School Psychology Quarterly

2015, Vol. 30, No. 1, 105122

2014 American Psychological Association

1045-3830/15/$12.00 http://dx.doi.org/10.1037/spq0000082

Examination of the Change in Latent Statuses in Bullying

Behaviors Across Time
Ji Hoon Ryoo

Cixin Wang

University of Virginia

University of California, Riverside

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Susan M. Swearer
University of Nebraska-Lincoln
Involvement in bullying and victimization has been mostly studied using crosssectional data from 1 time point. As such, much of our understanding of bullying and
victimization has not captured the dynamic experiences of youth over time. To examine
the change of latent statuses in bullying and victimization, we applied latent transition
analysis examining self-reported bullying involvement from 1,180 students in 5th
through 9th grades across 3 time points. We identified unobserved heterogeneous
subgroups (i.e., latent statuses) and investigated how students transition between the
unobserved subgroups over time. For victimization, 4 latent statuses were identified:
frequent victim (11.23%), occasional traditional victim (28.86%), occasional cyber and
traditional victim (10.34%), and infrequent victim (49.57%). For bullying behavior, 3
latent statuses were identified: frequent perpetrator (5.12%), occasional verbal/
relational perpetrator (26.04%), and infrequent perpetrator (68.84%). The characteristics of the transitions were examined. The multiple-group effects of gender, grade, and
first language learned on transitions across statuses were also investigated. The infrequent victim and infrequent perpetrator groups were the most stable, and the frequent
victim and frequent perpetrator groups were the least stable. These findings suggest
instability in perpetration and victimization over time, as well as significant changes,
especially during school transition years. Findings suggest that school-based interventions need to address the heterogeneity in perpetrator and victim experiences in
adolescence.
Keywords: latent transition analysis, bullying, victimization, middle and high schools
Supplemental materials: http://dx.doi.org/10.1037/spq0000082.supp

Bullying is a complex phenomenon, defined

by repeated aggressive behavior with the intent
to hurt others, as well as a perceived imbalance
of power between the bullies and the victims
(Olweus, 1994). It is characterized by disrup-

This article was published Online First August 11, 2014.

Ji Hoon Ryoo, Department of Educational Leadership,
Foundations and Policy, University of Virginia; Cixin
Wang, Graduate School of Education, University of California, Riverside; Susan M. Swearer, Department of Educational Psychology, University of Nebraska-Lincoln.
Correspondence concerning this article should be addressed
to Ji Hoon Ryoo, Department of Educational Leadership,
Foundations and Policy, University of Virginia, 417 Emmet
Street South, Charlottesville, VA 22904-4265. E-mail:
jr3gv@virginia.edu
105

tions in social, behavioral, and often, academic

functioning (Rueger & Jenkins, 2014; Swearer,
Siebecker, Johnsen-Frerichs, & Wang, 2010).
The bullying literature has exploded over the
past three decades; however, various methodological challenges plague the field. For example, different definitions and forms of bullying
(i.e., physical, verbal, relational, and cyber),
cut-off points, and time frames used to determine involvement, have made comparisons
across studies impossible. Meanwhile, understanding the changes in perpetration and victimization over time has been less studied. This
study sought to remedy this by examining fifth
through ninth grade students experiences with
bullying over three time points (i.e., semesters).
Specifically, we investigated the changes in sta-

106

RYOO, WANG, AND SWEARER

tuses with membership at each time point, and

individual characteristics that may moderate the
transition in bully/victim statuses over time.

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Prevalence of and Changes in Bullying and

Peer Victimization
Research suggests that the percentage of students engaging in bullying perpetration range
from 6.4% to 35% (Nansel et al., 2001; Swearer
et al., 2010), while 30% to 60% students reported being victimized, with weekly victimization ranging from 6% to 15% (Card & Hodges,
2008; Robers, Kemp, Truman, & Snyder,
2013). The wide range of prevalence rates are
likely due to the fact that researchers have used
different cut-off criteria to determine bullying
involvement (i.e., daily vs. weekly involvement,
or aggression/victimization scores falling 1 SD
above the mean; Schwartz, Proctor, & Chien,
2001; Swearer et al., 2010). Using arbitrary
cut-off criteria (e.g., 1 SD above the mean)
based on the score distribution of a particular
sample is problematic because it may lead to
different classification across studies. Furthermore, individuals who score close to the cut-off
points may be misclassified (Bettencourt, Farrell, Liu, & Sullivan, 2013). Person-oriented
approaches, such as latent class analysis (LCA)
or latent transition analyses (LTA) can remedy
this limitation by using response patterns of
observed variables to assign individuals to unobserved latent groups (Bye & Schechter, 1986;
Collins & Wugalter, 1992).
The difference in prevalence rates across
studies may also be due to researchers assessing
different types of bullying behaviors (i.e., physical, verbal, relational, and cyberbullying) independently or in combination (Swearer et al.,
2010). Although researchers generally agree
that physical, verbal, and relational bullying are
distinct constructs, the distinction between cyberbullying and traditional bullying (physical,
verbal, and relational bullying) is less clear.
There has been debate regarding whether cyberbullying is a unique phenomenon (Li, 2007;
Olenik-Shemesh, Heiman, & Eden, 2012).
Strong correlations between cyberbullying and
traditional bullying as well as cyber victimization and traditional victimization have been
documented (e.g., Li, 2007). However, few researchers have separated different subtypes of
traditional bullying when examining its rela-

tionship with cyberbullying. One study found

that students who engaged in cyberbullying belonged to a highly aggressive group who also
frequently engaged in other forms of bullying
(physical, verbal, and social; Wang, Iannotti, &
Luk, 2012). Further research is needed to advance our understanding of the relationship between cyberbullying/victimization and different
types of traditional bullying/victimization
(physical, verbal, and relational).
Only a few longitudinal studies have examined the changes in bullying and victimization
over time. Research has demonstrated an initial
increase in bullying after the transition into middle school, and then a general decreasing trend
afterward (Pellegrini & Bartini, 2000; Pellegrini
& Long, 2002). Studies have also shown a decrease in peer victimization over time (Nylund,
Bellmore, Nishina, & Graham, 2007; Smith,
Madsen, & Moody, 1999). When examining
different types of victimization over 3 years,
researchers found social victimization increased
significantly from seventh grade through ninth
grade only for girls and then decreased at tenth
grade; however, there were no significant
changes in overt victimization (both physical
and verbal victimization) between seventh
grade and tenth grade (Rosen, Beron, & Underwood, 2013). However, most extant studies
used variable-oriented approach (e.g., correlation or ANOVA) to examine stability/change in
aggression over time (Pellegrini & Long, 2002;
Strohmeier, Wagner, Spiel, & von Eye, 2010),
which fail to capture individuals transition
patterns. The correlation between time points
simply indicates the association between
those degrees but not the changes. LTA is a
person-oriented approach that classifies the
heterogeneous subgroups and traces the
changes in membership over time. It also
allows researchers to examine factors associated with group membership and transitions
across time (Bergman & Magnusson, 1997;
Bergman, Magnusson, & El-Khouri, 2003;
Collins & Lanza, 2010).
Fewer studies have examined the changes in
group membership/latent status over time using
LTA. One study followed fourth graders for 3
years and found four distinct groups: bully
(12%24%), victim (25%39%), bully/victim
(7%12%), and not involved (37% 40%), with
students depressive symptoms and antisocial
attitudes as significant predictors of status

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

change during the transition to middle school

(Williford, Boulton, & Jenson, 2014). Similarly,
one study with sixth graders over two time
points (2 years) also found four distinct classes:
nonvictimized aggressors (17%21%), aggressive-victims (21%24%), predominantly victimized (15%25%, the least stable group), and
well-adjusted youth (37%39%, the most stable
group; Bettencourt et al., 2013). Looking at
bullying perpetration specifically, one study followed fourth graders over a 3-year period and
found that three distinct groups of bullies existed, including one group (82%) with consistently low scores for bullying, one group (7%)
with high and consistent scores of bullying, and
one group (11%) with moderate but declining
levels of bullying behavior (Reijntjes et al.,
2013). To examine the stability of peer victimization during sixth to eighth grade, Nylund,
Bellmore, Nishina, and Graham (2007) found
three victim classes based on severity: frequently victimized, sometimes victimized, and
nonvictimized, and found adolescents tended to
move from a more frequently victimized class
into a less frequently victimized class over time.
Although above-mentioned studies have begun
to examine the changes in bully/victim statuses
using LTA, they only followed students in a
single grade over a short period of time (e.g., 2
to 3 years). Furthermore, to our knowledge,
researchers have not yet examined the effects of
grade and English language learner (ELL) status
on the transition between different bullying and
victimization groups. This study extends prior
research by using a sequential design with students from fifth to ninth grades over three semesters to examine the stability and transition in
class membership in bullying perpetration and
victimization as well as individual factors (gender, grade, and ELL status) that may impact the
transition. Because students involved in bullying
and victimization represent heterogeneous subgroups, using LTA to accurately identify different
subgroups of students involved in bullying, and
characteristics associated with each subgroup will
provide important information to guide bullying
prevention and intervention efforts.
Individual Characteristics Associated With
Bullying and Victimization
Research has shown that some individual
characteristics, such as gender, grade, immigra-

