Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/374501457

Trajectories of defending behaviors: Longitudinal association with normative


and social adjustment and self-perceived popularity

Article in Journal of School Psychology · October 2023


DOI: 10.1016/j.jsp.2023.101252

CITATIONS READS

2 141

4 authors:

Ana Bravo Christian Berger


University of Cordoba (Spain) Pontificia Universidad Católica de Chile
11 PUBLICATIONS 68 CITATIONS 76 PUBLICATIONS 1,534 CITATIONS

SEE PROFILE SEE PROFILE

Rosario Ortega-Ruiz Eva María Romera Félix


University of Cordoba (Spain) University of Cordoba (Spain)
404 PUBLICATIONS 13,239 CITATIONS 126 PUBLICATIONS 2,148 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ana Bravo on 06 October 2023.

The user has requested enhancement of the downloaded file.


Journal of School Psychology 101 (2023) 101252

Contents lists available at ScienceDirect

Journal of School Psychology


journal homepage: www.elsevier.com/locate/jschpsyc

Trajectories of defending behaviors: Longitudinal association with


normative and social adjustment and self-perceived popularity
Ana Bravo a, Christian Berger b, Rosario Ortega-Ruiz a, Eva M. Romera a, *
a
Department of Psychology, Universidad de Córdoba, Avenue San Alberto Magno, s/n, 14071 Córdoba, Spain
b
School of Psychology, Pontificia Universidad Católica de Chile, Campus San Joaquín, Vicuña Mackenna 4860, Macul, Región Metropolitana, Chile

A R T I C L E I N F O A B S T R A C T

Action Editor: Stephanie Fredrick Defending the victim in bullying situations is a moral behavior that has received increasing
Editor: Craig A. Albers attention in the literature. However, important questions about the development and mainte­
nance of defending behaviors remain unanswered. The present study adopted a longitudinal
Keywords: design with two main goals: (a) identifying trajectories of change in individuals’ defending
Defending
behavior over time and (b) describing and comparing the initial levels and dynamics of change in
Bullying
normative and social adjustment and self-perceived popularity between the different defending
Normative adjustment
Social adjustment trajectories. A total of 3303 students (49.8% girls; Wave 1 Mage = 12.61 years; SD = 1.47)
Self-perceived popularity participated in the study. Data were collected in four waves with self-report questionnaires. Using
growth mixture modeling, we found four defending trajectories (84% stable-high, 5% decrease,
4% increase, and 7% stable-low). Growth mixture model multigroup and comparative analyses
found that adolescents in the stable-high defending group exhibited the highest initial levels of
normative adjustment (Mintercept = 5.47), social adjustment (Mintercept = 5.48), and self-perceived
popularity (Mintercept = 5). Adolescents in the decrease defending group tended to reduce their
normative adjustment over time (Mslope = − 0.09), whereas the increase defending group
increased their social adjustment (Mslope = 0.18) and self-perceived popularity (Mslope = − 0.04).
The stable-low defending group showed low and stable levels of normative adjustment (Mintercept
= 5.01), social adjustment (Mintercept = 5.03), and self-perceived popularity (Mintercept = 4.4).
These results indicate a strong association between normative and social adjustment and self-
perceived popularity and involvement in defending behaviors. Bullying prevention programs
could improve by adding a stronger focus on the development of classroom dynamics that pro­
mote adjusted behaviors and class-group cohesion.

1. Introduction

Bullying is an intentional aggressive behavior maintained over time in which bullies develop a relationship of power imbalance
with the victim (Smith, 2016). Most bullying situations occur in the presence of peers who can react with different behaviors and take
up different roles, such as reinforcer, victim defender, or outsider (Salmivalli et al., 1996). Traditionally, adolescents were assigned
unique and stable categorical bullying roles (e.g., Romera, Bravo, et al., 2019; Salmivalli et al., 1996), but a growing number of studies

* Corresponding author.
E-mail addresses: ana.bravo.castillo@uco.es (A. Bravo), cberger@uc.cl (C. Berger), ortegaruiz@uco.es (R. Ortega-Ruiz), eva.romera@uco.es
(E.M. Romera).

https://doi.org/10.1016/j.jsp.2023.101252
Received 26 September 2022; Received in revised form 17 April 2023; Accepted 24 September 2023
Available online 5 October 2023
0022-4405/© 2023 The Authors. Published by Elsevier Ltd on behalf of Society for the Study of School Psychology. This is an open access article
under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

suggest that behavior of adolescents involved in bullying situations can change over time. Most of the studies taking this position have
focused on the bullying and/or victimization behaviors (e.g., de Vries et al., 2021; Romera et al., 2021), with another key behavior,
defending, receiving less attention. In two recent reviews about defending behavior in bullying situacions, only 11%-12% of the
included studies were longitudinal (Lambe et al., 2019; Ma et al., 2019). Some of these longitudinal studies found moderate stability in
defending behaviors (e.g., Barchia & Bussey, 2011; Gini et al., 2022; Mazzone et al., 2018; Sijtsema et al., 2014) whereas in other
studies, defending behaviors changed over time (Kollerová et al., 2018; Meter & Card, 2015; Saarento et al., 2015). The divergence of
results suggests that defending behaviors do not exhibit a homogeneous evolution among all adolescents but rather that adolescents
follow different patterns of change (i.e., trajectories). Hence, the first objective of this study was to explore whether adolescents’
defending behaviors exhibited different trajectories over time.
The involvement in and evolution of defending behaviors may depend on adolescents’ perceptions of their own psycho-social
characteristics and those of their classroom peers, but these perceptions may also be influenced by the level of involvement in
defending (e.g., Meter & Card, 2015; Pronk et al., 2020; van der Ploeg et al., 2017). Therefore, as a second objective, this study aimed
to describe and compare the longitudinal associations between adolescents’ defending trajectories, the trajectories of normative and
social adjustment, and self-perceived popularity. Achieving these research objectives may provide a better understanding of how to
promote defending behaviors among adolescents and to improve their adjustment to the classroom context.

2. Longitudinal changes in defending behaviors

Defending a victim of bullying is a prosocial act oriented to stop bullying and/or its negative consequences. It includes a variety of
behaviors such as directly stepping in, seeking adult support, and offering comfort to the victim (Salmivalli, 2010). As bullying is a
relational and social dynamic maintained over time, defending should also be explored from a longitudinal perspective. The few
longitudinal studies that have described how defending behaviors change over time used longitudinal designs with three or fewer
waves. Additionally, these studies had widely varying sample sizes, with most including students in Grades 6–8. Moreover, they largely
conducted variable-centered analyses (e.g., regressions, correlations, structural equation models) that only described associations
between variables. Studies using person-centered analyses (e.g., cross-path analysis, latent analysis) that enable classification of in­
dividuals into distinct groups or categories based on individual response patterns are less frequent and none have focused exclusively
on defending behavior (e.g., Demaray et al., 2021; Ingram et al., 2019; Pozzoli & Gini, 2021; Pronk et al., 2020).
Some studies’ results have identified moderate stability in defending behaviors (e.g., Barchia & Bussey, 2011; Gini et al., 2022;
Mazzone et al., 2018; Pronk et al., 2020; Rambaran et al., 2022; Sijtsema et al., 2014), whereas other studies have found that defending
behaviors changed over time. These changes included the emergence of new defenders (e.g., Saarento et al., 2015), increases in the
level of defending (e.g., Meter & Card, 2015), and decreases in the level of defending (e.g., Espelage et al., 2012; Kollerová et al., 2018).
Taken together, these results suggest that different trends of change in defending behaviors among adolescents may be expected. The
two existing studies that have explored this hypothesis, which requires a person-centered analysis, did not focus solely on defending
behaviors given that five different bullying roles were included in Demaray et al. (2021) and that Ingram et al. (2019) described the
willingness to intervene only among middle adolescent boys (i.e., ages 14–17 years). To address this gap in the existing literature, the
present study aimed to explore the evolution of defending behaviors using growth mixture modeling to identify distinct latent tra­
jectories (following Muthén, 2004). Another important conclusion from the existing literature regarding changes over time in bullying
behaviors suggests that victimization and bullying do not follow linear growth trajectories (e.g., de Vries et al., 2021; Zhou et al.,
2022). Thus, the same might be expected for defending behaviors. Such non-linear development could be due to these behaviors
occurring in interaction with others and changes in the group composition or social hierarchy could divert the evolution from a linear
pattern. We therefore explored both linear and quadratic patterns of change to identify which best describes the evolution tendency of
defending behaviors.
Finally, gender and age are likely to affect defending behaviors. Being a girl and being younger have been positively associated with
the level of defending (Ma et al., 2019). Gender differences could be explained by socialization in that stereotypical girls’ roles are
associated with caring, helping, and providing peer support (Rudman & Glick, 2021; Salmivalli et al., 1996). Age differences could be
due to the significant developmental changes in perspective taking, abstract thinking, and metacognitive reasoning that take place
during adolescence (Crone & Dahl, 2012). Thus, as adolescents grow older, they increasingly understand the consequences of their
own and others’ behaviors, which may explain the general reduction in bullying as adolescence progresses (Evans & Smokowski,
2015). Moreover, as individuals grow older, they also become more aware of the potential individual (e.g., being the next victim) and
social costs (e.g., being less liked and popular) of engaging in defending behaviors. Taken together, one could expect a peak of
defending during early adolescence that then decreases at older ages due to the lower prevalence of bullying and a greater awareness of
the possible consequences of defending for oneself.

