Footnote
*Article used in meta-analysis.
References
*Andreou, E., Didaskalou, E., & Vlachou, A. (2008).
Outcomes of a curriculum-based anti-bullying inter-
vention program on students’ attitudes and behavior.
Emotional & Behavioural Difficulties, 13, 235–248.
Astor, R. A., Meyer, H. A., Benbenishty, R., Marachi, R.,
& Rosemond, M. (2005). School safety interventions:
Best practices and programs. Children and Schools,
27(1), 17.
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Roth-
stein, H. R. (2005). Comprehensive meta-analysis,
Version 2. Englewood, NJ: Biostat.
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Roth-
stein, H. R. (2010). A basic introduction to fixed-effect
and random-effects models for meta-analysis. Re-
search Synthesis Methods, 1, 97–111.
Bosworth, K., Espelage, D., & Simon, T. (1999). Factors
associated with bullying behavior in middle school
students. The Journal of Early Adolescence, 19,
341–352.
Cohen, J. (2006). Social, emotional, ethical, and academic
education: Creating a climate for learning, participa-
tion in democracy, and well-being. Harvard Educa-
tional Review, 76, 201–237.
Cooper, H. (2010). Research synthesis and meta-analysis
(4th ed.). Thousand Oaks, CA: Sage Publications.
Cooper, H., & Hedges, L.V. (2009). Potentials and limi-
tations. In H. Cooper, L.V. Hedges, & J. C. Valentine
(Eds.), The handbook of research synthesis and meta-
analysis (pp. 562–571). New York: Sage Publication.
Cooper, H., Hedges, L. V., & Valentine, J. C. (2009). The
handbook of research synthesis (2nd ed.). New York:
Russell Sage Foundation.
Cowie, H. (2000). Bystanding or standing by: Gender
issues in coping with bullying in English schools.
Aggressive Behavior, 26, 85–97.
Cowie, H., & Hutson, N. (2006). Peer support: A strategy
to help bystanders challenge school bullying. Pastoral
Care in Education, 23(2), 40 – 44.
Craig, W. M. (1994). The relationship among bullying,
victimization, depression, anxiety, and aggression in
elementary school children. Personal Individual Dif-
ferences, 24(1), 123.
Craig, W. M., Pepler, D., & Atlas, R. (2000). Observa-
tions of bullying in the playground and in the class-
room. School Psychology International, 21(2), 22–36.
Duval, S., & Tweedie, R. (2000). A nonparametric “trim
and fill” method of accounting for publication bias in
meta-analysis. Journal of the American Statistical As-
sociation, 95(449), 89 –98.
Egger, M., Davey Smith, G., Schnieder, M., & Minder, C.
(1997). Bias in meta-analysis detected by a simple,
graphical test. British Medical Journal, 315, 629 – 634.
Espelage, D. L., Bosworth, K., & Simon, T. R. (2001).
Short-term stability and prospective correlates of bul-
lying in middle-school students: An examination of
potential demographic, psychosocial, and environmen-
tal factors. Violence and Victims, 16, 411– 426.
Espelage, D. L., & Swearer, S. M. (2003). Research on
school bullying and victimization: What have we
learned and where do we go from here? School Psy-
chology Review, 32, 365–385.
Espelage, D. L., Green, H. D., & Polanin, J. R. (in press).
Willingness to intervene in bullying episodes among
middle school students: Individual and peer-group in-
fluences. Journal of Early Adolescence.
*Evers, K., Prochaska, J., Van Marter, D., Johnson, J., &
Prochaska, J. (2007). Transtheoretical-based bullying
prevention effectiveness trials in middle schools and
high schools. [references]. Educational Research, 49,
397– 414.
Ferguson, C. J., Miguel, C. S., Kilburn, J. C., & Sanchez,
P. (2007). The effectiveness of school-based anti-bul-
lying programs: A meta-analytic review. Criminal Jus-
tice Review, 32, 401– 414.
*Fonagy, P., Twemlow, S. W., Vernberg, E. M., Nelson,
J. M., Dill, E. J., Little, T. D., et al. (2009). A cluster
randomized controlled trial of child-focused psychiat-
ric consultation and a school systems-focused interven-
tion to reduce aggression. Journal of Child Psychology
and Psychiatry, 50, 607– 616.
Frey, K. S., Hirschstein, M. K., Edstrom, L. V., & Snell,
J. L. (2009). Observed reductions in school bullying,
nonbullying aggression, and destructive bystander be-
havior: A longitudinal evaluation. Journal of Educa-
tional Psychology, 101, 466 – 481.
School Psychology Review, 2012, Volume 41, No. 1
62
*Frey, K. S., Hirschstein, M. K., Snell, J. L., Edstrom, L.,
Van Schoiack, L., MacKenzie, E. P., & Broderick,
C. J. (2005). Reducing playground bullying and sup-
porting beliefs: An experimental trial of the steps to
respect program. Developmental Psychology, 41,
479 – 491.