107

tion, or ELL status are associated with students

involvement in bullying perpetration and peer
victimization. However, how these characteristics impact the changes or transition in bully/
victim statuses is less clear.
Gender Differences
Research on gender differences in bullying
and aggression has yielded contradictory findings. Some studies have found that boys were
more likely to be involved in bullying than girls
(Cook, Williams, Guerra, Kim, & Sadek, 2010).
However, other studies have found that although boys engage in more physical and verbal
aggression than girls (e.g., Prinstein, Boergers,
& Vernberg, 2001), the gender differences in
indirect/relational aggression was close to zero
(d .06), with girls engaging in slightly higher
levels of indirect/relational aggression (Card,
Stucky, Sawalani, & Little, 2008).
When looking at different types of victimization, boys reported being overtly victimized
more than girls (Martin & Huebner, 2007; Prinstein et al., 2001). Girls have been found to
experience more relational victimization compared with boys (Dempsey, Fireman, & Wang,
2006; Rueger & Jenkins, 2014). However, the
absence of gender differences (Storch, MasiaWarner, Crisp, & Klein, 2005), and boys experiencing more relational victimization (Martin
& Huebner, 2007) have also been documented.
Regarding cyber victimization, the results are
also mixed, with some studies showing girls
experiencing more cyber victimization (e.g.,
Kowalski & Limber, 2007; Li, 2007), yet other
studies finding no gender difference (e.g., Jackson & Cohen, 2012). These differences might
be due to different data collection methods used
(self-report vs. peer/teacher report), different
age groups studied, and the use of different
definitions of bullying and victimization. Researchers have also suggested that in order to
better understand gender differences in bullying, it is important to move beyond the mean
level differences and to examine the process by
which bullying unfolds (Underwood & Rosen,
2011, p. 13). Thus, it is important to examine
how gender may influence the transition in bullying perpetration and victimization classes
over time. Limited research has suggested that
aggressive behavior is less stable among girls
than boys, while girls tend to transition out of

108

RYOO, WANG, AND SWEARER

serious aggression; boys involvement in aggression is more stable over time (Miller et al.,
2013).

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Grade Differences
Research has shown that bullying behaviors
tend to increase at the end of elementary school
to middle school years and then decrease during
the high school years (Olweus, 1993; Pepler et
al., 2006). Specifically, an increase in bullying
after the initial transition to middle school has
been documented, suggesting students may use
bullying as a way to gain social status with new
peers after school transitions (Pellegrini & Bartini, 2000; Pellegrini & Long, 2002). Furthermore, compared with younger students, older
students were more likely to be a bully, less
likely to be a victim (Cook et al., 2010), and
tended to exhibit lower rates of physical bullying and higher rates of relational bullying (Coie
& Dodge, 1998). Thus, as children mature, they
learn that there are greater negative consequences for physical bullying and fewer consequences for relational, verbal, and/or cyberbullying because they are more difficult to detect.
English Language Learner Differences
Immigrant students experience many challenges at school, including discrimination
(Shin, DAntonio, Son, Kim, & Park, 2011),
limited access to educational resources (Kozol,
2005), and increased risk for mental health difficulties (Sue, 1994). Considering that about
25% of children in U.S. have at least on foreignborn parent (U.S. Census Bureau, 2010), it is
important to examine how immigrant or ELL
status impacts students experiences with bullying and/or peer victimization. However, only a
few studies have examined the effect of ELL
status or immigration on bullying/victimization.
Some studies have shown that immigrants and
ELL students experience higher level of peer
victimization compared with native speakers,
possibly because of language barriers and cultural differences (von Grnigen, Perren, Ngele,
& Alsaker, 2010; Koo, Peguero, & Shekarkhar,
2012); however, other studies have not found
these disparities (Boulton, 1995; Strohmeier,
Krn, & Salmivalli, 2011). Furthermore, most
of those studies were conducted in European
countries and not in the United States.

The Present Study

To date, most studies have used predetermined cut-off criteria to define bully/victim
groups and have examined prevalence rates
across different age groups using crosssectional designs, which results in a lack of
understanding in the dynamic changes among
bully/victim statuses. In previous approaches
such as correlation studies using cut-off scores,
the degree of victimization and bullying were
measured with the assumption that there exists a
single homogeneous population for each construct that can be measured using a continuous
scale. The correlation between time points simply indicates the association between those degrees but not the changes. Previous research has
shown that using predetermined cut-off points
to define different victim groups may be problematic because it fails to take developmental
changes in peer victimization into consideration
(Nylund et al., 2007). The current study employed LTA and extends previous research by
using a sequential design with students from
fifth to ninth grades over three semesters to
estimate the prevalence and degree of transitioning between latent statuses. We identified
unobserved heterogeneous subgroups and observed how students transitioned between the
unobserved heterogeneous subgroups over time.
Using LTA, the current study was guided by the
following research questions and hypotheses:
1. Are there distinct subgroups of students
involved in bullying and victimization
who engage in particular patterns of behaviors? Based on the existing literature
(Nylund et al., 2007; Reijntjes et al.,
2013), we hypothesized that there are distinct victim classes and perpetrator classes
based on the frequency of involvement
instead of the type, for example, nonvictims, occasional victims, and frequent victims as well as nonperpetrators, occasional perpetrators, and frequent
perpetrators.
2. Is there change between latent statuses
across time? We hypothesized that there is
change between latent statuses across
time.
a. If so, how can this change be characterized? We hypothesized that each

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

transition can be classified either: decreasing, increasing, or stable.

b. Is the change in latent statuses affected
by student characteristics such as gender, grade, and ELL status? Based on
the existing literature (e.g., Pellegrini
& Long, 2002; Williford et al., 2014),
we hypothesized that during the transition to middle school, students will
move from a less involved victim or
perpetrator group into a more frequently involved group. Over time,
girls will become less involved in perpetration and victimization compared
with boys (Miller et al., 2013; Nylund
et al., 2007; Rosen et al., 2013). We
hypothesized that ELL students may
experience more victimization over
time due to their language and cultural
differences.
3. If an individual is in a particular latent
status at Time t, what is the probability
that the individual will be in that latent
status at Time t 1, and what is the
probability that the individual will be in a
different latent status? We hypothesized
that involvement in bullying and victimization is a dynamic experience over time;
however, due to limited research in the
literature, this research question is exploratory in nature and we do not have specific
hypothesis regarding the probability.
Method
Participants
Data for this study are part of a larger international longitudinal investigation involving researchers from the United States, Japan, Korea,
Australia, and Canada (Konishi et al., 2009).
Part of the first wave of the data collected in the
U.S. were published in a study examining the
different experiences of bullying among students in special and general education (Swearer,
Wang, Magg, Siebecker, & Frerichs, 2012). In
the current study, data were collected from fifth
to ninth graders over three semesters (Fall 2005,
Spring 2006, and Fall 2006). Participants were
1,180 students from fifth to ninth grades (mean
age 12.2, SD 1.29) in Fall 2005 (Time 1)
attending nine schools in a midwestern city in
the United States. Due to students school tran-

109

sitions, the number of schools increased to 22

over three semesters. Slightly more than half
(52.9%) of participants were female, 46.5%
were male, and gender information was not
available for 0.6% of participants at Time 1.
Grades were distributed from fifth to ninth grades:
fifth (10.0%), sixth (31.4%), seventh (26.4%),
eighth (21.0%), and ninth grade (10.6%). Among
the participants, 9.9% indicated that English was
not their first language. The ethnicity of the sample was predominantly Caucasian: Caucasian/
White (80.2%), Black/African American (7.1%),
Latino/Hispanic (5.4%), Asian American (2.4%),
other (1.7%), and missing (3.2%). The attrition
rates were 5.59% at Spring 2006 and 15.34% at
Fall 2006.
Procedure
Recruitment letters were distributed to all
parents with children from fifth grade to ninth
grade in the participating schools. Parents were
informed that the results would be confidential,
and that they could withdraw their consent at
any time without penalty. Approximately 53%
of the consent forms were returned by the parents and/or guardians and 81.1% of them gave
consent for their children to participate. Students were also given a youth assent form.
Almost all of the students (97%) assented to
participate in the study. Only the students
whose parents gave consent and the students
who gave assent were included in the current
study. Students completed the instruments in
large groups at school (e.g., classroom, lunchroom) during the regular school day. Trained
graduate research assistants gave clear instructions to the students, answered students questions during data collection, and checked the
measures for any missing data.
Measures
Each student completed a demographic questionnaire that included questions about gender,
age, grade, first language use, and race/
ethnicity. Then, students completed the PacificRim Bullying measure (PRBm; Konishi et al.,
2009; Swearer et al., 2012) which surveyed
students experiences and concerns about bullying and victimization without using the word
bullying in order to avoid misunderstanding
or different understandings of the bullying construct across countries and languages. Instead,