3. Adolescents’ defending behavior and their adjustment to classroom dynamics

According to social ecological theory (Bronfenbrenner, 1979), human development is the result of a complex interaction between
the individual and the many nested and interacting social environments of which they are a part. The notion of environment en­
compasses a set of interrelated contexts in which proximate processes mediate individuals’ experiences, cognitions, emotions, atti­
tudes, and behaviors (Bronfenbrenner & Morris, 2006). This model has been applied to study bullying and the victim’s defending
processes as these form complex interpersonal dynamics that are influenced by the social context (Espelage & Swearer, 2009). The
classroom is one of the most complex social contexts in which adolescents develop their social and emotional skills. Analyzing how

2
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

adolescents interact within diverse classroom dynamics, and their degree of adjustment to such dynamics, is essential for under­
standing not only differences in defending behaviors among adolescents, but also the individual tendencies for change in defending
behaviors over time. The present study focused on three individual characteristics associated with the classroom context, including
normative adjustment, social adjustment, and self-perceived popularity.
Normative adjustment refers to the development of behaviors that are considered appropriate and desirable in the classroom and that
promote a democratic school climate (Herrera et al., 2016) such as not disturbing others with their work, asking for a turn to speak, and
taking care of classroom materials. Only one study has explored the interaction between normative adjustment and defending
behavior; in Carmona-Rojas et al. (2023), normative adjustment at Wave 1 predicted defending behavior over time. Therefore, the
present study posited that adolescents with higher engagement in normative adjustment would also show greater defending behavior
levels and that an increase in defending behaviors over time would be accompanied by a concurrent increase in normative adjustment.
Both propositions were tested using a longitudinal approach.
Social adjustment is a multifaceted construct defined as the degree to which a person performs socially competent behaviors that
explain their fit in a specific peer group (Crick & Dodge, 1994; Gómez-Ortiz et al., 2017). It includes elements of social competence and
of social acceptance. Social competence refers to the cognitive evaluation of one’s own social skills and behaviors that enable positive
management of social relationships. Adolescents with greater social abilities such as cooperation, social responsibility, and engage­
ment have been found to show greater involvement in defending behaviors (e.g., Jenkins & Nickerson, 2019; van der Ploeg et al.,
2017). Social acceptance involves spending time with others, engaging in positive interactions, and having someone to provide support
and comfort (Gómez-Ortiz et al., 2017). More frequent defending behaviors have been observed when peer relationships were
perceived as positive, warm, and supportive (Thornberg et al., 2017). Despite the relevance of social adjustment for understanding
defending behaviors, only one prior longitudinal study (i.e., Carmona-Rojas et al., 2023) has explored the predictive effect of social
adjustment on defending behaviors, with their results suggesting that social adjustment may not predict defending behaviors. How­
ever, further analyses of the coevolution of both are neccesary; therefore, the present study aimed to both assess the association be­
tween defending behaviors and the level of social adjustment and test whether changes in defending behaviors would show concurrent
changes in social adjustment within the peer group.
Adolescents also tend to establish social hierarchies that are based on prestige, visibility, and power in the peer social network
(Cillessen & Marks, 2011). Consequently, some adolescents actively try to promote their popularity (Veenstra & Laninga-Wijnen,
2023). For instance, Forsberg et al. (2018) conducted a qualitative study that suggested that adolescents who witnessed peer
victimization evaluated the social position of the other involved peers in comparison to their own when deciding how to act. This
finding suggests that the decision to defend is likely to be affected by the awareness of one’s own social status, and indeed, adolescents
who defend or decide to defend have been found to enjoy higher peer-perceived popularity (e.g., Romera, Bravo, et al., 2019; Saarento
et al., 2015). However, given that defending is an active behavior, deciding to engage in defending may be associated more with self-
perceived rather than peer-perceived popularity.
Support for the idea that defending is mostly linked to self-perceived popularity can be found in Coie’s (1990) two-phase model that
proposed that the association between an individual’s actual behavior and social position goes through two phases. In the emergence
phase, an individual’s behavior guides their social position (and how they perceive their position). In the maintenance phase social
status becomes a stable individual characteristic. If an individual believes that their perceived (high) social position is due to the
previously developed behavior, they will decide to repeat and maintain such behavior. However, those who perceive a low social
position may attempt to change their behaviors and then the cycle for them can restart. Previous studies have confirmed this two-phase
cyclic model between popularity and aggressive behaviors (e.g., Mayeux & Cillessen, 2008), although only one recent empirical study
included self-perceived popularity to describe defending behaviors. Specifically, self-perceived popularity was explored as a potential
mediator between peer-perceived popularity and defending behaviors (Pozzoli & Gini, 2021). Their findings showed that peer-
perceived popularity at Wave 1 was associated with self-perceived popularity at Wave 2, which in turn was negatively associated
with defending behaviors at Wave 2. However, it should be noted that Pozzoli and Gini did not explore the longitudinal association
between self-perceived popularity and defending behaviors. The present study aimed to fill this gap in the bullying literature by
providing more information regarding the longitudinal association between defending behavior and self-perceived popularity.

4. The present study

To date, most bullying studies have utilized cross-sectional designs to examine defending behaviors and the associations between
defending behaviors and different social-contextual characteristics and adjustment. The few longitudinal studies that exist used
relatively short-term longitudinal designs and variable-centered approaches, and none have considered how defending behaviors
changed over time (see Lambe et al., 2019, and Ma et al., 2019, for more complete reviews of the literature). Therefore, our first
research objective was to explore whether adolescents’ defending behaviors followed distinct trajectories. Previous longitudinal
research has found both moderate stability in defending behavior—at high as well as at low defending levels—and changing behaviors
over time (e.g., Gini et al., 2022; Kollerová et al., 2018; Mazzone et al., 2018; Meter & Card, 2015; Saarento et al., 2015; Sijtsema et al.,
2014). We therefore expected to find groups that were stable at both high and low levels of defending and anticipated identifying at
least one group that showed changes in the level of defending (Hypothesis 1). We were also interested in potential differences in the
composition of the defending trajectory groups regarding gender and age. Based on earlier findings (e.g., Evans & Smokowski, 2015;
Rudman & Glick, 2021; Salmivalli et al., 1996), we expected early adolescents and girls reflect trajectories with more stable-high
defending levels than middle adolescents and boys (Hypothesis 2).
Our second research objective was to describe and compare the baseline levels (i.e., intercepts) and dynamics of change (i.e.,

3
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

slopes) of normative adjustment, social adjustment, and self-perceived popularity for the different defending trajectories. Based on
social ecological theory and previous studies, we proposed two hypotheses. Hypothesis 3 was that changes in the development of
defending behaviors would be associated with changes in the same direction in normative adjustment and social adjustment (Car­
mona-Rojas et al., 2023). The investigation into the longitudinal relationship between self-perceived popularity and defending be­
haviors was exploratory; consequently, no a priori hypothesis was formulated. The first part of Hypothesis 4 (Hypothesis 4a) was that
adolescents who initially had high/low defending levels would also have high/low initial normative and social adjustment levels
(Carmona-Rojas et al., 2023). Hypothesis 4b was that we expected a negative association between the initial levels of self-perceived
popularity and defending given Pozzoli and Gini’s (2021) recent results.

5. Method

5.1. Participants

The data for this analysis were obtained from a Spanish longitudinal study on risk and protective factors for bullying. Data were
collected over 2 academic years. Initially, 6082 students (Grades 5–10; 47.3% girls) ages 9–17 years (M = 13.14; SD = 1.72 at Wave 1)
were identified from 22 public schools (nine primary schools, 13 secondary schools) and participated in the overall study. Students
who did not participate in both years were excluded from the present analysis, including 1489 adolescents who did not participate in
the second academic year because they transitioned to secondary schools that were not part of the study or finished their compulsory
education between assessments and 1273 adolescents who did not participate in the first academic year because they were students
who started Grade 5 in the second year of data collection or who enrolled in participating secondary schools but were from primary
schools that were not part of the study. Another 17 students were excluded for not participating in at least three data collection waves.
The final analytic sample consisted of 3303 students (49.8% girls) ages 9–17 years (M = 12.61; SD = 1.47 at Wave 1). The first data
collection wave included 379 students from Grade 5 (2.4% grade retention), 225 from Grade 6 (2.7% grade retention), 938 from Grade
7 (9.5% grade retention), 876 from Grade 8 (14.3% grade retention), 820 from Grade 9 (7.7% grade retention), and 65 students from
Grade 10. In the Spanish educational system, the transition from primary to secondary school occurs between Grade 6 and Grade 7. The
average participation rate over the four waves was 89.23%. The participation rates included 92.7% at Wave 1 (n = 3061; Mage = 12.61
years, SD = 1.47), 87.3% at Wave 2 (n = 2882; Mage = 13.09 years, SD = 1.51), 89.9% at Wave 3 (n = 2970; Mage = 13.53 years, SD =
1.49), and 87% at Wave 4 (n = 2874; Mage = 14.07 years, SD = 1.49). Reasons for missing a wave included absence from school on the
day of data collection, having left the school (but after completing three waves), or having joined the study later due to a change of
schools (but completing three waves afterwards).

