Bullying Among Young Adolescents: The Strong, the Weak,
and the Troubled
Jaana Juvonen, PhD*; Sandra Graham, PhD‡; and Mark A. Schuster, MD, PhD§储¶
ABSTRACT. Objectives. Bullying and being bullied
have been recognized as health problems for children
because of their association with adjustment problems,
including poor mental health and more extreme violent
behavior. It is therefore important to understand how
bullying and being bullied affect the well-being and
adaptive functioning of youth. We sought to use multiple
data sources to better understand the psychological and
social problems exhibited by bullies, victims, and bully-
victims.
Design, Setting, and Participants. Analysis of data
from a community sample of 1985 mostly Latino and
black 6th graders from 11 schools in predominantly low
socioeconomic status urban communities (with a 79%
response rate).
Main Outcome Measures. Peer reports of who bullies
and who is victimized, self-reports of psychological dis-
tress, and peer and teacher reports of a range of adjust-
ment problems.
Results. Twenty-two percent of the sample was clas-
sified as involved in bullying as perpetrators (7%), vic-
tims (9%), or both (6%). Compared with other students,
these groups displayed school problems and difficulties
getting along with classmates. Despite increased conduct
problems, bullies were psychologically strongest and en-
joyed high social standing among their classmates. In
contrast, victims were emotionally distressed and so-
cially marginalized among their classmates. Bully-vic-
tims were the most troubled group, displaying the high-
est level of conduct, school, and peer relationship
problems.
Conclusions. To be able to intervene with bullying, it
is important to recognize the unique problems of bullies,
victims, and bully-victims. In addition to addressing
these issues directly with their patients, pediatricians can
recommend school-wide antibullying approaches that
aim to change peer dynamics that support and maintain
bullying. Pediatrics 2003;112:1231–1237; bullying, youth,
victims.
B
ullying and being victimized by bullies have
been recognized recently as health problems
for school children because of their association
with a range of adjustment problems, including poor
mental health and violent behavior.
1
Large studies
suggest that 20% to 30% of students are frequently
involved in bullying as perpetrators and/or vic-
tims.
2–6
For example, in a nationally representative
study of 6th- to 10th-grade US students (n ⫽ 15 686),
13% were identified as bullies, 11% as victims, and
6% as bully-victims (ie, bullies who also get bullied
by others).
5
Bullying includes a range of behaviors that result
in an imbalance of power between the aggressor and
the victim.
7
Such behaviors include not only physical
aggression but also verbal harassment and public
humiliation (eg, name-calling and spreading ru-
mors). Indeed, emotional bullying by peers is espe-
cially a concern for youth,
8
and recent school shoot
-
ings suggest that it is not physical abuse by peers but
inability to cope with social ridicule and personal
rejection that can fuel extreme outbursts of violence.
9
Large-scale US studies on bullying and victimiza-
tion have relied mainly on self-reports when identi-
fying bully/victim groups and assessing their adjust-
ment problems, and none that we know of have
combined assessments from the perspective of
youth, their peers, and their teachers. Studies that
rely solely on self-reports suggest that bullies, vic-
tims, and bully-victims all share psychosocial adjust-
ment difficulties such as depression and psychoso-
matic problems.
2–4
Despite showing a rather uniform
picture of these 3 groups in some studies, other in-
vestigations have found differences: compared with
victims, bullies are more likely to manifest defiant
behavior
4–6,9
and negative attitudes toward school
2,5
and use drugs.
5
Victims, in turn, report feeling more
insecure and lonely than bullies.
2,10
There is less
information on bully-victims, although this group
seems to display the most severe problems.
2,4,5
Un
-
derstanding the characteristics of these groups is
important for identification and intervention.
However, there is concern that youth involved in
bullying may not accurately self-identify and that
others, namely their classmates, would provide a
more accurate assessment of who is involved in bul-
lying, and their teachers can provide important in-
formation about their adjustment. Indeed, if only
some youth accurately self-identify as bullies or vic-
tims, reports of their mental health could mislead
efforts to develop interventions to address their
needs.
Therefore, we conducted a study that included all
3 perspectives: self-, peer, and teacher reports. Most
importantly, we relied on classmates to report which
From the Departments of *Psychology and ‡Education, 储Department of
Pediatrics, School of Medicine, and ¶Department of Health Services, School
of Public Health, University of California, Los Angeles, California; and
§Rand, Santa Monica, California.