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

110

RYOO, WANG, AND SWEARER

students were provided with the following instruction to capture three primary distinguishing
characteristics of bullyingintentionality, repetition, and power differential: Students can be
very mean to one another at school. Mean and
negative behavior can be especially upsetting
and embarrassing when it happens over and
over again, either by one person or by many
different people in the group. We want to know
about times when students use mean behavior
and take advantage of other students who cannot defend themselves easily. Data were also
collected from school records that included demographics, students GPA, and office referral
data. As part of the longitudinal study, students
also completed additional counterbalanced
measures on internalizing symptoms and cognitive functioning.
Students were asked to respond to six items
from the PRBm to measure peer victimization:
In the past 2 months, how often have other
students been mean or negative to you (a) by
pushing, hitting, or kicking or other physical
ways (jokingly)?; (b) by pushing, hitting, or
kicking or other physical ways (on purpose)?;
(c) by taking things from them or damaging
their property?; (d) by teasing, calling them
names, threatening them verbally, or saying
mean things to you?; (e) by excluding or
ignoring them, spreading rumors or saying
mean things about them to others, or getting
others not to like them?; (f) by using computer, e-mail, or phone text messages? The
same six items tapped bullying perpetration.
Response options were based on a 4-point Likert-type scale, ranging from never, once or
twice, about once a week, to several times a
week. In the current study, because bullying was
defined as a purposeful aggressive behavior,
one item by pushing, hitting or kicking or other
physical ways (jokingly) was not included in
the analyses, because the word, jokingly negated the purposeful aggressive nature of bullying. Instead of using all four responses (never,
once or twice, about once a week, and several
times a week), the last two response options
were combined into once or more a week, because
they both represent regular weekly involvement in
victimization and bullying perpetration. Internal
consistencies for the self-reported victimization
scale were .72 (Fall 2005), .75 (Spring 2006),
and .78 (Fall 2006). Internal consistencies for
the self-reported bullying perpetration scale

were .73 (Fall 2005), .74 (Spring 2006), and .77

(Fall 2006). Additional information regarding
the validity of PRBm can be found in Swearer,
Wang, Magg, Siebecker, and Frerichs (2012).
For example, the bullying perpetration mean
score in PRBm correlated significantly with aggression measured by Childrens Social Behavior Scale (Crick & Grotpeter, 1995) and Bully
Survey-Short (Swearer, 2006). The victimization mean score in PRBm correlated significantly with victimization measured by Childrens Social Experiences Questionnaire (Crick
& Grotpeter, 1995) and Bully Survey-Short
(Swearer, 2006).
Statistical Analyses
We conducted LTA, to represent the complex
array of response proportions in this data set in
a format that is more parsimonious and easier to
comprehend, and at the same time to reveal
important scientific information contained in
the data. Using the same notations as that in
Collins and Lanza (2010), the LTA with three
time lags in this study can be written as
P(Y y)

Rj

I(y r )
k)
k (k k )(k k )t1
j1
r 1 (j,r
k 1 k 1 k 1
1

j,t
j,t

j,t

j,t
t

where k 1, . . . , K are the number of latent

classes, j 1, . . . , J are observed variables
having r j 1, . . . , R j response categories (rj
3 for items for bullying and victimization measures), ks are probability of membership in the
kth latent class, s are the item-response probabilities, I is an indicator function, k j1 k j represents the probability of a transition to latent
status k at time j 1, conditional on membership in latent status k at time j. Multiple-Groups
LTA, denoted by g, can also be written as
P(Y y G g)

Rj

I(y r )
k ,g) .
k ,g(k k ),g(k k ),gt1
j1
r 1 (j,r

k11 k21 k31

j,t
j,t

j,t
t

j,t

In the analyses, we restricted parameters of

item-response probabilities, , across times and
Iy j,tr j
Iy j,tr j,t
j,tr j,t
groups, that is, Iy
j,r j,t kt j,r j kt and j,r j,t kt,g
Iy j,tr j
j,r
,
respectively.
The
purposes
of
assuming
j kt
measurement invariance across times, groups,
and latent statuses are the same as in Collins and
Lanza (2010; pp. 212213), because it is easy to

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

interpret the latent transition models and the

other is to help stabilize estimation and improve
model identification.
To select the optimal number of latent statuses for bullying and victimization, we compared LTA models from two latent classes to six
latent classes by applying information criteria
(AIC and BIC). In the case of disagreement of
AIC and BIC, we selected the optimal number
by looking at trends of changes in both information criteria. The optimal number was used in
subsequent latent transition analyses with grouping variables: gender, grade, and ELL status.
Next, the testing hypotheses about change of
transition probabilities between times was conducted by using likelihood ratio difference test
(LRDT). That is, we tested if a single transition
matrix was enough to explain the change of
latent statuses across times. According to the
work by Read and Cressie (1988; as cited in
Collins & Lanza, 2010), the LRDT statistic,
G2 G22 G21, with df df 2 df 1, is likely
to be approximated well by the 2 distribution
with df when df is relatively small. In this
study, there are 28 dfs that vary from 2 to 96
with median 12.
In multiple Groups LTA, four models were
compared to obtain the best fitting models in
terms of restrictions on latent status prevalence
at Time 1 and transition probabilities (see Table
1). In Model 1, both prevalence and transitions
were free to vary across groups. In Model 2, the
prevalence was constrained across groups. In
Model 3, the transitions were constrained across
groups. In Model 4, both prevalence and transitions were constrained. These hypothesis tests
allowed us to examine the possible variances in
terms of latent status prevalence and transition

111

probabilities according to grouping variables.

To compare four models, AIC and BIC were
also considered in addition to the LRDT because the LRDT works only for nested models
and those four models are not fully nested. We
compared Models 2, 3, and 4 with Model 1
unless either Model 2 or Model 3 fit equally
well into the data and fit better than Model 1.
When Model 2 or Model 3 fit equally well into
the data and fit better than Model 1, we compared Model 4 with the selected model via the
results of the information criteria.
Results
Students Experiences of Victimization
To identify the number of latent statuses,
two-status to six-status models were compared
using the information criteria, AIC and BIC,
and the interpretability of latent classes in terms
of item probabilities was considered. The lowest AIC was found for a six-status model, while
the lowest BIC was found for a four-status
model. The four-status model, however, was
more parsimonious and conceptually more appealing than the six-status model in terms of
item probabilities. We classified four different
groups of students who were bullied. Frequent
victims referred to the regularly (weekly) bullied students: they were regularly physically
harmed (47% probability of endorsing the response), their property was regularly taken or
damaged (39%), they were regularly and verbally bullied (82%), and they were regularly
isolated from peers due to rumors (66%). However, they were not highly involved with cyberbullying (25% answered weekly, 18% answered

Table 1
Model Comparison of Selecting the Best Fitting Model
Latent status
prevalences at Time 1

Item-response probabilities
LTA without grouping
variable
Model 1
Model 2
Multiple groups LTA
Model 1
Model 2
Model 3
Model 4

Equal across times and latent statuses

Equal across items and latent statuses
Equal
Equal
Equal
Equal

across
across
across
across

times,
times,
times,
times,

groups
groups
groups
groups

and
and
and
and

latent
latent
latent
latent

Transition
probabilities

Free
Equal across times
statuses
statuses
statuses
statuses

Free
Equal across groups
Free
Equal across groups

Free
Free
Equal across groups
Equal across groups

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

112

RYOO, WANG, AND SWEARER

once or twice during the past 2 months, and

58% answered never). Occasional traditional
victims referred to the students who were occasionally bullied physically, verbally, and relationally, but not electronically. They answered
high probabilities for once or twice during the
past 2 months in terms of being physically
harmed (36%), their property was once or twice
taken or damaged (34%), they were once or
twice verbally bullied (59%), and they were
once or twice isolated from peers due to rumors
(42%). However, they were not involved with
cyberbullying (99% answered never). Occasional cyber and traditional victims referred to
the occasional verbal, relational, and cyber victims: They were once or twice verbally bullied
(69%), isolated from peers due to rumors
(69%), and cyber bullied (57%), but were less
likely to experience physical bullying behavior
(pushing/hitting/kicking, 24%) and property
damage (38%). Infrequent victims referred to
the rarely bullied students. These statuses were
mutually exclusive and exhaustive. Item probabilities ( estimates) were summarized in Table
2. The main difference between the occasional
traditional victims and the occasional cyber and
traditional victims were that the occasional traditional victims did not report experiencing cyber victimization.
The probability of being in the frequent, occasional traditional, and occasional cyber and
traditional victims statuses was 50% at Fall
2005, 51% at Spring 2006, and 40% at Fall
2006 (see estimates in Table 3). These results

suggested that after students returned from summer break, they were less likely to be victimized. Specifically, the decrease mainly occurred
among the occasional traditional victim group
(from 29% to 18%). The probability of being an
occasional cyber and traditional victim increased from 10.3% (Fall 2005) to 12.2%
(Spring 2006) and 12.8% (Fall 2006).
In this article, we focus our discussion on the
difference of 20% or higher between the probabilities at two different time points because the
transition probability of 20% is relatively high
in this sample. We found that the total number
of frequent victims mainly stayed level from
Fall 2005 to Spring 2006, but their group composition changed over time. Specifically, 24%
of members in the frequent victim group
changed into the occasional traditional victim
group, meanwhile, 14% of members in the occasional traditional victim group and 14% occasional cyber and traditional victims changed
into the frequent victim group. There were
greater changes in statuses from Spring 2006 to
Fall 2006. About 35.9% of members in the
frequent victim group changed into the occasional traditional victim group, 46.1% of members in the occasional traditional victim group
changed into the infrequent victim group, and
about 28.4% of members in the occasional cyber and traditional victim group changed into
the infrequent victim group (see Table 3). This
suggests that after the summer break, most students experienced less victimization.