5.2. Procedure

The study was developed in accordance with the Declaration of Helsinki and ethical approval was obtained from the research ethics
committee of the Universidad de Córdoba. Schools were selected based on accessibility. Authorization was obtained from the school
principals and active parental informed consent was required for all participating students. Of the contacted parents, 67 (2%) did not
agree to their children’s participation. Participation was voluntary, anonymous, and confidential; students (or their parents) could
decide to leave the study at any time. To ensure data anonymity but still be able to link the different data collection waves, participants
were asked to create a personal code from the initial letters of their name and date of birth. Data were collected in four waves between
2017 and 2019 at 6-month intervals, with two waves at the beginning of the academic year (October–November) and two at the end
(May–June). Transition to the next academic year occurred between the second and third waves. Participants received standardized
descriptions of all study measures and completed a paper and pencil questionnaire that was administered by interviewers during
regular school lessons. The interview team consisted of four doctoral psychology students who were trained and had participated in
previous data collection exercises. On average, participants completed the questionnaires in 30 min.

5.3. Measures

5.3.1. Defending behavior


Defending behavior was measured using the defender subscale of the three item Participant Role Questionnaire (PRQ; Salmivalli &
Voeten, 2004; i.e., “I try to make the others stop bullying”, “I comfort the victim or encourage them to tell the teacher about the
bullying”, and “I tell the others to stop bullying”). As in other studies (e.g., Barchia & Bussey, 2011; Troop-Gordon et al., 2019), we
administered the PRQ as a self-report questionnaire. The items were translated into Spanish following a parallel back translation
process. Students were asked to indicate on a 5-point Likert scale (0 = never, 4 = always) how often they adopted the behavior
described in each item during the last months. A composite score was created by averaging the scores over all items. The internal
consistency of the scale in the present study was good (ωT1 = 0.69, ωT2 = 0.80, ωT3 = 0.82, ωT4 = 0.82). A tau-equivalent model was
applied instead of the standard congeneric model to determine fit indices. This was done because the standard model had zero degrees
of freedom given the number of items and thus model fit would not yield meaningful results (Czerwiński & Atroszko, 2021). Tau-
equivalence indicates that all items in the model load unto one factor equally, which increases the number of degrees of freedom
to two. The Tau-equivalent one-factor model showed the following fit indices at T1: χ2/df = 9.44, CFI = 0.995, TLI = 0.992, RMSEA =
0.053, 90% CI [0.033, 0.076], SRMR = 0.012.

4
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

5.3.2. Normative and social adjustment


Normative and social adjustment were measured with the two subscales of the Spanish version of the Adolescent Multidimensional
Social Competence Questionnaire (AMSC-Q; Gómez-Ortiz et al., 2017). The normative adjustment subscale is composed of five items
(e.g., “I take care of the material and facilities of the school”, “I ask for the floor and wait for my turn to speak”), whereas the social
adjustment subscale is composed of eight items (e.g., “My classmates or friends come to me when they have a problem”, “My class­
mates or friends count on me when there is an activity to be organized”, “I am well liked by my classmates”). Responses were provided
using a 7-point Likert scale (1 = not at all true, 7 = very true). Both subscales exhibited good internal consistency in the present study
(normative adjustment subscale: ωT1 = 0.80, ωT2 = 0.83, ωT3 = 0.82, ωT4 = 0.83; social adjustment subscale: ωT1 = 0.83, ωT2 = 0.86,
ωT3 = 0.88, and ωT4 = 0.89). Confirmatory factor analysis of the current data showed a proper adjustment to the two-dimension
structure at T1, χ2/df = 13.636, CFI = 0.967, TLI = 0.96, RMSEA = 0.064, 90% CI [0.06, 0.068], SRMR = 0.038.

5.3.3. Self-perceived popularity


Self-perceived popularity was measured using the single item of “I am popular in my classroom” on a 7-point Likert scale (1 = not at
all true, 7 = very true). Such perceptual measures based on a single item have been used in previous studies (e.g., Mayeux & Cillessen,
2008).

5.3.4. Gender and developmental stage


Gender was included as a binary categorical variable (0 = boys, 1 = girls), with only nine (0.27%) missing values. Students were
grouped into two developmental stages based on age following the World Health Organization’s guidelines (World Health Organi­
zation, 2014) and to aid comparison of potential differences. The first group included participants ages 9–13 years at the start of the
study (0 = early adolescence; n = 1449) and the second group included participants ages 14–17 years (1 = middle adolescence; n = 1854).
There were no missing age data after combining age entered at the first and second data collection waves.

5.4. Analytical plan

Pre-analyses, descriptive analyses, and correlations were performed in SPSS Version 28. The remaining analyses were conducted in
MPlus version 8.8 (Muthén & Muthén, 2017). Unconditional latent growth curve models (LGCM) were run to identify the form of the
average growth curve and to confirm whether there was significant heterogeneity in the growth factors (i.e., intercept and slope),
which would indicate that trajectories in defending differed between participants (Curran & Bollen, 2001). Fixed effects, represented
by the average intercept and slopes, and random effects, represented by the variances over individual lines, were calculated (Curran &
Bollen, 2001). As the study included four measurement moments (or waves), potential curvilinear patterns of changes in defending
behavior were based on known patterns for other bullying behaviors (e.g., de Vries et al., 2021; Zhou et al., 2022). Both linear and
quadratic solutions were compared using the Satorra-Bentler χ2 scaled difference test (Satorra & Bentler, 2001, 2010) and the Akaike
information criterion (AIC), the Bayesian information criterion (BIC), and the sample-size adjusted Bayesian information criterion
(SABIC) model fit indices. Lower values of these indices indicate a better fit of the model (Geiser, 2012). Absolute model fit was
confirmed using the root mean square error of approximation (RMSEA), the comparative fit index (CFI), and the Tucker Lewis index
(TLI). Two out of these three criteria need to meet the following threshold levels for model fit to be considered adequate, including
RMSEA ≤0.08, CFI ≥ 0.90, and TLI ≥ 0.90 (Hu & Bentler, 1999).
We used growth mixture modeling (GMM; Muthén, 2004; Muthén & Shedden, 1999), which is a type of person-centered variable
mixture model. GMM explores whether different (groups of) growth trajectories exist among the participants. For this, the model
examines differences in the growth parameter means (i.e., the means of the intercepts and slope coefficients). These growth parameters
were freely estimated within each group/latent trajectory class (i.e., having its own mean, variance, covariance, and residual vari­
ance). GMM is a model testing process; thus, multiple models were run with increasing numbers of latent trajectory classes (k). Each
model was compared against the previous model (k− 1) to find the optimal number of classes in the data. This optimum was determined
based on different fit indices (Masyn, 2013; Nylund et al., 2007). First, if AIC, BIC, and SABIC have lower values in the current model
than in the previous model, this is indicative of better model fit. Second, significant p-values in the Lo–Mendell–Rubin likelihood ratio
test (LMR-LRT) and the Vuong–Lo–Mendell–Rubin likelihood ratio test (VLMRT) when adding an additional class indicate an
improvement in model fit (Lo et al., 2001). Third, the level of entropy and posterior classification probabilities reflect the ability of the
latent trajectory model to provide a relevant partition of the data. Both range from 0 to 1, with values of 0.80 and above indicating a
profile classification with minimal uncertainty (Ram & Grimm, 2009). Besides using these statistical criteria to determine the optimal
number of latent groups, the decision was also based on theoretical expectations. Although no previous studies have explored the
evolution of defending behavior using a latent class design, existing longitudinal research did provide certain expectations about
patterns of change.
We then examined the proportions of girls and boys and the proportions of early and middle adolescents in the identified model.
Both covariates (more appropriately called auxiliary variables in this type of analysis) were included to predict class membership using
the three-step approach with adjustment for classification errors method (Asparouhov & Muthén, 2014; Vermunt, 2010). The three-
step method relies on the modal profile membership that is saved from the final unconditional model, which is the optimal model from
the GGM (Step 1). Modal profile membership is a nominal variable then used to estimate a new latent profile solution where the
classification logits (i.e., error probabilities) are fixed in values to account for classification uncertainty and retain the probability-
based classification from the optimal unconditional model (Step 2). This solution based on nominal variables is used in a multino­
mial logistic regression in which each pairwise comparison of latent class membership is estimated to assign participants across the