Glass, G. V. (1976). Primary, secondary, and meta-anal-
ysis of research. Educational Researcher, 5(10), 3– 8.
Glew, G. M., Fan, M., Katon, W., Rivara, F. P., &
Kermic, M. A. (2005). Bullying, psychosocial adjust-
ment, and academic performance in elementary school.
Pediatric Adolescent Medicine, 159, 1026 –1031.
Gini, G., Albiero, P., Benelli, B., & Altoe`, G. (2007).
Does empathy predict adolescents’ bullying and de-
fending behavior? Aggressive Behavior, 33, 467– 476.
Hawkins, D. L., Pepler, D. J., Craig, W. M. (2001).
Naturalistic observations of peer interventions in bul-
lying. Social Development, 10, 543–555.
Hedges, L. V. (1981). Distribution theory for glass’s es-
timator of effect size and related estimators. Journal of
Educational Statistics, 6, 107–128.
Hedges, L.V., & Olkin, I. (1985). Statistical methods for
meta-analysis. Orlando, FL: Academic Press.
Hedges, L. V., & Pigott, T. D. (2004). The power of
statistical tests for moderators in meta-analysis. Psy-
chological Methods, 9, 426 – 444.
Hedges, L. V., & Vevea, J. L. (1998). Fixed- and random-
effects models in meta-analysis. Psychological Meth-
ods, 3, 486 –504.
Howell, D. C. (2010). Statistical methods for psychology
(7th ed.). Pacific Grove, CA: Duxbury.
Hymel, S., & Swearer, S. (2010). Bullying: An age-old
problem that needs new solutions. Education.com. Re-
trieved from http://www.education.com/reference/
article/bullying-about-power-and-abuse-of-power/
Kaltiala-Heino, R., Rimpela, M., Marttunen, M., Rimpela,
A., & Rantanen, P. (1999). Bullying, depression, and
suicidal ideation in Finnish adolescents: School sur-
vey. British Medical Journal, 319, 348 –351.
*Karna, A., Voeten, M., Little, T. D., Poskiparta, E., Kal-
jonen, A., & Salmivalli, C. (2011). A large-scale evalu-
ation of the KiVa antibullying program: Grades 4 – 6.
Child Development, 82, 311–330.
Kochenderfer, B. J., & Ladd, G. W. (1996). Peer victim-
ization: Manifestations and relations to school adjust-
ment in kindergarten. Journal of School Psychology,
34, 267–283.
Lagerspetz, K. J., Bjorkqvist, K., Bertz, M., & King, E.
(1982). Group aggression among school children in
three schools. Scandinavian Journal of Psychology,
23, 45–53.
Levine, T., Asada, K. J., & Carpenter, C. (2009). Sample
sizes and effect sizes are negatively correlated in meta-
analyses: Evidence and implications of a publication
bias against nonsignificant findings. Communication
Monographs, 76, 286 –302.
Lipsey, M. W. (2009). Identifying interesting variables
and analysis opportunities. In H. Cooper, L. V.
Hedges, & J. C. Valentine (Eds.), The handbook of
research synthesis and meta-analysis (pp. 147–158).
New York: Sage Publication.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-
analysis (2nd ed.). Thousand Oaks, CA: Sage
Publications.
Matt, G. E., & Cook, T. D. (2009). Threats to the validity
of generalized inferences. In H. Cooper, L. V. Hedges,
& J. C. Valentine (Eds.), The handbook of research
synthesis and meta-analysis (pp. 538 –557). New York:
Sage Publication.
McDougal, P., Hymel, S., & Vaillancourt, T. (2009).
What happens over time to those who bully and those
who are victimized? Education.com. Retrieved from
http://www.education.com/reference/article/Ref_
What_Happens_Over/
*McLaughlin, L. P. (2009). The effect of cognitive behav-
ioral therapy and cognitive behavioral therapy plus
media on the reduction of bullying and victimization
and the increase of empathy and bystander response in
a bully prevention program for urban sixth-grade stu-
dents. Unpublished doctoral dissertation, University of
Toledo, Digital Dissertations.
*Menesini, E., Codecasa, E., Benelli, B., & Cowie, H.
(2003). Enhancing children’s responsibility to take ac-
tion against bullying. Aggressive Behavior, 29, 1–14.
*Merrell, R. (2004). The impact of drama intervention
program on response of the bystander to bullying
situations. Unpublished doctoral dissertation, Univer-
sity of Rochester, Digital Dissertations.
Merrell, K. W., Gueldner, B. A., Ross, S. W., & Isava,
D. M. (2008). How effective are school bullying inter-
vention programs? A meta-analysis of intervention re-
search. School Psychology Quarterly,23, 26 – 42.