Table 2
Item Probabilities of the Latent Statuses ( Estimates) on the Victimization Item (Item 18)
Response
category
Latent status
All
Item18
Item18
Item18
Item18
Item18

(b)
(c)
(d)
(e)
(f)

Never

Once or twice

Once or more a week

LS1

LS2

LS3

LS4

LS1

LS2

LS3

LS4

LS1

LS2

LS3

LS4

0.2168
0.2565
0.0377
0.1265
0.5767

0.5994
0.6366
0.2629
0.5012
0.9932

0.7342
0.6119
0.2767
0.2216
0.3507

0.9371
0.9354
0.8836
0.8872
0.9505

0.3133
0.3510
0.1424
0.2103
0.1767

0.3583
0.3438
0.5884
0.4146
0.0000

0.2426
0.3815
0.6921
0.6907
0.5703

0.0597
0.0605
0.1023
0.1093
0.0439

0.4699
0.3925
0.8199
0.6632
0.2466

0.0423
0.0196
0.1487
0.0842
0.0068

0.0231
0.0066
0.0312
0.0877
0.0790

0.0032
0.0041
0.0142
0.0035
0.0055

Note. LS latent status; LS1 Frequent victim; LS2 Occasional traditional victim; LS3 Occasional cyber victim; LS4
Infrequent victim. Item 18: In the past 2 months, how often have other students been mean or negative to you (b) by pushing,
hitting, or kicking or other physical ways (on purpose)?; (c) by taking things from them or damaging their property?; (d) by
teasing, calling them names, threatening them verbally, or saying mean things to them?; (e) by excluding or ignoring them,
spreading rumors, or saying mean things about them to others, or getting others not to like them?; (f) by using computer, e-mail,
or phone text messages? The bold values indicate the relatively higher probabilities that are greater than 0.3333.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

113

Table 3
Latent Class Prevalence ( Estimate) and Transition Matrix Estimates ( Estimates) Over 3 Time Points
on the Victimization Item
estimate

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Time

LS1

LS2

LS3

estimate(a)
LS4

LS1

All
F05 0.1123 0.2886 0.1034 0.4957 LS1 0.5268
S06 0.1227 0.2914 0.1222 0.4936 LS2 0.1366
F06 0.0908 0.1827 0.1283 0.5982 LS3 0.1417
LS4 0.0180

estimate(b)

LS2

LS3

LS4

0.2388
0.7181
0.0172
0.1122

0.1429
0.0505
0.6844
0.0420

0.0915
0.0928
0.1568
0.8278

LS1
LS2
LS3
LS4

LS1

LS2

LS3

LS4

0.4197
0.0457
0.1043
0.0285

0.3588
0.4096
0.0000
0.0415

0.0533
0.0836
0.6116
0.0490

0.1682
0.4611
0.2842
0.8810

Note. LS latent status; LS1 Frequent victim; LS2 Occasional traditional victim; LS3 Occasional cyber and traditional
victim; LS4 Infrequent victim; (a) Transition matrix from Fall 2005 to Spring 2006; (b) Transition matrix from Spring 2006
to Fall 2006. The bold values indicate the relatively higher transitions that are greater than 0.2000.

Gender. Model comparison results suggest

that each gender group had different initial latent
status prevalence and different transition matrix.
The probabilities of item parameters ( estimates)
were almost identical to those in the overall sample, which indicated that the characteristics of
latent statuses were also identical to each other,
and verified the assumption of measurement
invariance.
At Fall 2005, girls were more likely to be occasional cyber and traditional victims (31%) than
boys (3%); boys were more likely to be occasional traditional victims (28%) than girls (12%),
and there were more infrequent victims among
boys (58%) than girls (46%). From Fall 2005 to
Spring 2006, 26% of frequent victim boys moved
into the occasional traditional victim group, and
22% of occasional traditional victim boys also
changed into frequent victims. On the other hand,
29% and 19% of girls in the frequent victim group
became occasional cyber and traditional victims
and occasional traditional victims. From Spring
2006 to Fall 2006, all boys in the occasional cyber
and traditional victim group moved to either the
frequent victim group (20%) or the infrequent
victim group (80%). For occasional cyber and
traditional victim girls, the majority (56%) stayed in
that group, and 38% transitioned into the infrequent
victim group. Most girls in frequent victims (62%)
became occasional traditional victims (21%),
occasional cyber and traditional victims (20%),
and infrequent victims (22%). In addition, 39%
of occasional traditional victims and 38% of
occasional cyber and traditional victims became infrequent victims. At both time points,
girls (38% to 40%) were less likely to remain in
the frequent victim class than boys (43% to

65%). Boys (0% to 52%) were less likely to

remain in the occasional cyber and traditional
victim class than girls (56% to 79%). Further
information on changes can be found in supplemental materials (see Tables S9 and S10).
Grade. Results of four model comparisons
suggest the trend of being a member in each latent
status was different across grade, but their changes
in latent statuses were the same. On the other
hand, the probabilities of item parameters ( estimates) were almost identical to those of overall
group, which also supported the measurement invariance.
Fifth graders indicated the highest probability
of being frequent victims (21%), followed by sixth
graders (12%), eighth graders (11%), seventh
graders (9%), and ninth graders (6%). On the
other hand, ninth graders had a different trend
from other students. Ninth graders reported lower
probability of being in the occasional traditional
victim group (11%) than other graders (ranging
from 27% to 39%), while reporting higher probability of being in the occasional cyber and traditional victim status (26%) than other grades
(ranging from 4% to 13%). In terms of probability
of being in a specific latent status in Fall 2005,
there were three clusters indicating similar prevalence: fifth and sixth graders, seventh and eighth
graders, and ninth graders. The transition probabilities were the same across grades, which indicated that all students stayed in the same latent
status except for 24% of the frequent victims who
changed into the occasional traditional victim
group from Fall 2005 to Spring 2006. This trend
continued at the transition from Spring 2006 to
Fall 2006. In addition, 45% of the occasional
traditional victims and 30% of the occasional

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

114

RYOO, WANG, AND SWEARER

cyber and traditional victims moved into the infrequent victim group. Further information on
changes can be found in supplemental materials
(Tables S12 and S13).
ELL status. Consistent with the results of
the grade analyses, the probability of being a
member in each latent status was different for the
language groups (ELL and non-ELL groups) but
their changes of latent statuses were the same. On
the other hand, the probabilities of item parameters ( estimates) were fairly similar compared
with those of overall group, which also supports
the measurement invariance.
We found differences in the probabilities of the
occasional victim and the infrequent victim groups
across language groups: the English speaking
group had higher probability in the occasional
victim group (33%) than the ELL group (18%),
and the ELL group had a higher probability to be
in the infrequent victim (62%) group than in the
English speaking group (47%). Results also indicated that English speaking group had a slightly
higher probability to be in the occasional cyber
and traditional victim group (9%) than in the ELL
group (6%), but a lower probability to be in the
frequent victim group (11%) than in the ELL
group (13%). The probabilities from frequent victims to occasional traditional victims and from
occasional cyber and traditional victims to the
infrequent victims were relatively higher (46%
and 24%, respectively) after controlling for ELL
status than probabilities from frequent victims to
occasional traditional victims and from occasional cyber and traditional victims to infrequent
victims in the overall sample (24% and 16%, respectively). Further information on changes can
be found in supplemental materials (see Tables
S15 and S16).
Students Experience of Bullying
Perpetration
Results indicated a three-status model for
bullying perpetration (instead of a four-status
model in victimization). Frequent perpetrator
referred to the group engaging in bullying behaviors frequently (once or more a week) physically (46%), verbally (66%), relationally
(50%), and online (30%), as well as by destroying others property (30%). Infrequent perpetrator referred to the group of students who rarely
engaged in bullying behavior (less than 5%).
Occasional verbal/relational perpetrator re-