5
A. Bravo et al.
Table 1
Descriptive Statistics and Correlations Between the Study Variables.
Descriptives

D1 D2 D3 D4 NA1 NA2 NA3 NA4 SA1 SA2 SA3 SA4 SoP1 SoP2 SoP3 SoP4

M 2.93 3.01 3.03 3.01 5.9 5.83 5.94 5.91 5.68 5.7 5.73 5.77 4.54 4.47 4.4 4.44
SD 0.91 0.92 0.93 0.92 0.99 1.04 0.98 0.98 0.92 0.95 0.95 0.94 1.73 1.71 1.74 1.69
S − 0.8 − 0.96 − 1.05 − 0.99 − 1.21 − 1.13 − 1.18 − 1.16 − 1.14 − 1.29 − 1.27 − 1.33 − 0.46 − 0.46 − 0.41 − 0.46
K 0.24 0.59 0.86 0.77 1.65 1.32 1.47 1.24 1.81 2.33 2.51 2.66 − 0.44 − 0.45 − 0.54 − 0.41
Missing % 11.1 15.3 12.9 16.3 14.2 18 14.8 16.7 17.7 19.3 15.6 17 10.6 14.3 12.1 14.2

Correlations

D1 D2 D3 D4 NA1 NA2 NA3 NA4 SA1 SA2 SA3 SA4 SoP1 SoP2 SoP3 SoP4

D2 0.49 –
D3 0.47 0.59 –
6

D4 0.45 0.55 0.63 –


NA1 0.24 0.22 0.20 0.21 –
NA2 0.20 0.30 0.24 0.23 0.66 –
NA3 0.17 0.23 0.28 0.25 0.59 0.66 –
NA4 0.14 0.21 0.23 0.26 0.56 0.64 0.66 –
SA1 0.24 0.20 0.17 0.17 0.23 0.16 0.11 0.12 –
SA2 0.18 0.23 0.19 0.18 0.20 0.29 0.20 0.17 0.62 –
SA3 0.18 0.20 0.27 0.19 0.17 0.19 0.24 0.15 0.52 0.60 –
SA4 0.16 0.19 0.21 0.25 0.17 0.20 0.19 0.27 0.49 0.55 0.62 –
SoP1 0.11 0.10 0.06 0.08 − 0.07 − 0.11 − 0.12 − 0.12 0.47 0.31 0.31 0.27 –
SoP2 0.09 0.09 0.08 0.08 − 0.10 − 0.05 − 0.08 − 0.10 0.36 0.47 0.34 0.30 0.58 –
SoP3 0.07 0.07 0.09 0.07 − 0.11 − 0.13 − 0.10 − 0.12 0.31 0.32 0.44 0.34 0.55 0.60 –

Journal of School Psychology 101 (2023) 101252


SoP4 0.07 0.07 0.07 0.09 − 0.13 − 0.15 − 0.14 − 0.12 0.29 0.29 0.34 0.44 0.50 0.55 0.65 –

Note. D1 = Defending W1; D2 = Defending W2; D3 = Defending W3; D4 = Defending W4; NA1 = Normative Adjustment W1; NA2 = Normative Adjustment W2; NA3 = Normative Adjustment W3; NA4 =
Normative Adjustment W4; SA1 = Social Adjustment W1; SA2 = Social Adjustment W2; SA3 = Social Adjustment W3; SA4 = Social Adjustment W4; SoP1 = Self-perceived popularity W1; SoP2 = Self-
perceived popularity W2; SoP3 = Self-perceived popularity W3; SoP4 = Self-perceived popularity W4; M = Mean; SD = Standard Deviation; S = Skewness; K = Kurtosis. All correlations were significant at
p < .001.
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

latent classes and determine the distribution of gender and development stage over the classes (Step 3; Morin et al., 2020). The
“Auxiliary = (R3STEP)” option in Variable commander was included to run this procedure.
Finally, we performed three GMM with known classes (or GMM multigroup analyses) exercises to estimate the longitudinal tra­
jectories of normative adjustment, social adjustment, and self-perceived popularity of each defending trajectory group. To do this, the
participant assignment based on the best fitting trajectory model for defending was treated as observed groups; that is, we created a
categorical variable and each participant was assigned to one of the different trajectory groups depending on their individual evolution
tendency (the “save = CPROB” option in Savedata commander was set to identify the most likely class for each participant). Whether
the trajectory indicators (intercept and slope coefficients) of the three variables differed significantly between the defending groups
was determined through Wald tests (Muthén & Muthén, 2017). Prior to conducting the analyses in our last step, unconditional LGCMs
were run for each of the three variables (i.e., normative adjustment, social adjustment, and self-perceived popularity) to confirm that
there were changes over time and that growth trajectories differed between participants. Linear and quadratic solutions were
compared using the Satorra-Bentler χ2 scaled difference test. The best solution was used in the GMM for each of the three variables.
Because adolescents were nested in different classrooms, the “Cluster” option in Variable command to take account of the de­
pendencies in the data. The “Type = COMPLEX” option in Analysis command is necessary when the “Cluster” option is specified.

6. Results

6.1. Pre-analyses, selection of estimator, and missing data

Inspection of the QQ plots and histograms and the associated statistical parameters of kurtosis, skewness, and standard deviations
of the study variables (see Table 1) suggested that the data were not normally distributed. Moreover, Mardia’s test (Mardia, 1970)
results (p < .001) confirmed that multivariate skewness and kurtosis were not consistent with a multivariate normal distribution.
Hence, analyses were conducted using the MLR-estimator, which is a robust maximum likelihood estimator that uses a Huber-White
sandwich approach to adjust standard errors for non-normality (Vermunt & Magidson, 2002). Moreover, the MLR-estimator is suf­
ficiently robust to a failure of meeting the assumption of independence of the observations, which is assumed when the “Cluster”
option is selected (McNeish et al., 2017; Muthén & Muthén, 2017).
Little’s test (Little, 1988) of the study variables had a significant result (χ2 (2523) = 3387.56, p < .001), suggesting the data were
not missing completely at random (MCAR). If true, listwise deletion of cases with one or more missing values may bias results (Enders
& Bandalos, 2001). However, as χ2 is sensitive to the sample size, the normed version of χ2 was used to adjust this result, χ2/df = 1.34.
Normed χ2 values below three suggest the data can be considered missing at random (Bollen, 1989). To summarize, we conducted our
analyses using the full information maximum likelihood with the MLR estimator. MLR-estimation retains all participants including
those with missing data in all analyses, except when covariates are included because Mplus applies listwise deletion to missing data.
Given that only nine (0.27%) of the participants had missing data on covariates (i.e., gender), we excluded these participants from the
analyses.

6.2. Descriptive statistics and correlational results

Descriptive statistics and correlations of the study variables are presented in Table 1. In each wave, roughly 4% of participants
indicated zero defending behavior and approximately 54% indicated that they always or very often defended their peers, but only
20.4% (n = 675; 62.5% girls; Mage = 12.17 years; SD = 1.51 at W1) reported such high levels of defending in all waves. Correlational
analysis within each wave showed that defending behavior correlated positively with normative adjustment (rs = 0.24–0.30), social
adjustment (rs = 0.24–0.27), and self-perceived popularity (rs = 0.09–0.11). Correlations within each variable over time were sig­
nificant, with similarly sized correlation found for all variables, including high to moderate for defending behavior (rs = 0.45–0.63),
and high for normative adjustment (rs = 0.56–0.66), social adjustment (rs = 0.49–0.62), and self-perceived popularity (rs =
0.50–0.65).

6.3. Unconditional growth models

To test the type of change occurring in defending, both linear and quadratic unconditional LGCMs were run. In each model, the
growth parameters were free to covary with each other. Both linear (χ2(5) = 29.89, RMSEA = 0.04, 90% CI [0.03, 0.05], CFI = 0.99,
TLI = 0.99) and quadratic models (χ2(1) = 1.304, RMSEA = 0.01, 90% CI [0.00, 0.05], CFI = 1, TLI = 0.99) showed adequate fit
indices, but the Satorra-Bentler χ2 scaled difference test showed a better fit for the quadratic than for the linear model, Δχ2(4) = 29.54,
p < .001. This was confirmed by the AIC, BIC, and SABIC indices (see Table 2); we selected the quadratic latent growth model for the
subsequent analyses. In this model, defending linearly increased (β = 0.10, p < .001), but the negative quadratic tendency (β = − 0.03
p = .002) suggested that an initial positive change flattens or turns negative at some point. We found significant heterogeneity in the
intercepts (σ 2 = 0.40, p < .001) and quadratic tendency (σ 2 = 0.01, p = .034), but no significant heterogeneity in the linear tendency
(σ2 = 0.13, p = .067). That we found change over time and heterogeneity in starting level and growth tendencies validated proceeding
with the subsequent GGM.

7
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Table 2
Unconditional Latent Growth Curve Model Fit Indices for the Independent Variables.
Model χ2 (df) scf RMSEA CFI TLI AIC BIC SABIC
[90% C.I.]