Received for publication Apr 14, 2003; accepted Jul 14, 2003.
This work does not necessarily represent the opinions of the funding
organizations or the institutions with which the authors are affiliated.
Reprint requests to (J.J.) Department of Psychology, University of California
Los Angeles, Hilgard Ave, Los Angeles, CA 90095. E-mail: juvonen@
psych.ucla.edu
PEDIATRICS (ISSN 0031 4005). Copyright © 2003 by the American Acad-
emy of Pediatrics.
PEDIATRICS Vol. 112 No. 6 December 2003 1231
students were involved in bullying, because they
have ample opportunities to observe peers’ behavior
in situations where bullying is most likely to take
place (ie, when adult supervision is absent or mini-
mal). Specifically, we used a labor-intensive peer
nomination methodology that is rarely used in such
large studies: each student provides confidential re-
ports on which classmates bully others and which
are victims of bullying. Individual nominations are
then combined to determine the strength of reputa-
tions. Peer nominations have been found to be very
reliable over time and predictive of a variety of de-
velopmental outcomes.
11–13
For example, one study
showed that peer nominations for negative charac-
teristics (such as those describing bullying) in 3rd
grade predicted psychiatric problems 11 to 13 years
later better than any other traditional predictors (eg,
teacher ratings, self-reports, or achievement data).
14
We used our data to compare self-reported psy-
chological distress, peer-reported social adaptation,
and teacher-rated adjustment of bullies, victims, bul-
ly-victims, and youth uninvolved in bullying. We
also focused on a demographic population of stu-
dents who are considered at risk for violence: youth
in predominantly urban, low-income communities
with a high representation of Latino and black
youth.
15
With one exception, large-scale studies have
not included these 2 racial/ethnic groups.
5
To our
knowledge, the current study is the largest investi-
gation on bullying and victimization among ethically
diverse urban adolescents.
METHODS
Study Design
We conducted a study of bullying among 6th-grade students in
11 public middle schools in the greater metropolitan area of Los
Angeles. All the schools were in low-income communities and
qualified for Title I compensatory education funds. Student eligi-
bility for free or reduced-price lunch programs ranged from 47%
to 87%. Three of the schools consisted primarily of black students,
3 were mostly Latino, and 5 had no majority group.
The study does not include students in special education, lim-
ited English proficiency, and gifted-student programs. Parental
permission was required; consent forms were sent home with
students in English and Spanish. The return rate for all schools
was 78%; 90% of returned forms granted permission. Students
provided assent. As part of the assent procedure, students were
assured confidentiality.
The final sample consisted of 1985 students (mean age: 11.5
years, 46% male). The racial/ethnic distribution was 45% Latino,
26% black, 10% white, 11% Asian, and 8% other.
Data Collection
Data were collected from 11 schools in 4 school districts split
over the fall of 2000 and the fall of 2001. Research team members
conducted the data collection. Self-administered student surveys
took ⬃1 hour to complete. Teachers who had daily contact with a
class rated students’ social behavior and academic engagement.
The survey and procedures were approved by the University of
California Los Angeles Human Subjects Protection Committee
and the school districts.
Measures and Variables
Bullies and Victims
We classified youth involved in bullying with categories devel-
oped in prior studies.
2,3,5
We used peer nominations whereby
students listed up to 4 classmates from a class roster who fit
descriptions for bullying (“starts fights and pushes other kids
around,”“puts down and makes fun of others,” and “spreads
nasty rumors about others”) and victimization (“gets pushed
around,”“is put down or made fun of,” and “about whom nasty
rumors are spread”). The 3 bullying nominations received were
strongly correlated (Cronbach’s
␣
⫽ 0.90) and were therefore
summed for each student. The same was true for the 3 victimiza-
tion nominations (
␣
⫽ 0.87). The total nominations received for
each then were standardized within classrooms. We used cutoff
values from prior research to classify students into the bully/
victim groups.
16
Students who fell 0.5 standard deviations above
the sample mean on bully nominations and below the sample
mean on victim nominations were classified as “bullies”; students
whose victim nominations were 0.5 standard deviations above the
sample mean and whose bully nominations fell below the mean
were classified as “victims”; and students whose peer nomina-
tions for bullying and victimization were both 0.5 standard devi-
ations above the mean were identified as “bully-victims.” We
identified nonaggressive, nonvictimized youth (“uninvolved”)as
those falling below the sample mean on both bully and victim
nominations. The rest of the sample was classified as “borderline.”