ferred to the group of students who were involved in bullying behaviors occasionally (once
or twice in the past 2 months) mainly verbally
(61%) and relationally (39%), and to a lesser
degree through physical harm to others (27%),
destroying property (13%), and online methods
(16%). The item probabilities ( estimates)
were summarized in Table 4.
Latent status prevalence indicates that most
students were in the infrequent perpetrator
group ranging from 66% to 72%, followed by
the occasional verbal/relational perpetrator
group ranging from 22% to 30%, and the frequent perpetrator group ranging from 4% to 6%
(see estimates in Table 5). The infrequent
perpetrators and the occasional verbal/relational perpetrators tended to stay in their
groups from Fall 2005 to Spring 2006 (88% and
75%, respectively), and the frequent perpetrators spread out to other latent statuses: 24% to
the infrequent perpetrators and 35% to the occasional verbal/relational perpetrators. In addition, 19% of occasional verbal/relational perpetrators moved into the infrequent perpetrator
group. From Spring 2006 to Fall 2006, the infrequent perpetrators (91%) tended to stay in
their group, and the frequent perpetrators and
the occasional verbal/relational perpetrators
spread out to other latent statuses. For example,
34.9% of occasional verbal/relational perpetrators moved into the infrequent perpetrator
group (see estimates in Table 5).
Gender. The best fitted model with the
gender covariate was the most parsimonious
one constraining both latent status prevalence
and transition probabilities, suggesting there
were no gender differences.
Grade. Different from gender, the trends of
change varied across grades. In terms of latent
status prevalence, sixth and ninth graders were
less likely to be frequent perpetrators (3% and
1%, respectively) than other grades (probabilities ranging from 7% to 9%) in Fall 2005. On
the other hand, eighth graders were less likely to
be infrequent perpetrators (55%) than other
graders (probabilities of being an infrequent
perpetrator ranging from 66% to 77%). Similarly, eighth graders were more likely to be in
the occasional verbal/relational perpetrators
(38%) group than other graders (probabilities
ranging from 20% to 27%).
Most frequent perpetrators among sixth and
ninth graders remained in the same group (85%

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

115

Table 4
Estimate of Probabilities of Item Parameters on the Bullying Item
Response
category

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Latent status
All
Item22
Item22
Item22
Item22
Item22

(b)
(c)
(d)
(e)
(f)

Never

Once or twice

Once or more a week

LS1

LS2

LS3

LS1

LS2

LS3

LS1

LS2

LS3

0.2644
0.3943
0.1206
0.2398
0.4717

0.9740
0.9886
0.9513
0.9604
0.9803

0.7217
0.8605
0.3371
0.5953
0.8297

0.2754
0.3068
0.2226
0.2605
0.2234

0.0253
0.0114
0.0429
0.0394
0.0178

0.2678
0.1331
0.6131
0.3862
0.1579

0.4603
0.2989
0.6568
0.4996
0.3049

0.0006
0.0000
0.0058
0.0002
0.0019

0.0105
0.0064
0.0498
0.0185
0.0124

Note. LS latent status; LS1 Frequent perpetrator; LS2 Infrequent perpetrator; LS3 Occasional verbal/relational
perpetrator. Item 22: In the past 2 months, how often have you been mean or negative toward others? (b) by pushing, hitting,
or kicking or other physical ways (on purpose)?; (c) by taking things from them or damaging their property?; (d) by teasing,
calling them names, threatening them verbally, or saying mean things to them?; (e) by excluding or ignoring them, spreading
rumors, or saying mean things about them to others, or getting others not to like them?; (f) by using computer, e-mail, or phone
text messages? The bold values indicate the relatively higher probabilities that are greater than 0.3333.

and 100%, respectively) during the school year

(from Fall 2005 to Spring 2006); however, only
28% and 0% of the frequent perpetrators
among sixth and ninth graders remained in the
same status from Spring 2006 to Fall 2006. That
is, the total number of frequent perpetrators
among sixth and ninth graders remained stable
during the school year and then decreased as
those students became seventh and tenth graders. However, 78% and 62% of the frequent
perpetrators among fifth and eighth graders
tended to remain in the same status when they
became sixth and ninth graders, after the school
transition. Another interesting finding was that
38% of the occasional verbal/relational perpetrators among fifth graders moved into the frequent perpetrator group after transiting into
middle schools. Furthermore, about 5% of fifth
graders were in the frequent perpetrator group
in Spring 2006, but the probability increased

16% in Fall 2006, which was the biggest increase across grades. Further information on
changes can be found in supplemental materials
(see Tables S25 and S26).
ELL status. Results of the model comparison suggested group difference for both latent
status prevalence and transition matrix probabilities. The ELL group had slightly higher
probability of being in the frequent perpetrator
group and the occasional verbal/relational perpetrator group, 8% and 31%, than those in
English speaking groups, 5% and 25%, respectively. That is, the ELL group reported slightly
more involvement in bullying perpetration. On
the other hand, the frequent perpetrators in the
ELL group tended to either stay in the same
status (30% at Spring 2006 and 54% at Fall
2006) or move to the infrequent perpetrator
group (70% at Spring 2006 and 46% at Fall
2006), but did not move to the occasional ver-

Table 5
Latent Class Prevalence ( Estimate) and Transition Matrix Estimates ( Estimates) Over 3 Time Points
on the Bullying Item
estimate

estimate(b)

estimate(a)

Time

LS1

LS2

LS3

All
F05
S06
F06

0.0512
0.0392
0.0564

0.6884
0.6642
0.7215

0.2604
0.2966
0.2222

LS1
LS2
LS3

LS1

LS2

LS3

0.4171
0.0093
0.0444

0.2387
0.8750
0.1907

0.3466
0.1157
0.7648

LS1
LS2
LS3

LS1

LS2

LS3

0.4569
0.0265
0.0702

0.3133
0.9117
0.3492

0.2297
0.0617
0.5806

Note. LS latent status; LS1 Frequent perpetrator; LS2 Infrequent perpetrator; LS3 Occasional verbal/relational
perpetrator; (a) Transition matrix from Fall 2005 (F05) to Spring 2006 (S06); (b) Transition matrix from Spring 2006 to Fall 2006
(F06). The bold values indicate the relatively higher transitions that are greater than 0.2000.

116

RYOO, WANG, AND SWEARER

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

bal/relational perpetrator group. Frequent perpetrators among the English speaking group
tended to move into the occasional verbal/
relational perpetrator group (40% at Spring
2006 and 28% at Fall 2006) and infrequent
perpetrator (18% at Spring 2006 and 27% at
Fall 2006). Further information on changes can
be found in supplemental materials (see Tables
S28 and S29).
Discussion
Few studies have explored changes in bullying and victimization over three time points.
Previous studies have mainly used observed
data of frequency measures instead of using
latent status to capture the changes in bullying
and victimization, which neither focuses on individual change over time nor reveals the
change of latent statuses representing the true
membership in terms of bully/victim statuses.
This study used LTA with five perpetration
items and five victimization items to model
individual changes in bullying behavior and
victimization and examined the effects of gender, grade, and ELL status on the transition.
Because LTA is a data-driven method, results
may vary if different measures and different
populations are used. However, based on previous research (e.g., Reijntjes et al., 2013) and on
our results, it is likely that the groups will be
partially based on frequency (not involved,
sometimes involved, and frequently involved).
It is unlikely that the groups will be based
solely on the type of bullying/victimization
because students tend to engage in different
types of bullying behaviors/victimization simultaneously, making it impossible to separate different groups only by the type of bullying/victimization.
Prevalence Rates
Results from the current study indicate that
most students (66% to 72%) rarely participated
in bullying perpetration, 22% to 30% of students occasionally engaged in bullying behavior
toward others, and only 4% to 6% students
regularly bullied others. On the other hand,
most students (49% to 60%) were rarely involved with victimization, 18% to 29% students
were occasional traditional victims, 10% to
13% students were occasional cyber and tradi-

tional victims, and 9% to 12% students were

frequent victims. The finding is somewhat consistent with previous research which found that
a small percentage of individuals accounted for
the majority of antisocial behaviors (Falk et al.,
2014). Our findings highlight the importance of
latent statuses and their transitions, as the prevalence rates vary based on the cut-off criteria
(weekly vs. monthly) and different forms of
bullying/victimization measured (Swearer et al.,
2010). Traditional bullying interventions focus
on teaching the bullying perpetrators social
skills, and group formats are often used. However, having a group of frequent perpetrators in
a social skills group may actually promote aggression and reinforce bullying behaviors
(Dishion, McCord, & Poulin, 1999; Dodge,
Dishion, & Lansford, 2006). Considering the
small number of frequent perpetrators, individualized interventions may be more appropriate
to target specific behavioral deficiencies and
adjustment difficulties among bully perpetrations.
Victimization and Bullying Classes
Using latent class analysis, Nylund et al.
(2007) suggested that victim groups are better
classified by the severity of victimization instead of types. Partially consistent with Nylund
et al. (2007), results from the current investigation indicated four distinct victim groups based
on both severity (weekly vs. monthly) and type
(traditional vs. cyber). Specifically, we found
two groups of occasionally victimized students
who differed in the types of victimization they
experienced. One group reported experiencing
cyber and relational and verbal victimization, as
well as physical victimization (although to a
lesser degree). The other group reported experiencing physical, verbal, and relational victimization, but not cyber victimization. The finding
supports the claim that cyber victimization is a
distinct construct from physical, verbal, and relational victimization (Smith, 2012), as it differentiated the two occasional victim groups in
our study. Furthermore, findings from the current study suggest that both severity and type
are important in distinguishing different victim
groups.
Our results identified three groups of bullying
perpetrators based on the severity instead of the
type, namely frequent perpetrators, occasional