Dl 29.89 (5) 1.56 0.04 [0.03, 0.05] 0.99 0.99 26,887.54 26,966.87 26,925.56
Dq 1.30 (1) 1.80 0.01 [0.00, 0.05] 1 0.99 26,924.36 26,979.28 26,950.69
SNl 34.91 (5) 1.69 0.04 [0.03, 0.06] 0.98 0.98 26,630.17 26,685.07 26,656.47
SNq1 33.86 (4) 1.71 0.05 [0.03, 0.06] 0.98 0.98 26,631 26,691.99 26,660.22
SAl 11.48 (5) 1.80 0.02 [0.00, 0.04] 1 1 25,850 25,904.89 25,876.29
SAq 0.16 (1) 1.55 0.00 [0.00, 0.03] 1 1 25,837.56 25,916.84 25,875.54
SoPl 18.30 (5) 1.28 0.03 [0.02, 0.04] 0.99 0.99 41,058.26 41,113.17 41,084.58
SoPq 1.61 (1) 1.13 0.01 [0.00, 0.05] 1 1 41,044.65 41,123.97 41,082.66

Note. 1The variances of the quadratic slope growth factors were fixed at zero. This implies that covariances between the quadratic slope and the
intercept or linear slope were also fixed. Dl = linear model for defending behavior; Dq = quadratic model for defending behavior; NAl = linear model
for normative adjustment; NAq = quadratic model for normative adjustment; SAl = linear model for social adjustment; SAq = quadratic model for
social adjustment; SoPl = linear model for self-perceived popularity; SoPq = quadratic model for self-perceived popularity; scf = scaling correction
factor for MLR; AIC = Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample-size adjusted Bayesian information
criterion.

6.4. Determining latent defending trajectory classes

Table 3 shows the model fit indices for the GGM estimation of defending trajectories.1 The model fit indices AIC, BIC, SABIC, and
the level of entropy showed a slightly better fit for the 5-class solution than for fewer-class models. However, LMRT and VLMRT p-
values were non-significant for the 5-class model, which suggest this solution was not an improvement upon the previous model. After
considering all criteria, we opted for the 4-class model solution.
The four defending trajectories are depicted in Fig. 1. The first group (n = 2786, 84% of the sample) had high defending levels in all
four waves (intercept = 3.17, p < .001; linear slope = 0.09, p = .014; quadratic slope = − 0.02, p = .042). The second group (n = 168,
5%) showed a high level of defending at baseline with a decreasing tendency over time (intercept = 2.71, p < .001; linear slope =
− 0.19, p = .399; quadratic slope = − 0.11, p = .141). The third group (n = 128; 4%) had low levels of defending at baseline with an
increasing tendency over time (intercept = 1.40, p < .001; linear slope = 1.76, p < .001; quadratic slope = − 0.41, p < .001). Finally,
the fourth group (n = 221; 7%) had no to low defending levels across the four waves (intercept = 1.84, p < .001; linear slope = − 0.77,
p = .001; quadratic slope = 0.28, p = .002). Wald test analyses confirmed that differences in both the intercepts and the slopes between
the groups were significant. Thus, the four defending trajectories differed significantly in their estimated baseline levels of defending
and in the changes in these levels over time.

6.5. Multinomial logistic regression

The associations between the covariates (i.e., gender and development stage) and defending group membership were assessed
through multinominal logistic regression, which formed the last step in the three-step procedure. Table 4 shows the different prob­
abilities of being a member of a particular group for boys and girls and for early and middle adolescents. Relative to boys, girls were
less likely to fall into the decreasing and stable-low groups than into the stable-high group (β = 1.03, SE = 0.23, p < .001; β = 0.92, SE
= 0.19, p < .001, respectively). No other gender differences were found. Relative to early adolescents, middle adolescents were more
likely to fall within the stable-low group than in the other three groups (stable-high group: β = 0.88, SE = 0.25, p = .001; decrease
group: β = 0.97, SE = 0.35, p = .005; increase group: β = 1.6, SE = 0.39, p < .001) and were less likely to be members of the increase
group compared to the stable-high group (β = − 0.71, SE = 0.293, p = .013).

6.6. GMM multigroup analyses for normative adjustment, social adjustment, and self-perceived popularity

As a preliminary step for the comparative analysis and to test the growth tendencies in the three variables, both linear and quadratic
unconditional LGCMs were run using the same procedure as used in the defending LGCM (see Table 2 for a complete description of
model fit indices). On the Satorra-Bentler χ2 scaled difference test, the linear model outperformed the quadratic model for normative
adjustment (Δχ2(1) = 0.73, p = .400) and social adjustment (Δχ2(4) = 10.96, p = .05), but the quadratic model performed better for
self-perceived popularity (Δχ2(4) = 17.27, p = .003). The linear models for both normative and social adjustment showed significant
heterogeneity in the intercepts (σ 2 = 0.73, p < .001; σ 2 = 0.61, p < .001, respectively) and linear slope tendencies (σ2 = 0.03, p < .001;

1
From the 3-class model solution upward, a warning message indicated that the latent variable covariance matrix in all classes were not positive
definite. Non-convergence and plausible-values problems are omnipresent in analyses that allow the mean and covariance parameters to be
completely class-specific (Diallo et al., 2016). A common solution and the Mplus default parameterization is to apply equality constraints to
covariance parameters across classes (Muthén & Muthén, 2017; Wickrama et al., 2016). Concretely, the variances of linear and quadratic slope
estimates were fixed at zero. This implies that covariances between the intercept, linear slope estimates, and quadratic slope estimates were also
fixed, and only the variance of the intercept growth factor was estimated.

8
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Table 3
GGM Defending Fit Indices for 1-Class to 5-Class Solutions.
AIC BIC SABIC LMRT VLMRT Entropy Pb %

1 26,887.54 26,966.87 26,925.56


2 26,489.34 2659.08 26,539.07 < 0.001 < 0.001 0.79 0.83–0.95 12%/88%
3 26,309.8 26,407.44 26,356.6 0.046 0.049 0.82 0.77–0.95 5%/7%/88%
4 26,162.9 26,284.95 26,221.4 0.007 0.008 0.79 0.76–0.92 4%/5%/7%/84%
5 26,046.33 26,192.79 26,116.53 0.170 0.177 0.81 0.74–0.91 1%/4%/5%/7%/83%

Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample-size adjusted Bayesian information criterion; LMRT
= Lo-Mendell-Rubin test; VLMRT = Vuong-Lo-Mendell-Rubin likelihood ratio test; Pb = Posterior classification probabilities for most likely latent
class member; % = percentage of participants per group in ascending order.

Fig. 1. Model Estimated Means from the GMM for the Latent Trajectories of Defending.

σ2 = 0.04, p < .001, respectively). The quadratic model for self-perceived popularity showed significant heterogeneity in the intercepts
(σ2 = 1.61, p < .001) and quadratic slope tendency (σ 2 = 0.04, p = .048), but not in the linear slope tendency (σ 2 = 0.08, p = .732).
Together, these results confirmed that the three variables tended to change over time and that both baseline levels and growth ten­
dencies varied within the participant sample. This heterogeneity in the variance suggests that adolescents describe different tendencies
of change that could be related to the different defending trajectories that adolescents describe according to our previous analyses.
Table 5 and Fig. 2 show the results of the GMM multigroup analyses. Significant change tendencies were found in the stable-high
group, where social adjustment linearly increased (β = 0.09; SE = 0.02; p < .001); in the decrease group, where normative adjustment
showed a linear decrease (β = − 0.09; SE = 0.04; p = .024); and in the increase group, where social adjustment linearly increased (β =
0.18; SE = 0.04; p < .001), as did self-perceived popularity (β = 0.43; SE = 0.19; p = .022).
Finally, comparative analyses were run to test the differences between defending trajectory groups in the baseline (i.e., intercepts)
and growth tendency (i.e, slopes) values for normative adjustment, social adjustment, and self-perceived popularity (see Table 5 and
Fig. 2). First, regarding the comparisons of the intercept values, adolescents in the stable-high group had higher baseline level of
normative adjustment than for adolescents in the decrease group (χ2 = 8.56, p = .003) and the stable-low group (χ2 = 24.13, p < .001),
the baseline level of social adjustment was higher than in all other groups (decrease group: χ2 = 14.48, p < .001; increase group: χ2 =
12.69, p < .001; stable-low group: χ2 = 28.47, p < .001), and the baseline level of self-perceived popularity was also higher in stable-
high than in all other groups (decrease group: χ2 = 10.79, p < .001; increase group: χ2 = 6.62, p = .01; stable-low group: χ2 = 20.9, p <
.001). No other significant differences between groups were found in the baseline levels (i.e., intercepts). Comparisons of the linear
slopes across defending groups showed that adolescents in the stable-high group exhibited a significantly larger positive growth in
social adjustment than adolescents in the decrease group (χ2 = 5.92, p = .015), whereas this group showed lower growth than the
increase group (χ2 = 6.43, p = .011). Adolescents in the decrease group were the only ones who declined in normative adjustment over
time with the negative growth coefficient for this variable being significantly lower than for all other defending groups. The increase
group displayed significantly larger positive growth in social adjustment than the stable-high group (see above) and the decrease group
(decrease group: χ2 = 10.71, p = .001). The increase group also had higher linear growth in self-perceived popularity compared to
these groups (stable-high group: χ2 = 5.23, p = .022; decrease group: χ2 = 5.05, p = .025). Finally, the comparison of the quadratic
slopes for self-perceived popularity did not result in significant differences.