As with any research that uses cutoff scores to categorize
participants into groups, it is important to determine if specific
cutoff values affect the findings. Therefore, we also conducted
sensitivity analyses using different cutoffs.
Self-Reported Psychological Distress
We used 3 indicators of self-reported psychological distress.
1) Depression was measured with the 10-item Children’s Depres-
sion Inventory Short Form.
17
For each of the 10 items, respondents
were asked to choose the option that best described how they had
been feeling during the past 2 weeks (eg, “I do most things right,”
“I do many things wrong,” or “I do everything wrong”). Item
scores (range: 0–2) were summed (
␣
⫽ 0.80). 2) Social anxiety was
assessed with a combination of 2 subscales (12 items,
␣
⫽ 0.82)
from the Social Anxiety Scale for Adolescents
18
measuring fear of
negative evaluation (eg, “I worry about what others think of me”)
and social avoidance (eg, “I’m afraid to invite others to do things
with me because they might say no”). Each of the 12 items were
rated on a 5-point scale from “not at all” to “all the time.” 3) A
16-item loneliness measure
19
(eg, “I feel alone” or “I have nobody
to talk to”) had a 5-point scale from “not true at all” to “always
true.” Scores were averaged (
␣
⫽ 0.85).
Peer Reports of Adjustment
Peer nominations were used to assess social adaptation within
the peer group. Respondents nominated up to 4 classmates they
considered the “coolest” kids in their class (indicating social status
or rank) and up to 4 they did not like to hang out with (indicating
avoidance). We included the question about social avoidance (also
called peer rejection), because high social rank does not preclude
classmates from avoiding a student. Nominations received were
summed for each student and standardized within classroom.
Teacher-Rated Adjustment
Teachers with daily classroom contact rated students on behav-
ior by using 11 interpersonal competence items with a 7-point
scale with item-specific anchors (eg, “never sad” to “always
sad”).
20
These items yielded 3 subscales with 3 items each: inter
-
nalizing problems (sad, worries, cries a lot;
␣
⫽ 0.61); conduct
problems (starts fights, argues, gets in trouble;
␣
⫽ 0.89); and
popularity (popular with boys, popular with girls, has lots of
friends;
␣
⫽ 0.79).
Teachers also rated school engagement with 6 items from the
Teacher Report of Engagement Questionnaire
21
(eg, “in my class
this student concentrates on doing his/her work”). Ratings were
on 4-point scales (“not at all characteristic of this student” to “very
characteristic”). Item scores were averaged (
␣
⫽ 0.89). For the
analyses, we reverse-coded this measure such that it indicates
problems similar to most of the other variables; we relabeled it as
disengagement from school.
Statistical Methods
The differences among the bully/victim groups regarding psy-
chosocial distress, social adjustment, and school engagement were
analyzed by using analyses of variance with SPSS. Statistically
1232 BULLYING AMONG YOUNG ADOLESCENTS
significant group differences (P ⬍ .01) were followed up with
pairwise comparisons (Tukey test, significant difference: P ⬍ .05).
To facilitate the interpretation of group differences in measures
from multiple informants and with different response scales, all
the variables were converted into standard scores with a mean of
0 and standard deviation of 1. As standardized scores, the means
portrayed in Figs 1, 2, and 3 can also be interpreted as percentiles:
Values of 0 are at the 50th percentile; positive scores are above the
50th percentile; and negative scores are below the 50th percentile.
Scores of 1 represent approximately the 85th percentile. The scales
for y axes vary across self-, peer, and teacher reports to best depict
the relative differences among the 5 groups within each data
source.
RESULTS
Classification of Bully/Victim Groups
Students were classified as bullies (7%), bully-vic-
tims (6%), victims (9%), borderline (22%), or unin-
volved (56%) (Table 1). Boys were twice as likely as
girls to be classified as bullies (10% vs 5%), ⬎3 times
as likely to be classified as bully-victims (10% vs 3%),
and almost twice as likely to be classified as victims
(12% vs 7%; Table 1).