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

verbal/relational perpetrators, and infrequent

perpetrators. These results are consistent with
previous findings (Reijntjes et al., 2013; Wang,
Iannotti, & Luk, 2012) in that type is not an
important factor in distinguishing different perpetrator groups. It appears that all perpetrators
use different forms of bullying behaviors to
target their victims, although perpetrators differ
in the frequency of their involvement.
Changes in Bullying Behaviors
and Victimization
Consistent with previous research (e.g., Nylund et al., 2007; Smith et al., 1999), we found
that bullying and victimization decreased over
time and students tended to transition from a
more frequent perpetrator class into a less frequent perpetrator class over time. This finding
suggests other malleable factors in the environment may have an influence on group membership (i.e., maturation, peer norms). However,
the patterns were different during the school
transition year. The prevalence and frequency
of bullying perpetration increased from fifth to
sixth grades. During the sixth grade year, most
of the frequent perpetrators remained in the
same status, and then became less involved in
perpetration when they became seventh graders.
It is likely that the new sixth graders used bullying behaviors as a means to gain social status
after their school transition. Our findings are
consistent with social dominance theory and
previous studies, which have found that bullying tends to increase after the school transition,
and then decreases afterward (Pellegrini & Bartini, 2000; Pellegrini & Long, 2002). Considering the unique pattern during school transitions,
carefully designed interventions should target
new sixth grade students as they adjust to the
middle school environment and to new peer
groups. Interventions should focus on teaching
social-emotional learning skills to students
(e.g., www.casel.org) and appropriate ways to
navigate new peer groups and social hierarchies
(Faris & Felmlee, 2014).
An interesting finding is that although victimization decreased in general, cyber victimization increased over time. The increase in cyber
victimization may be related to increased access
to the Internet and mobile devices as adolescents get older. The distinct pattern of cyber
victimization also suggests that it is a unique

117

form of peer victimization (Smith, 2012).

Adults need to educate students about cyber
safety and monitor students technology use to
help prevent cyber victimization.
We found that although bullying behaviors
and victimization are highly stable for the infrequently involved groups (83%91% infrequent perpetrators and victims tended to stay in
the same group), it is less stable among the
frequently involved groups (54%58% frequent
perpetrators and 47%58% frequent victims
transitioned into other groups). This finding
suggests that bullying behaviors and victimization are dynamic phenomena, which challenges
the notion that aggression and victimization are
highly stable from late childhood to adolescence (Scholte, Engels, Overbeek, de Kemp, &
Haselager, 2007). Early intervention is critical
in order to teach bullying perpetrators replacement prosocial behaviors before their aggression becomes stable (Bettencourt et al., 2013).
Gender differences. Different from Miller
et al. (2013), current results did not indicate any
gender differences in bullying perpetration to
support the assertion that aggressive behavior is
less stable among girls than boys. However, we
found that girls were more likely to experience
verbal/relational and cyber victimization than
boys, and boys were more likely to be physically victimized. This gender difference in verbal/relational and cyber victimization is consistent with previous research (Kowalski &
Limber, 2007; Li, 2007), but different from
other studies finding no gender differences (e.g.,
Jackson & Cohen, 2012; Storch et al., 2005). In
addition, different from Cook, Williams,
Guerra, Kim, and Sadek (2010), we found boys
were less likely to be victimized in general
compared with girls. The inconsistent findings
in this study and previous studies may speak to
the complexity of the role of gender in peer
victimization. Other factors (e.g., grade, school
climate, peer norms) in addition to gender are
likely to play a role in peer victimization. Consistent with previous research (Nylund et al.,
2007), we found girls were less likely to remain
in the frequent victim class than boys over time.
Furthermore, girls were more likely to remain in
the occasional cyber and traditional victim
class than boys, which is consistent with previous findings that girls experience more cyber
victimization than boys (Kowalski & Limber,
2007; Li, 2007). Over time, peer victimization

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

118

RYOO, WANG, AND SWEARER

decreased in frequency and developed into more

verbal/relational and cyber forms, especially for
girls. In general, our findings provide support to
Gilligans theory (Gilligan, 1982) that girls are
more concerned with relationships than boys.
Grade differences. Our results indicated
that the youngest students at school (sixth graders in the middle school and ninth graders in the
high school) were less likely to be the frequent
perpetrators and the oldest students at school
(eighth graders in the middle school) were more
likely to engage in bullying perpetration, specifically, verbal and relational bullying behaviors. Considering that eighth graders were the
oldest students at their middle school, they were
more likely to be taller, stronger, and have more
power compared with younger students, which
may put them at a physical and social advantage
to engage in bullying behaviors toward younger
students. Comparisons across grade levels indicated the probability of being a frequent victim
decreased from fifth to ninth grade, which also
supports the decrease in peer victimization over
time (Smith et al., 1999). In addition, we found
that ninth graders were more likely to be occasional cyber and traditional victims compared
with other students. It is possible that ninth
graders have more access to Internet and cell
phones, which increases their chance of being
cyber and traditional victims.
The effect of ELL status. Few studies
have examined the relationship between first
language spoken and involvement in bullying/
victimization over time. Although some previous studies found that ELL students experienced higher level of peer victimization than
native speakers (von Grnigen et al., 2010),
results from the current study did not support
this difference. Instead, ELL students were
more likely to be infrequent victims and to be
bully perpetrators (frequent or occasional verbal/relational perpetrators) than native speakers; however, we were unable to test whether
the difference is significant at each latent status
because the latent membership was not reported
in the LTA. We do not know whether ELL
students mainly bullied other students of different ethnic backgrounds or within their own ethnic groups. It is possible that as the minority
ethnic group in their schools, ELL students engaged in bullying behaviors toward others
slightly more often than their peers because they
felt frustrated about language and cultural dif-

ferences. Future studies (e.g., qualitative studies

interviewing ELL students about the reasons of
their perpetration and victimization) are needed
to better understand this phenomenon. Overtime, native speakers were more likely to transition from the frequent perpetrators to the occasional verbal/relational perpetrators than
ELL students. However, it is not clear whether
this is related to their English language competencies or not. As ELL students become more
proficient in English, are they more likely to
engage in relationally bullying behaviors? This
is an interesting question that deserves further
exploration.
Limitations and Future Directions
There are several limitations of the current
study. First, only students self-report were used.
Students may have underreported their experiences of bullying and peer victimization due to
social desirability (Ivarsson, Broberg, Arvidsson,
& Gillberg, 2005). However, one may argue that
students views regarding bullying and victimization are most important because they reflect their
personal experiences and are possibly more accurate because the bullying may be undetected by
teachers (Card & Hodges, 2008). Future studies
should integrate information from multiple informants (e.g., peers and teachers) to provide better
construct validity. Furthermore, students in the
current study were recruited from one city in the
Midwest and the majority of students were European American. The findings may not be readily
generalizable to students living in rural areas or
other socially and politically different areas. Furthermore, we did not assess ELL students English
language competency. Future studies should examine ELL students English language competencies to rule out whether language difficulties may
contribute to their experiences of bullying/
victimization. Future researchers may also consider translating measures into languages other
than English in order to capture the experiences of
victimization among students with limited English
language competencies. In addition, the PRB
measure has only shown acceptable psychometric
properties, with internal consistency above .70
and somewhat limited support for the validity of
the measure (only correlations with other measures). It is important to mention that different
items in the PRB measure capture different subtypes of bullying behaviors, which contributes to