Mishna, F., Cook, C., Saini, M., Wu, M. J., & MacFadden,
R. (2010). Interventions to prevent and reduce cyber
abuse of youth. Research on Social Work Practice,
1(1), 1–10.
Musher-Eizeman, D. R., Boxer, P., Danner, S., Dubow,
E. F., Goldstein, S. E., & Herretick, D. M. I. (2004).
Social-cognitive mediators of the relation of environ-
mental and emotion regulation factors to children’s
aggression. Aggressive Behavior, 30, 389 – 408.
Nansel, T. R., Overpeck, M., Pilla, R. S., Ruan, W. J.,
Simons-Morton, B., & Scheidt, P. (2001). Bullying
behaviors among US youth: Prevalence and associa-
tion with psychosocial adjustment. Journal of Ameri-
can Medical Association, 285, 2094 –2100.
Nickerson, A. B., Mele, D., & Princiotta, D. (2008).
Attachment and empathy as predictors of roles as de-
fenders or outsiders in bullying interactions. Journal of
School Psychology, 46, 687–703.
O’Connell, P., Pepler, D., & Craig, W. (1999). Peer in-
volvement in bullying: Insights and challenges for
intervention. Journal of Adolescence,22, 437– 452.
Olweus, D. (1973). Personality and aggression. In J. K.
Cole & D. D. Jensen (Eds.), Nebraska Symposium on
Motivation (pp. 260 –321). Lincoln: University of Ne-
braska Press.
Olweus, D. (2002). Bullying at school. Malden, MA:
Blackwell Publishers.
Raudenbush, S. W. (2009). Analyzing effect sizes: Ran-
dom-effects models. In H. Cooper, L. V. Hedges, &
J. C. Valentine (Eds.), The handbook of research syn-
thesis and meta-analysis (pp. 295–314). New York:
Sage Publication.
Rigby, K. (1999). Peer victimization at school and the
health of secondary school students. The Journal of
School Psychology, 69, 95–104.
Rigby, K., & Slee, P. T. (1993). Dimensions of interper-
sonal relation among Australian children and implica-
A Meta-Analysis of School-Based Bullying
63
tions for psychological well-being. British Journal of
Educational Psychology, 133, 33– 42.
Rivers, I., Poteat, V. P., Noret, N., & Ashurst, N. (2009).
Observing bullying at school: The mental health im-
plications of witness status. School Psychology Quar-
terly, 24, 211–223.
Rosenthal, R. (1979). The “file drawer problem” and
tolerance for null results. Psychological Bulletin, 86,
638 – 641.
Ross, S. W., & Horner, R. H. (2009). Bullying prevention
in positive behavior support. Journal of Applied Be-
havior Analysis, 42, 747–759.
Rothstein, H. R., Sutton, A. J., & Borenstein, M. (2005).
Publication bias in meta-analysis. West Sussex, Eng-
land: John Wiley & Sons Ltd.
Ryan, W., & Smith, D. J. (2009). Antibullying programs
in schools: How effective are evaluation practices?
Prevention Science, 10, 248 –262.
Salmivalli, C., Lagerspetz, K., Bjorkqvist, K., Osterman,
K., & Kaukianen, A. (1996). Bullying as a group
process: Participant roles and their relations to social
status within the group. Aggressive Behavior, 22, 1–12.
Salmivalli, C., Kaukianinen, A., & Voeten, M. (2005).
Anti-bullying intervention: Implementation and out-
come. British Journal of Educational Psychology, 75,
465.
Sanchez-Meca, J., Marin-Martinez, F., & Chacon-Mos-
coso, S. (2003). Effect-size indices for dichotomized
outcomes in meta-analysis. Psychological Methods, 8,
448 – 462.
Scholte, R. H. J., Engels, R. C. M. E., Overbeek, G.,
Kemp, R. A. T., & Haselager, G. J. T. (2007). Stability
in bullying and victimization and its association with
social adjustment in childhood and adolescence. Jour-
nal of Abnormal Child Psychology, 35, 217–228.
*Schumacher, P. J. (2007). To what extent will a sixty-
second video on bullying produced by high school
students increase students’ awareness of bullying and
change their attitudes to reduce acceptance of bullying
in a high school environment? Unpublished master’s
thesis, University of Pennsylvania, Digital
Dissertations.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002).
Experimental and quasi-experimental designs. New
York: Houghton Mifflin.
Sharp, S., Sellors, A., & Cowie, H. (1994). Time to listen:
Setting up a peer counseling service to help tackle the
problem of bullying in schools. Pastoral Care Educa-
tion, 12, 3– 6.
Smith, J. D., Twemlow, S. W., & Hoover, D. W. (1999).