9
A. Bravo et al.
Table 4
Descriptive and Multinomial Logistic Regression Results: Comparing the Defending Latent Classes by Gender and Development Stage.
Distribution among defending classes n (%)a Comparative analyses

2 vs. 1 3 vs. 1 4 vs. 1 3 vs. 2 4 vs. 2 4 vs. 3

1 2 3 4 OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI


Gender Boys 1321 54 (6.9%) 57 78 (8.7%) 0.36 [0.23, 0.69 [0.40, 0.40 [0.28, 0.53 [0.26, 1.13 [0.62, 1.73 [0.9, 3.34]
(80.1%) (4.3%) 0.56] 1.20] 0.58] 1.09] 2.06]
10

Girls 1456 114 71 143


(88.5%) (3.3%) (3.5%) (4.7%)
Dev. Early 1234 75 (5.2%) 73 (5%) 67 (4.6%) 0.91 [0.58, 0.49 [0.28, 2.41 [1.46, 1.95 [0.99, 2.64 [1.34, 0.20 [0.09,
stage (85.2%) 1.44] 0.86] 3.95] 3.84] 5.21] 0.43]
Middle 1552 93 (5%) 55 (2%) 154
(83.7%) (9.3%)

Note. a Reported percentages refer to the row distribution. Dev. Stage = developmental stage. Gender: 0 = boys, 1 = girls. Development stage: 0 = early adolescence, 1 = middle adolescence. Latent classes:
1 = Stable-high, 2 = Decrease, 3 = Increase, 4 = Stable-low.

Journal of School Psychology 101 (2023) 101252


A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Table 5
Results of the GMM Multigroup Analyses and Differences between Defense Group Growth Parameters for Normative Adjustment, Social Adjustment,
and Self-Perceived Popularity.
Normative Adjustment

Results of the Model Comparison

Intercept Linear Intercept Linear

Est. SE Est. SE Groups χ2 χ2


1 Stable-high 5.47 0.07 0.01 0.02 1–2 8.56*** 10.17***
2 Decrease 5.21 0.10 − 0.09* 0.04 1–3 3.35 1.53
3 Increase 5.27 0.12 0.06 0.04 1–4 24.13*** 0.93
4 Stable-low 5.01 0.12 0.04 0.04 2–3 0.21 10.42***
2–4 2.67 9.07**
3–4 3.71 0.14

Social Adjustment

Results of the Model Comparison

Intercept Linear Intercept Linear

Est. SE Est. SE Groups χ2 χ2


1 Stable-high 5.48 0.05 0.09*** 0.02 1–2 14.48*** 5.92*
2 Decrease 5.21 0.08 –0.01 0.04 1–3 12.69*** 6.43*
3 Increase 5.13 0.11 0.18*** 0.04 1–4 28.47*** 0.01
4 Stable-low 5.03 0.10 0.09 0.04 2–3 0.50 10.71***
2–4 2.94 3.34
3–4 0.74 3.42

Self-Perception of Popularity

Results of the Model Comparison

Intercept Linear Quadratic Intercept Linear Quadratic


2 2
Est. SE Est. SE Est. SE Groups χ χ χ2
1 Stable-high 5 0.10 0.04 0.12 0 0.03 1–2 10.79*** 1.02 0.62
2 Decrease 4.57 0.15 − 0.10 0.17 0.04 0.06 1–3 6.62* 5.23* 3.21
3 Increase 4.59 0.18 0.04* 0.20 –0.10 0.06 1–4 20.90*** 2.97 1.54
4 Stable-low 4.4 0.16 0.29 0.18 − 0.05 0.06 2–3 0.01 5.05* 3.23
2–4 0.83 3.55 2.09
3–4 0.79 0.42 0.50

Note. Est. = estimate; SE = standard error; χ2 = Wald Test. * p < .05. ** p < .01. *** p < .001.

7. Discussion

The present study adopted a longitudinal design with the first main objective being to explore whether adolescents differed in the
way their defending behaviors changed over time (i.e., whether distinct trajectories of change could be identified). The second main
objective was to explore differences between these defending trajectories in the estimated baseline levels and the growth patterns of
normative adjustment, social adjustment, and self-perceived popularity levels.

7.1. Latent longitudinal defending trajectories

The unconditional LGCM results suggested that, overall, there was a tendency for defending to increase, although the increase was
not stable over the four data collection waves as there was an inverse curve in the last data collection wave. More studies are needed to
confirm whether the non-linear tendency of change in defending behaviors is a universal phenomenon. Ideally, such studies should
include at least five repeated measurements that would allow us to conduct a completely random cubic model and have more than one
inflex point. Additionally, our findings also emphasized that defending behaviors should be described as a “role” that the individual
assumes that may vary depending on the social context characteristics. This type of change in behavioral direction has also been
observed in other bullying behaviors (e.g., de Vries et al., 2021; Zhou et al., 2022). Future studies should explore how changes in
trajectories may be explained (i.e., why adolescents who tend to increase in their defending behaviors stop defending, or adolescents
who do not defend begin to do so). Finally, by identifying non-linear growth in the level of defending behaviors, the LGCM results also
highlighted the existence of significant differences between participants’ growth curves, thereby supporting the pertinence of our
exploration of subgroups about changes in defending behavior.
The optimal GMM solution distinguished four trajectory groups for defending. Consistent with Hypothesis 1, the model showed (a)

11
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Fig. 2. GMM Multigroup Analyses: Model Estimated Means of Normative Adjustment, Social Adjustment, and Self-Perceived Popularity per
Defending Trajectory.
Note. HD = Stable-high defending; DD = Decrease in defending; ID = Increase in defending; LD = Stable-low defending. The distribution of par­
ticipants by each trajectory is shown in parentheses (%).

a stable-high trajectory that formed the normative group (including >80% of the participants), (b) a stable-low group (7% of par­
ticipants), and (c) two trajectories of changing levels of defending behaviors (10% of participants combined). Previous research has
also found a high prevalence of defending. In a recent study by Malamut et al. (2021), approximately only 20% of adolescent par­
ticipants were not nominated as defenders by their peers. Demaray et al. (2021) used latent profile analysis to examine five types of
behavior in bullying situations and found that approximately half had an uninvolved–occasional defending profile. Moreover, our
findings emphasize that most adolescents are relatively consistent in their engagement in defending behaviors, which is aligned with
previous literature (Gini et al., 2021). Despite this, and even though adolescents who change their defending behaviors constitute a
minority, it would be interesting to examine this group to better understand what drives these changes (e.g., psychosocial and
contextual characteristics) and consequently to inform intervention and prevention programs.
Differences in the composition of the defending trajectory groups by gender and age (Hypothesis 2) were found to be aligned with
previous studies (see Lambe et al., 2019; Ma et al., 2019). Compared to girls, boys had a higher probability to be in the stable-low group
and the decrease group than to be in the stable-high group. Within the present study, it appears that boys may perceive defending
behavior as a slightly less desirable behavior type. As mentioned in the introduction, the traditional socialization processes often assign
different roles to each gender (Rudman & Glick, 2021). The few educational programs developed to date that target gender stereotypes
have focused primarily on reworking roles for girls, but less attention has been given to redefining the traditional masculinization
process (e.g., Gonzalez-Perez et al., 2020). Remaining stereotypical roles that perhaps are stronger for boys may be a reason for the
differences found in our study, which would align with results from Ingram et al. (2019), who found that characteristics of traditional
masculinity are associated with a lower willingness to intervene among male early adolescents. Future educational programs should
focus on incorporating new socialization methods that promote critical thinking and advocate an absence of gender stereotypes and
their associated behavioral patterns (Espinoza-Catalán & Albornoz, 2023). Looking at relative differences between early and middle
adolescents, 84%–85% of participants in both age groups were classified as stable-high. Middle adolescents were relatively more likely
to fall into the stable-low group than in the three other groups, whereas early adolescents were relatively more likely to be classified as
being in the increase group rather than in the stable-high group. These results are aligned with previous findings that have suggested an
increase in bullying from an early age with a peak in early adolescence and a decline thereafter (Evans & Smokowski, 2015). To be able
to study these development patterns more fully, future research using GMM to describe the evolution of defending behavior would
benefit from more extensive analysis of developmental changes including childhood and late adolescence.