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

relatively low reliability. Future studies are needed

to continue examining the psychometric properties of this measure, for example, using confirmatory factor analysis to examine the factor structure
of the measure and reliabilities within the factor
structure (Raykov, 1997, 2004). Last but not least,
the occasional involvement groups refer to the
students who self-reported bullying and/or victimization once or twice in the past 2 months, which
may fail to capture the repetitive nature of bullying. This is a limitation of the PRB measure,
which was designed to assess bullying involvement in students across several countries (i.e.,
Canada, Japan, Australia, Korea, and the United
States). Given that the length of the school year is
different across countries and the structure of the
school day also varies across countries, the Pacific
Rim research team delimited involvement to the
past 2 months in order to create a measure that
was relevant across several Pacific Rim countries.
Implications
Because students involved in bullying and victimization represent heterogeneous subgroups,
with different degrees of involvement and types of
bullying, it is important to develop prevention and
intervention efforts that address these differences.
For example, interventions for the occasional cyber victim group should include psychoeducation
on cyber safety to promote awareness, while interventions for the frequent victims should focus
on strategies for dealing with physical, verbal, and
relational victimization. Interventions for frequent
perpetrators should be more intense and involve
psychotherapy and coordination between home
and school (Swearer, Wang, Collins, Strawhun, &
Fluke, 2014). Meanwhile, because results indicated that cyber victimization tended to co-occur
with traditional victimization, we suggest that educators (a) identify such a group of students who
experience pervasive (all different types of) victimization for specific interventions (e.g., psychoeducation on the appropriate use of technology; and (b) continue their efforts on reducing
general victimization instead of shifting the focus
from traditional victimization to cyberspace
(Salmivalli, Sainio, & Hodges, 2013).
Considering that the oldest students (eighth
graders in the middle school) were more likely to
engage in bullying behaviors, and bullying perpetration increased after students transitioned into
middle school, school personnel should focus their

119

intervention resources on students in the sixth and

eighth grades. Furthermore, it is important for
school personnel to be aware of and pay attention
to verbal/relational and cyberbullying, because results suggest that peer victimization develops into
more verbal/relational and cyber types over time,
especially for girls. Considering the gender differences in peer victimization, different interventions
may be warranted for boys and girls. Interventions
for girls may focus on relationship issues and
appropriate use of social media, whereas interventions for boys may address physical bullying. It is
important for teachers and parents to talk to adolescents about cyber safety and to supervise adolescents Internet and mobile device use to help
prevent cyber victimization. It is also important
for adults to take reports of verbal/relational bullying and cyberbullying seriously and to intervene
not only during physical bullying, but also with
types of bullying that might be less obvious. Considering that a small percentage of students are
responsible for the majority of bullying behaviors,
it is important to consider the use of individualized
specific interventions for those frequent perpetrators. Only when bullying interventions are developmentally based, gender- and culturally sensitive, and address all types of bullying, will
American schools be free from bullying and cruel
behaviors.

References
Bergman, L. R., & Magnusson, D. (1997). A personoriented approach in research on developmental
psychopathology. Development and Psychopathology, 9, 291319. doi:10.1017/S0954579
49700206X
Bergman, L. R., Magnusson, D., & El-Khouri, B. M.
(2003). Studying individual development in an interindividual context: A person-oriented approach. Mahwah, NJ: Erlbaum.
Bettencourt, A., Farrell, A., Liu, W., & Sullivan, T.
(2013). Stability and change in patterns of peer
victimization and aggression during adolescence.
Journal of Clinical Child and Adolescent Psychology, 42, 429 441. doi:10.1080/15374416.2012
.738455
Boulton, M. J. (1995). Patterns of bully/victim problems in mixed race groups of children. Social
Development, 4, 277293. doi:10.1111/j.14679507.1995.tb00066.x
Bye, B. V., & Schechter, E. S. (1986). A latent
Markov model approach to the estimation of response error in multiwave panel data. Journal of

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

120

RYOO, WANG, AND SWEARER

the American Statistical Association, 81, 375380.

doi:10.1080/01621459.1986.10478281
Card, N. A., & Hodges, E. V. E. (2008). Peer victimization among schoolchildren: Correlations,
causes, consequences, and considerations in assessment and intervention. School Psychology
Quarterly, 23, 451 461. doi:10.1037/a0012769
Card, N. A., Stucky, B. D., Sawalani, G. M., & Little,
T. D. (2008). Direct and indirect aggression during
childhood and adolescence: A meta-analytic review of gender differences, intercorrelations, and
relations to maladjustment. Child Development,
79, 11851229. doi:10.1111/j.1467-8624.2008
.01184.x
Coie, J. K., & Dodge, K. A. (1998). Aggression and
antisocial behavior. In W. Damon & N. Eisenberg
(Eds.), Handbook of child psychology: Vol. 3. Social, emotional, and personality development (5th
ed., pp. 779 862). New York, NY: Wiley.
Collins, L., & Lanza, S. (2010). Latent class and
latent transition analysis: With applications in The
social, behavioral, and health sciences. New York,
NY: Wiley.
Collins, L. M., & Wugalter, S. E. (1992). Latent class
models for stage-sequential dynamic latent variables. Multivariate Behavioral Research, 27, 131
157. doi:10.1207/s15327906mbr2701_8
Cook, C. R., Williams, K. R., Guerra, N. G., Kim,
T. E., & Sadek, S. (2010). Predictors of bullying
and victimization in childhood and adolescence: A
meta-analytic investigation. School Psychology
Quarterly, 25, 65 83. doi:10.1037/a0020149
Crick, N. R., & Grotpeter, J. K. (1995). Relational
aggression, gender, and social-psychological adjustment. Child Development, 66, 710 722. doi:
10.2307/1131945
Dempsey, J., Fireman, G., & Wang, E. (2006). Transitioning out of peer victimization: Victims level
of relational aggression. Journal of Psychopathology and Behavioral Assessment, 28, 273282. doi:
10.1007/s10862-005-9014-5
Dishion, T. J., McCord, J., & Poulin, F. (1999).
When intervention harm: Peer groups and problem
behavior. American Psychologist, 54, 755764.
doi:10.1037/0003-066X.54.9.755
Dodge, K. A., Dishion, T. J., & Lansford, J. E. (Eds.).
(2006). Deviant peer influences in programs for
youth. New York, NY: New York: Guilford Press.
Falk, , Wallinius, M., Lundstrm, S., Frisell, T.,
Anckarster, H., & Kerekes, N. (2014). The 1% of
the population accountable for 63% of all violent
crime convictions. Social Psychiatry and Psychiatric Epidemiology, 49, 559 571. doi:10.1007/
s00127-013-0783-y
Faris, R., & Felmlee, D. (2014). Casualties of social
combat: School networks of peer victimization and
their consequences. American Sociological Re-

view, 79, 228 257. doi:10.1177/0003122

414524573
Gilligan, C. (1982). In a different voice: Psychological theory and womens development. Cambridge,
MA: Harvard University Press.
Ivarsson, T., Broberg, A. G., Arvidsson, T., & Gillberg, C. (2005). Bullying in adolescence: Psychiatric problems in victims and bullies as measured
by the Youth Self Report (YSR) and the Depression Self-Rating Scale (DSRS). Nordic Journal of
Psychiatry, 59, 365373. doi:10.1080/08039
480500227816
Jackson, C. L., & Cohen, R. (2012). Childhood victimization: Modeling the relation between classroom victimization, cyber victimization, and psychosocial functioning. Psychology of Popular
Media Culture, 1, 254 269. doi:10.1037/a0029
482
Konishi, C., Hymel, S., Zumbo, B. D., Li, Z., Taki,
M., Slee, P., . . . Kwak, K. (2009). Investigating the
comparability of a self-report measure of childhood bullying across countries. Canadian Journal
of School Psychology, 24, 8293. doi:10.1177/
0829573509331614
Koo, D. J., Peguero, A. A., & Shekarkhar, Z. (2012).
The Model Minority victim: Immigration, gender, and Asian American vulnerabilities to violence at school. Journal of Ethnicity in Criminal
Justice, 10, 129 147. doi:10.1080/15377938.2011
.609405
Kowalski, R. M., & Limber, S. P. (2007). Electronic
bullying among middle school students. Journal of
Adolescent Health, 41, S22S30. doi:10.1016/j
.jadohealth.2007.08.017
Kozol, J. (2005). The shame of the nation: The restoration of apartheid schooling in America. New
York, NY: Crown.
Li, Q. (2007). New bottle but old wine: A research on
cyberbullying in schools. Computers in Human
Behavior, 23, 17771791. doi:10.1016/j.chb.2005
.10.005
Martin, K. M., & Huebner, E. S. (2007). Peer victimization and prosocial experiences and emotional well-being of middle school students. Psychology in the Schools, 44, 199 208. doi:10.1002/
pits.20216
Miller, S., Williams, J., Cutbush, S., Gibbs, D., Clinton-Sherrod, M., & Jones, S. (2013). Dating violence, bullying, and sexual harassment: Longitudinal profiles and transitions over time. Journal of
Youth & Adolescence, 42, 607 618. doi:10.1007/
s10964-013-9914-8
Nansel, T. R., Overpeck, M. D., Pilla, R. S., Ruan,
W. J., Simons-Morton, B., & Scheidt, P. (2001).
Bullying behavior among U.S. youth: Prevalence
and association with psychosocial adjustment.
Journal of the American Medical Association, 285,
2094 2100. doi:10.1001/jama.285.16.2094

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

CHANGE IN LATENT STATUSES IN BULLYING BEHAVIORS

Nylund, K., Bellmore, A., Nishina, A., & Graham, S.