Type of involvement also varied by race/ethnicity
(Table 1). Black and other youth were most likely and
Asian least likely to be classified as bullies (11% and
10% vs 3%, respecetively). Other and white were
more likely and Latino least likely to be classified as
victims (13% and 12% vs 7%, respectively). Black
youth were most likely to be classified as bully-
victims (10%).
Self-Reported Psychological Distress
Significant differences appeared on all psycholog-
ical adjustment indicators, with bullies reporting the
lowest and victims reporting the highest levels of
depression, social anxiety, and loneliness (Fig 1). Bul-
ly-victims generally fell in between, with elevated
levels of depression and loneliness but average levels
of social anxiety.
Peer-Rated Adjustment
Bullies were regarded as the highest and victims
the lowest in social status (Fig 2). However, class-
mates avoided both bullies and victims and espe-
cially bully-victims more than they avoided other
classmates.
Teacher-Rated School Engagement and Social
Adjustment
The 5 groups also showed differences in teacher
ratings of school engagement and social adjustment
(Fig 3). Consistent with peer nominations (Pearson
r ⫽ 0.37; P ⬍ .001 between teacher ratings of popu-
larity and peer nominations of social status), teacher
ratings indicated that bullies were most popular and
victims were least popular. Teachers also rated vic-
tims as displaying more internalizing problems (eg,
sadness or anxiety) than bullies or bully-victims (but
not more than the uninvolved). Teachers ranked bul-
ly-victims as manifesting by far the most and unin-
volved students the least conduct problems. All 3
groups of students involved with bullying, especially
bully-victims, were rated as more disengaged in
school than their classmates.
Multivariate Analyses
Multivariate analyses of covariance predicting
each of the measures of self-, peer-, and teacher-rated
adjustment by controlling for gender and race/eth-
nicity differences were conducted. The results were
essentially identical to the bivariate analyses.
Sensitivity Analyses
We conducted 2 sets of sensitivity analyses by
changing the 0.5 cutoff for the standard deviation
above the mean cutoff to 0.75 and 1.00, respectively.
Although these modifications necessarily identified
smaller groups more intensely involved in bullying,
the pattern of differences across the 5 groups was the
same for all analyses with no evidence of the findings
being dependent on the specific cutoff criterion.
DISCUSSION
Among youth involved in bullying in a commu-
nity sample of ethnically diverse middle school stu-
dents, we found that bullies manifest the fewest
number of adjustment problems. Specifically, bullies
are psychologically stronger than classmates not in-
volved in bullying, and they enjoy high social status
among their classmates (although the classmates
tend to avoid their company). Victims, on the other
hand, suffer not only emotional distress but also
social marginalization (ie, classmates avoid them,
and they have low social status). Finally, those who
both bully and get bullied (ie, bully-victims) are es-
pecially troubled. They are by far the most socially
ostracized by their peers, most likely to display con-
duct problems, and least engaged in school, and they
also report elevated levels of depression and loneli-
ness.
TABLE 1. Percentage of 6th Grade Students Identified as Bullies, Bully-Victims, Victims, Border-
line, and Uninvolved (by Demographic Characteristics)
Bully
(%)
Bully-Victims
(%)
Victim
(%)
Borderline
(%)
Uninvolved
(%)
Total (n ⫽ 1985) 7 6 9 22 56
Gender
Boys (n ⫽ 904) 10 10 12 24 44
Girls (n ⫽ 1081) 5 3 7 19 66
Race/ethnicity
Latino (n ⫽ 910) 6 5 7 21 61
Black (n ⫽ 511) 11 10 10 23 46
Asian (n ⫽ 212) 3 5 10 20 62
White (n ⫽ 188) 6 6 12 21 55
Other (n ⫽ 164) 10 4 13 22 51
ARTICLES 1233
Our findings about bullies are distinct from the
findings in some studies that have reported that
bullies tend to be depressed and have psychological
distress.
2–4
However, these studies have depended
on self-reports of being a bully, and it is unlikely that
bullies as a group provide accurate self-reports of
how they treat others. Instead, we used the well-
validated peer nomination method to collect reports
from classmates about the roles students play. This
approach provided identification of bullies (and vic-
tims) based on a consensus of a large group of stu-
dents who know and observe all students in a class.