7.2. Defending behavior and adjustment to classroom dynamics

The GMM multigroup analyses were used to test our third and fourth hypotheses. In Hypothesis 3, we posited that adolescents who

12
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

showed rising or falling levels of defending behavior would exhibit changes in normative adjustment and social adjustment in the same
direction. We did not have any prior expectations regarding self-perceived popularity in this respect. Adolescents in the decrease
defending group indeed exhibited a significantly decreasing tendency in their normative adjustment. Moreover, this tendency was
significantly different from the normative adjustment patterns of the other three defending groups. Adolescents in the increase
defending group showed an increasing tendency in their social adjustment and self-perceived popularity. Both tendencies were
significantly different from the patterns in the stable-high defending group and the decrease defending group. These results partially
support Hypothesis 3 because a strong association between defending behaviors, normative adjustment, and social adjustment was
found, but the association appears to differ depending on whether adolescents increase or decrease their engagement in defending
behaviors. A decrease appeared to be associated with a decrease in the self-perception that one’s behavior is conforming to classroom/
school norms, whereas those who increased their defensive involvement had a heightened perception of their own social skills and they
perceived greater cohesion and group belonging (Crick & Dodge, 1994; Gómez-Ortiz et al., 2017). These results are consistent with and
extend those findings found by previous studies examining the associations between prosocial, moral, and civic behaviors (e.g., Berger
& Andaur, 2022; Carmona-Rojas et al., 2023; Herrera et al., 2016). Moreover, previous longitudinal studies found a negative longi­
tudinal association with bullying behaviors (Romera, Carmona-Rojas, Ortega-Ruiz and Camacho, 2022). Future studies should explore
whether other individual (e.g., social goals, bullying implications) and contextual (e.g., peer group norms) variables mediate the
association between these variables, which could offer clues as to why there was an association with normative adjustment for those
decreasing their defending behavior and with social adjustment for those increasing their defending behaviors.
Besides the hypothesized associations with the adjustment variables, the increase defending group also perceived themselves as
more popular over time. This outcome is consistent with Coie’s (1990) two-phase model as described iearlier. Analogous to Coie’s
model, adolescents who started or increased their defending behaviors may have perceived themselves as more popular and attributed
the improved popularity to their defending behavior, thereby strengthening their defending behaviors. Adolescents who already
defended and thus perceived themselves as more popular may have continued with their defending behavior to secure and maintain
their position within the group hierarchy. However, further longitudinal studies are needed to explore and confirm these ideas as
Coie’s two-phase model was originally proposed for the association between social status and peer rejection and not defending
behaviors.
For Hypothesis 4a (i.e., adolescents who initially have high/low defending levels would also have high/low initial normative
adjustment and social adjustment levels) and Hypothesis 4b (i.e., there would be a negative association between the initial levels of
self-perceived popularity and defending behavior), we consider the comparative results of the GMM multigroup analyses. Adolescents
in the stable-high defending group showed higher estimated baseline levels in normative adjustment, social adjustment, and self-
perceived popularity than adolescents in the decrease group and the stable-low group; however, compared to the increase defend­
ing group, they only showed higher baseline levels in social adjustment and self-perceived popularity. No differences were found in the
baseline estimated levels of normative adjustment, social adjustment, and self-perceived popularity between the decrease defending
group, the increase defending group, and the stable-low defending group. Together, these results partially supported Hypothesis 4a
and rejected Hypothesis 4b. This leads to two important conclusions. First, the association between high levels in the three variables
and defending behaviors might require a degree of longitudinal stability in both. Adolescents who are stable in their defending be­
haviors may have a greater internalization and awareness of the importance of performing these socially and morally appropriate
behaviors (Pozzoli et al., 2016) and adolescents who perceive themselves as fitting in better, being socially competent, and being
popular within the group may feel more encouraged and responsible to defend (Carmona-Rojas et al., 2023). Such a positive inter­
pretation of their adjustment to and position within the classroom dynamics could motivate them to maintain these high levels of
defending and would be consistent with the maintenance phase of the two-phase model (Coie, 1990). Second, findings indicating that
adolescents who reduced or increased their defending behaviors had similar baseline levels of normative adjustment, social adjust­
ment, and self-perceived popularity suggests that changes in defending behaviors are not related to the baseline levels of both ad­
justments and self-perceived popularity. For future research examining what influences decisions to defend, we also suggest including
other variables, such as moral sensitivity, moral courage, and self-efficacy to defend (Romera, Casas, et al., 2019). This could offer
further insights into the processes behind engagement in and maintenance of defending behaviors.
The rejection of Hypothesis 4b (i.e., being in the stable-high group was associated with higher rather than the expected lower self-
perceived popularity) is not aligned with Pozzoli and Gini (2021), who previously explored the static association between both. This
lack of consistency with this study may be because in the present study, participants who were initially high in defending behavior
were described as belonging to two separate groups, consisting of the stable-high defending group and the decrease in defending group,
finding significant differences in their perceptions of popularity. In Pozzoli and Gini, the participants described as defenders could
include both groups; therefore, the results on their perceptions of popularity could mask (i.e., under/overestimate) their actual levels.
More longitudinal studies are necessary, with a particular need for studies that combine both peer-perceived and self-perceived
popularity. Moreover, this result is also aligned with the maintenance phase of the two-phase model (Coie, 1990) and suggests that
participants who defend may positively associate this behavior with popularity. More research is required to better understand the
relationship between defending and self-perceived popularity, including how adolescents’ perceived associations between defending
and popularity differ and in what way factors such as the social and relational contexts influence these differences.

7.3. Strengths, limitations, and practical implications

Strengths of our research include its large sample size and inclusion of four data collection waves over the course of 2 years.
Compared to other studies (e.g., Gini et al., 2022; Kollerová et al., 2018; Mazzone et al., 2018; Meter & Card, 2015; Sijtsema et al.,

13
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

2014), this is a relatively long period with multiple (more than two) data points. Additionally, to our knowledge, the present study is
the first study to explore different latent defending trajectories in adolescence. Therefore, this study treated defending behavior as a
continuous dynamic behavior without arbitrary cutoff points as compared to a stable categorical variable, thus avoiding the risk of
oversimplifying the bullying phenomenon. This study also combined different complex longitudinal analyses to provide a more
complete and comprehensive description of defending behavior and its longitudinal associations with normative adjustment, social
adjustment, and self-perceived popularity. However, the present study also has several limitations. First, despite the overall statistical
power of the analysis, the relatively small proportion of participants included in three of the groups (i.e., decrease, increase, and stable-
low in defending behaviors) suggests the need to consider the results carefully and to develop similar studies to confirm these results.
Second, it instrumentalized defending as a single concept, but adolescents’ specific defending behavior (e.g., offering support to the
victim, telling the teacher, or facing the bully) could differ depending on the context, as has been shown in other studies (e.g., Gini
et al., 2021; Lambe & Craig, 2020). Therefore, future studies might consider separately exploring different types of defending be­
haviors to identify defending trajectories. Third, the present study only included self-reports of defending behaviors, whereas a
combination of self- and peer-reports or their comparison might provide a more thorough understanding. Finally, although self-
perceived popularity previously has been measured with a single item (i.e., “I am popular in my classroom”), the complexity of the
concept itself suggests that may be necessary to develop more reliable assessment instruments that include different items. The
development of a scale that considers the complexity of popularity and how students define it when answering questionnaires should
be a priority for both self and peer reporters.
The present student also leads to potential practical implications. First, the present study showed that defending may be a malleable
behavior, but that associations with other variables are not always present initially. Prevention programs therefore should consist of
broad strategies that include follow-up activities and sessions that can be implemented by teachers as reinforcement of core concepts
and abilities initially promoted and developed within the program. Second, teachers should be trained in the promotion of positive and
affective interpersonal dynamics, socio-emotional skills, and moral skills that allow for the development of more cohesive and caring
environments given the associations of social adjustment with increasing and maintained defending behaviors. Third, our results
indicate that normative adjustment was associated with defending behaviors; therefore, schools should consider implementing stra­
tegies that promote democratic participation and prosocial behaviors among students to promote the well-being of all participants
within the school daily (Encina & Berger, 2021). Finally, although popularity may be less related to changes in defending, it seems to
be a characteristic of adolescents who maintain a high involvement in these types of behaviors. Therefore, future intervention pro­
grams should include popularity as an objective of work with schoolchildren by trying to associate assertive and caring behaviors, such
as defending with high positions in the hierarchical dynamics of the classroom.

8. Conclusions

Our results indicated that defending behavior showed varying patterns of change for different aged adolescents. Therefore, this
behavior should be viewed as an individual process that depends on psycho-social and classroom characteristics. Our results revealed
concurrent, but differential, associations that illustrate how social dynamics and behaviors that coexist in the classroom can co-evolve
and reinforce each other. Adolescents who reduced or increased their defending behaviors over time exhibited similar initial (i.e.,
moderate to low) levels of normative and social adjustment and self-perceived popularity. However, the change in these variables
differed significantly, with adolescents who reduced their defending behaviors perceiving a decrease in their adjustment to the norms
of the classroom, whereas adolescents who increased defending behaviors perceived themselves as more socially adjusted and being
more popular within the peer group.

Funding statement

This study was supported by the Spanish Research Agency, Ministerio de Ciencia e Innovación (PID2020-113911RB-I00, PI: Eva M.
Romera) (www.mineco.gob.es).

Ethics approval statement

All procedures performed in studies involving human participants were in accordance with the ethical standards of the 1964
Helsinki declaration and its later amendments or comparable ethical standards. The project was reviewed and approved by Biosafety
and Bioethics Committee of the University of Córdoba. Code: CEIH-21-17.