Bullies, victims, and bystanders: A method of in-
school intervention and possible parental contribu-
tions. Child Psychiatry and Human Development, 30,
29 –37.
Smith, J. D., Schneider, B. H., Smith, P. K., & Ananiadou,
K. (2004). The effectiveness of whole-school antibul-
lying programs: A synthesis of evaluation research.
School Psychology Review,33, 547–560.
Sourander, A., Helstela, L., Helenius, H., & Piha, J.
(2000). Persistence of bullying from childhood to ad-
olescence—A longitudinal 8-year follow-up study.
Child Abuse & Neglect, 24, 873– 881.
SPSS. (2010). Statistical package for social sciences,
Version 19.0.0. Chicago, IL: IBM.
Sutton, A. J. (2005). Evidence concerning the conse-
quences of publication and related biases. In H. R.
Rothstein, A. J. Sutton, & M. Borenstein (Eds.), Pub-
lication bias in meta-analysis (pp. 175–192). West
Sussex, England: Wiley & Sons.
*Stevens, V., Oost, P. V., & Bourdeaudhuij, I. D. (2000).
The effects of an anti-bullying intervention programme
on peers’ attitudes and behaviour. Journal of Adoles-
cence, 23, 21–34.
Swearer, S. M., Espelage, D. L., Vaillancourt, T., &
Hymel, S. (2010). What can be done about school
bullying? Linking research to educational practice. Ed-
ucational Researcher, 39, 38 – 47.
Sweeting, H., Young, R., West, P., & Der, G. (2006). Peer
victimization and depression in early-mid adolescence:
A longitudinal study. British Journal of Educational
Psychology, 76, 577–594.
Trikalinos, T. A., & Ioannidis, J. P. A. (2005). Assessing
the evolution of effect sizes over time. In H. R. Roth-
stein, A. J. Sutton, & M. Borenstein (Eds.), Publication
bias in meta-analysis. West Sussex, England: John
Wiley & Sons Ltd.
Ttofi, M. M., & Farrington, D. P. (2009). What works in
preventing bullying: Effective elements of anti-bully-
ing programmes. Journal of Aggression, 1, 13–24.
Ttofi, M. M., & Farrington, D. P. (2011). Effectiveness of
school-based programs to reduce bullying: A system-
atic and meta-analytic review. Journal of Experimental
Criminology, 7, 27–56.
Twemlow, S. W., Fonagy, P., & Sacco, F. C. (2004). The
role of bystander in social architecture of bullying and
violence in schools and communities. Annals of the
New York Academy of Sciences,1036, 215–232.
Unnever, J. D., & Cornell, D. G. (2003). The culture of
bullying in middle school. Journal of School Violence,
2, 5–27.
Wang, J., Iannotti, R. J., & Nansel, T. R. (2009). School
bullying among adolescents in the United States: Phys-
ical, verbal, relational, cyber. Journal of Adolescent
Health, 45, 368 –375.
*Whitaker, D. J., Rosenbluth, B., Valle, L. A., & Sanchez,
E. (2004). Expect respect: A school-based intervention
to promote awareness and effective responses to bul-
lying and sexual harassment. In D. L. Espelage & S. M.
Swearer (Eds.), Bullying in American schools (1st ed.,
pp. 327–350). Mahwah, NJ: Lawrence Erlbaum
Associates.
Williford, A., Boulton, A., Noland, B., Little, T. D.,
Karna, A., & Salmivalli, C. (2011). Effects of the KiVa
anti-bullying program on adolescents’ depression, anx-
iety, and perception of peers. Journal of Abnormal
Child Psychology. Retrieved from http://www.spring-
erlink.com/content/ 2532380182216828/
Date Received: February 1, 2011
Date Accepted: September 7, 2011
Action Editor: Joseph Betts 䡲
School Psychology Review, 2012, Volume 41, No. 1
64
Joshua R. Polanin is a research methodology doctoral student at Loyola University
Chicago. His interests include methodological improvements to meta-analysis and hier-
archical linear modeling. He currently serves as the methodologist for a large, school-
based bullying prevention program and as the managing editor of the Campbell Collab-
oration’s Methods Group.
Dorothy L. Espelage is a professor in the Department of Educational Psychology at the
University of Illinois, Urbana-Champaign. Her research programs include investigations
of bullying, sexual harassment, and dating violence among adolescents for almost two
decades. She is engaged in a large randomized clinical trial of a school-based bullying
prevention program.
Therese D. Pigott is a professor at the School of Education, Loyola University Chicago.
Her research interests are statistical methods for meta-analysis and methods for handling
missing data in statistical analysis. She is a co-chair and editor of the Methods Group of
the Campbell Collaboration, an international organization that produces systematic re-
views of social interventions.
A Meta-Analysis of School-Based Bullying
65
Copyright of School Psychology Review is the property of National Association of School Psychologists and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.