Whereas prior studies had the unexpected finding of
a fairly uniform psychological and social picture of
bullies and victims,
2–4
we found that despite some
common characteristics (eg, school disengagement
problems), these groups are distinct, which has im-
plications for identifying them and intervening. For
example, although prior studies could be interpreted
as indicating a need to focus on depression in ad-
dressing the needs of bullies, this approach may be
useful only for the subset most likely to self-identify,
not for those who do not admit to bullying.
Although the use of peer nominations to classify
Fig 1. Group means on standardized scores of self-reported psychological distress. The y axis indicates standardized scores with the
mean of 0 (that represents approximately the 50th percentile of the sample) and a standard deviation of 1. Positive values indicate groups
means greater than the 50th percentile, whereas negative values indicate group means less than the 50th percentile of the sample. The
letter subscripts on the bottom of the figures indicate pairwise group similarities and differences. Groups sharing letter subscripts do not
differ significantly from one another, whereas those that do not share a subscript differ from one another (P ⬍ 0.05).
Fig 2. Group means on standardized scores of peer-reported social adjustment problems.
1234 BULLYING AMONG YOUNG ADOLESCENTS
youth into bully, victim, and bully-victim groups is a
strength of the study, there can be biases in the
manner respondents identify classmates as bullies or
victims. Who is reported as a bully or victim may be
determined not only by individuals’ behavior but
also by the biases of the observer.
22
For example, the
behaviors of 2 classmates acting in the same way
may be reported differently by other students de-
pending on the implicit stereotype associated with
the 2 students’ social category (eg, race or gender).
Future studies are needed to help us better under-
stand the conditions under which racial and gender
biases are most likely to emerge. How such biases
may influence adolescent behavior, psychological
well-being, and school functioning is another ques-
tion for future research.
The superior mental health (lack of psychological
stress) of bullies documented in the current study
can in part be understood in light of the social pres-
tige that they enjoy among their classmates. Devel-
opmental research shows that in early adolescence,
social status is one of the strongest predictors of
positive self-views and psychological well-being.
23,24
Hence, it is likely that teenaged bullies do not feel
depressed, anxious, or lonely because they have high
social status within their peer collective. It is never-
theless important to keep in mind that, despite bul-
lies’ high social status, classmates would rather not
spend time with them. Thus, it may be that the social
prestige of bullies is motivated in part by fear.
Regardless of the reasons behind the favorable
social ranking of bullies, the high status imposes a
challenge for addressing bullying problems. When
bullies are considered the “coolest,” bullying be-
havior is encouraged. This finding underscores the im-
portance of addressing bullying as a systemic problem
that involves the whole peer collective. Comprehen-
sive, school-wide antibullying programs
25,26
aim to
change peer dynamics that encourage and maintain
bullying by raising the awareness of how bystanders
contribute to the problem of bullying.
27
Perspective-
taking exercises (including videotaped vignettes of
typical incidents) depict the way in which bystanders
encourage bullies. These same exercises may be use-
ful for increasing empathy for victims such that stu-
dents become more conscious of the plight of the
victim.
28
The current findings indicate that groups of teens
who report elevated psychological distress are the
same ones who others report as having trouble fit-
ting in or being accepted by their peers. This finding
underscores the importance of also understanding
the social plight of victims: they are not only targeted
by bullies but also ostracized by many of their class-
mates. To address the adjustment difficulties of vic-
tims, it may be that psychological services that focus
solely on the emotional problems are inadequate;
efforts to deal with the social issues are also needed.
This means that clinicians working outside of the
school, who might be most qualified to provide such
individually focused psychological services, may
have limited capacity to intervene. The lack of con-
nections between clinicians, who see youth outside
of school, and teachers and other school staff, who
see youth in school, is a problem that complicates
interventions. One of the challenges with bullying
among school children therefore is to try to build and
maintain links between clinicians and school staff.
Compared with bullies and victims, the bully-vic-
tim group seems to have the worst of both worlds
and a unique risk profile. Their high levels of social
avoidance, conduct problems, and school difficulties
suggest that they are a particularly high-risk group.
2,5
Indeed, previous research indicates that bully-
victims are most vulnerable to both concurrent and
subsequent psychiatric disorders.
4
Victims who
bully others also best fit the profiles of seriously
violent offenders. For example, a recent in-depth case
analysis of 37 intended and conducted school shoot-
ings revealed that ⬃2/3 of the alleged perpetrators
Fig 3. Group means on standardized scores of teacher-rated social adjustment problems.
ARTICLES 1235