(2007). Subtypes, severity, and structural stability
of peer victimization: What does latent class analysis say? Child Development, 78, 1706 1722. doi:
10.1111/j.1467-8624.2007.01097.x
Olenik-Shemesh, D., Heiman, T., & Eden, S. (2012).
Cyberbullying victimsation in adolescence: Relationships with loneliness and depressive mood.
Emotional and Behavioral Difficulties, 17, 361
374. doi:10.1080/13632752.2012.704227
Olweus, D. (1993). Victimization by peers: Antecedents and long-term outcomes. In K. H. Rubin & J.
Asendorf (Eds.), Social withdrawal, inhibition,
and shyness (pp. 315341). Hillsdale, NJ: Erlbaum.
Olweus, D. (1994). Annotation: Bullying at school:
Basic facts and effects of a school based intervention program. Journal of Child Psychology and
Psychiatry, 35, 11711190. doi:10.1111/j.14697610.1994.tb01229.x
Pellegrini, A. D., & Bartini, M. (2000). A longitudinal study of bullies, victims, and peer affiliation
during the transition from primary school to middle school. American Educational Research Journal, 37, 699 725. doi:10.3102/00028312
037003699
Pellegrini, A. D., & Long, J. (2002). A longitudinal
study of bullying, dominance, and victimization
during the transition from primary to secondary
school. British Journal of Developmental Psychology, 20, 259 280. doi:10.1348/026151002166442
Pepler, D. J., Craig, W. M., Connolly, J. A., Yuile,
A., McMaster, L., & Jiang, D. (2006). A developmental perspective on bullying. Aggressive Behavior, 32, 376 384. doi:10.1002/ab.20136
Prinstein, M. J., Boergers, J., & Vernberg, E. M.
(2001). Overt and relational aggression in adolescents: Social-psychological adjustment of aggressors and victims. Journal of Clinical Child Psychology, 30, 479 491. doi:10.1207/S15374424
JCCP3004_05
Raykov, T. (1997). Estimation of composite reliability for congeneric measures. Applied Psychological Measurement, 21, 173184. doi:10.1177/
01466216970212006
Raykov, T. (2004). Behavioral scale reliability and
measurement invariance evaluation using latent
variable modeling, Behavior Therapy, 35, 299
331. doi:10.1016/S0005-7894(04)80041-8
Read, T. R. C., & Cressie, N. A. C. (1988). Goodness-of-fit statistics for discrete multivariate data.
New York, NY: Springer. doi:10.1007/978-14612-4578-0
Reijntjes, A., Vermande, M., Goossens, F. A.,
Olthof, T., van de Schoot, R., Aleva, L., & van der
Meulen, M. (2013). Developmental trajectories of
bullying and social dominance in youth. Child

121

Abuse & Neglect, 37, 224 234. doi:10.1016/j

.chiabu.2012.12.004
Robers, S., Kemp, J., Truman, J., & Snyder, T. D.
(2013). Indicators of School Crime and Safety:
2012 (NCES 2013 036/NCJ 241446). Washington, DC: National Center for Education Statistics,
U.S. Department of Education, and Bureau of Justice Statistics, Office of Justice Programs, U.S.
Department of Justice. Retrieved from http://nces
.ed.gov
Rosen, L. H., Beron, K. J., & Underwood, M. K.
(2013). Assessing peer victimization across adolescence: Measurement invariance and developmental change. Psychological Assessment, 25,
111. doi:10.1037/a0028985
Rueger, S. Y., & Jenkins, L. N. (2014). Effects of
peer victimization on psychological and academic
adjustment in early adolescence. School Psychology Quarterly, 29, 77 88. doi:10.1037/spq000
0036
Salmivalli, C., Sainio, M., & Hodges, E. (2013).
Electronic victimization: Correlates, antecedents,
and consequences among elementary and middle
school students. Journal of Clinical Child & Adolescent Psychology, 42, 442 453. doi:10.1080/
15374416.2012.759228
Scholte, R. H. J., Engels, R. C. M. E., Overbeek, G.,
de Kemp, R. A. T., & Haselager, G. J. T. (2007).
Stability in bullying and victimization and its association with social adjustment in childhood and
adolescence. Journal of Abnormal Child Psychology, 35, 217228. doi:10.1007/s10802-006-9074-3
Schwartz, D., Proctor, L. J., & Chien, D. H. (2001).
The aggressive victim of bullying: Emotional and
behavioral dysregulation as a pathway to victimization by peers. In J. Juvonen & S. Graham (Eds.),
Peer harassment in schools: The plight of the
vulnerable and victimized (pp. 147174). New
York: NY: New York: Guilford Press.
Shin, J. Y., DAntonio, E., Son, H., Kim, S., & Park,
Y. (2011). Bullying and discrimination experiences among Korean-American adolescents. Journal of Adolescence, 34, 873 883. doi:10.1016/j
.adolescence.2011.01.004
Smith, P. K. (2012). Cyberbullying and cyber aggression. In S. R. Jimerson, A. B. Nickerson, M. J.
Mayer, & M. J. Furlong (Eds.), Handbook of
school violence and school safety: International
research and practice (pp. 93103). New York,
NY: Routledge.
Smith, P. K., Madsen, K., & Moody, J. (1999). What
causes the age decline in reports of being bullied at
school? Towards a developmental analysis of risks
of being bullied. Educational Research, 41, 267
285. doi:10.1080/0013188990410303
Storch, E. A., Masia-Warner, C. L., Crisp, H., &
Klein, R. G. (2005). Peer victimization and social
anxiety in adolescence: A prospective study. Ag-

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

122

RYOO, WANG, AND SWEARER

gressive Behavior, 31, 437 452. doi:10.1002/ab

.20093
Strohmeier, D., Krn, A., & Salmivalli, C. (2011).
Intrapersonal and interpersonal risk factors for
peer victimization in immigrant youth in Finland.
Developmental Psychology, 47, 248 258. doi:
10.1037/a0020785
Strohmeier, D., Wagner, P., Spiel, C., & von Eye, A.
(2010). Stability and constancy of bully-victim
behavior: Looking at variables and individuals.
Zeitschrift Fur Psychologie-Journal of Psychology, 218, 185193. doi:10.1027/0044-3409/
a000028
Sue, S. (1994). Mental health. In N. Zane, D. T.
Takeuchi, & K. Young (Eds.), Confronting critical
health issues of Asian and Pacific Islander Americans (pp. 266 288). Newbury Park, CA: Sage.
Swearer, S. M. (2006). Bully Surveyshort. Unpublished measure. University of NebraskaLincoln,
Lincoln, NE.
Swearer, S. M., Siebecker, A., Johnsen-Frerichs, L.,
& Wang, C. (2010). Assessment of bullying/
victimization: The problem of comparability
across studies and across methodologies. In S. R.
Jimerson, S. M. Swearer, & D. L. Espelage (Eds.),
Handbook of bullying in schools: An international
perspective (pp. 305327). New York, NY: Routledge.
Swearer, S. M., Wang, C., Collins, A., Strawhun, J.,
& Fluke, S. (2014). Bullying: A school mental
health perspective. In M. D. Weist, N. A. Lever,
C. P. Bradshaw, & J. S. Owens (Eds.), Handbook
of school mental health: Research, training practice, and policy (2nd ed., pp. 341354). New York,
NY: Springer. doi:10.1007/978-1-4614-7624-5_25

Swearer, S. M., Wang, C., Magg, J. W., Siebecker,

A. B., & Frerichs, L. (2012). Understanding the
bullying dynamic among students in special and
general education. Journal of School Psychology,
50, 503520. doi:10.1016/j.jsp.2012.04.001
Underwood, M. K., & Rosen, L. H. (2011). Gender
and bullying: Moving beyond mean differences to
consider conception of bullying, processes by
which bullying unfolds, and cyberbullying. In
D. L. Espelage & S. M. Swearer (Eds.), Bullying in
North American schools (pp. 1322). New York,
NY: Routledge.
U.S. Census Bureau. (2010). The foreign-born population in the United States. Retrieved from https://
www.census.gov/newsroom/pdf/cspan_fb_slides
.pdf
von Grnigen, R., Perren, S., Ngele, C., & Alsaker,
F. D. (2010). Immigrant childrens peer acceptance
and victimization in kindergarten: The role of local
language competence. British Journal of Developmental Psychology, 28, 679 697. doi:10.1348/
026151009X470582
Wang, J., Iannotti, R. J., & Luk, J. W. (2012). Patterns of adolescent bullying behaviors: Physical,
verbal, exclusion, rumor, and cyber. Journal of
School Psychology, 50, 521534. doi:10.1016/j.jsp
.2012.03.004
Williford, A., Boulton, A. J., & Jenson, J. M. (2014).
Transitions between subclasses of bullying and
victimization when entering middle school. Aggressive Behavior, 40, 24 41. doi:10.1002/ab
.21503
Received December 4, 2013
Revision received June 12, 2014
Accepted June 18, 2014