Informed consent written

Informed consent was obtained from the parents.

Declaration of Competing Interest

The authors declare no competing interests.

14
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Data availability

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corre­
sponding author on reasonable request.

References

Asparouhov, T., & Muthén, B. O. (2014). Auxiliary variables in mixture modeling: Three-step approaches using Mplus. Structural Equation Modeling: A Multidisciplinary
Journal, 21(3), 329–341. https://doi.org/10.1080/10705511.2014.915181.
Barchia, K., & Bussey, K. (2011). Predictors of student defenders of peer aggression victims: Empathy and social cognitive factors. International Journal of Behavioral
Development, 35(4), 289–297. https://doi.org/10.1177/0165025410396746.
Berger, C., & Andaur, A. (2022). Integrating prosocial and proenvironmental behaviors: The role of moral disengagement and peer social norms. Psychology, Society,
and Education, 14(3), 18–28. https://doi.org/10.21071/psye.v14i3.15113.
Bollen, K. A. (1989). Structural equations with latent variables. John Wiley & Sons, Inc. https://doi.org/10.1002/9781118619179.
Bronfenbrenner, U. (1979). Contexts of child rearing: Problems and prospects. American Psychologist, 34(10), 844–850. https://doi.org/10.1037/0003-066X.34.10.
844.
Bronfenbrenner, U., & Morris, P. A. (2006). The bioecological model of human development. In W. Damon, R. M. Lerner, & R. M. Lerner (Eds.), Handbook of child
psychology: Vol. 1. Theoretical models of human development (6th edition; pp. 793–828). Wiley.
Carmona-Rojas, M., Ortega-Ruiz, R., Romera, E., & Bravo, A. (2023). Aggressive and defensive behaviour, normative, and social adjustment in the complex dynamics
of school bullying. Psychosocial Intervention, 32(3), 165–175. https://doi.org/10.5093/pi2023a11.
Cillessen, A. H. N., & Marks, P. E. L. (2011). Conceptualizing and measuring popularity. In A. H. N. Cillessen, D. Schwartz, & L. Mayeux (Eds.), Popularity in the peer
system (pp. 25–56). Guilford.
Coie, J. D. (1990). Toward a theory of peer rejection. In S. R. Asher, & J. D. Coie (Eds.), Peer rejection in childhood (pp. 365–401). Cambridge University Press.
Crick, N. R., & Dodge, K. A. (1994). A review and reformulation of social information-processing mechanisms in children’s social adjustment. Psychological Bulletin,
115(1), 74–101. https://doi.org/10.1037/0033-2909.115.1.74.
Crone, E. A., & Dahl, R. E. (2012). Understanding adolescence as a period of social–affective engagement and goal flexibility. Nature Reviews Neuroscience, 13(9),
636–650. https://doi.org/10.1038/nrn3313.
Curran, P. J., & Bollen, K. A. (2001). The best of both worlds: Combining autoregressive and latent curve models. In L. M. Collins, & A. G. Sayer (Eds.), New methods for
the analysis of change (pp. 107–135). American Psychological Association.
Czerwiński, S. K., & Atroszko, P. A. (2021). A solution for factorial validity testing of three-item scales: An example of tau-equivalent strict measurement invariance of
three-item loneliness scale. Current Psychology, 42, 1652–1664. https://doi.org/10.1007/s12144-021-01554-5.
Demaray, M. K., Malecki, C. K., Ryoo, J. H., & Summers, K. H. (2021). Deconstructing bullying roles: A longitudinal latent profile analysis of bullying participant
behaviors for students in grades 4 through 12. Journal of School Psychology, 86, 32–48. https://doi.org/10.1016/j.jsp.2021.02.006.
Diallo, T. M. O., Morin, A. J. S., & Lu, H. (2016). Impact of misspecifications of the latent variance–covariance and residual matrices on the class enumeration accuracy
of growth mixture models. Structural Equation Modeling: A Multidisciplinary Journal, 23(4), 507–531. https://doi.org/10.1080/10705511.2016.1169188.
Encina, Y., & Berger, C. (2021). Civic behavior and sense of belonging at school: The moderating role of school climate. Child Indicators Research, 14, 1453–1477.
https://doi.org/10.1007/s12187-021-09809-0.
Enders, C., & Bandalos, D. (2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models.
Structural Equation Modeling: A Multidisciplinary Journal, 8(3), 430–457. https://doi.org/10.1207/S15328007SEM0803_5.
Espelage, D. L., Green, H., & Polanin, J. (2012). Willingness to intervene in bullying episodes among middle school students: Individual and peer-group influences. The
Journal of Early Adolescence, 32(6), 776–801. https://doi.org/10.1177/0272431611423017.
Espelage, D. L., & Swearer, S. M. (2009). A social–ecological model for bullying prevention and intervention: Understanding the impact of adults in the social ecology
of youngsters. In S. R. Jimerson, S. M. Swearer, & D. L. Espelage (Eds.), Handbook of bullying in schools: An international perspective (pp. 61–72). Routledge. https://
doi.org/10.4324/9780203842898.
Espinoza-Catalán, A. M., & Albornoz, N. (2023). Sexismo en Educación Superior: ¿Cómo se Reproduce la Inequidad de Género en el contexto Universitario? Psykhe, 32
(1). https://doi.org/10.7764/psykhe.2021.35613.
Evans, C. B. R., & Smokowski, P. R. (2015). Prosocial bystander behavior in bullying dynamics: Assessing the impact of social capital. Journal of Youth and Adolescence,
44(12), 2289–2307. https://doi.org/10.1007/s10964-015-0338-5.
Forsberg, C., Wood, L., Smith, J., Varjas, K., Meyers, J., Jungert, T., & Thornberg, R. (2018). Students’ views of factors affecting their bystander behaviors in response
to school bullying: A cross-collaborative conceptual qualitative analysis. Research Papers in Education, 33(1), 127–142. https://doi.org/10.1080/02671522.2016.
1271001.
Geiser, C. (2012). Data analysis with Mplus (methodology in the social sciences). Guilford Press.
Gini, G., Pozzoli, T., Angelini, F., Thornberg, R., & Demaray, M. K. (2022). Longitudinal associations of social-cognitive and moral correlates with defending in
bullying. Journal of School Psychology, 91, 146–159. https://doi.org/10.1016/j.jsp.2022.01.005.
Gini, G., Pozzoli, T., Jenkins, L., & Demaray, M. K. (2021). Participant roles in bullying. In P. K. Smith, & J. O’Higgins Norman (Eds.), Handbook of bullying. A
comprehensive and international review of research and intervention. Volume 1: Characteristcs, risks and outcomes (pp. 76–98). Wiley-Blackwell.
Gómez-Ortiz, O., Romera-Félix, E.-M., & Ortega-Ruiz, R. (2017). Multidimensionality of social competence: Measurement of the construct and its relationship with
bullying roles. Revista de Psicodidáctica (English Ed.), 22(1), 37–44. https://doi.org/10.1387/RevPsicodidact.15702.
Gonzalez-Perez, S., Mateos de Cabo, R., & Sainz, M. (2020). Girls in STEM: Is it a female role-model thing? Frontiers in Psychology, 11, 2204. https://doi.org/10.3389/
fpsyg.2020.02204.
Herrera, M., Romera, E., Ortega-Ruiz, R., & Gómez-Ortiz, O. (2016). Influence of social motivation, self-perception of social efficacy and normative adjustment in the
peer setting. Psicothema, 28(1), 32–39. https://doi.org/10.7334/psicothema2015.135.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation
Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118.
Ingram, K. M., Davis, J. P., Espelage, D. L., Hatchel, T., Merrin, G. J., Valido, A., & Torgal, C. (2019). Longitudinal associations between features of toxic masculinity
and bystander willingness to intervene in bullying among middle school boys. Journal of School Psychology, 77, 139–151. https://doi.org/10.1016/j.jsp.2019.10.
007.
Jenkins, L. N., & Nickerson, A. B. (2019). Bystander intervention in bullying: Role of social skills and gender. The Journal of Early Adolescence, 39(2), 141–166. https://
doi.org/10.1177/0272431617735652.
Kollerová, L., Yanagida, T., Mazzone, A., Soukup, P., & Strohmeier, D. (2018). “They think that I should defend”: Effects of peer and teacher injunctive norms on
defending victimized classmates in early adolescents. Journal of Youth and Adolescence, 47(11), 2424–2439. https://doi.org/10.1007/s10964-018-0918-2.
Lambe, L. J., Cioppa, V. D., Hong, I. K., & Craig, W. M. (2019). Standing up to bullying: A social ecological review of peer defending in offline and online contexts.
Aggression and Violent Behavior, 45, 51–74. https://doi.org/10.1016/j.avb.2018.05.007.
Lambe, L. J., & Craig, W. M. (2020). Peer defending as a multidimensional behavior: Development and validation of the defending behaviors scale. Journal of School
Psychology, 78, 38–53. https://doi.org/10.1016/j.jsp.2019.12.001.
Little, R. J. A. (1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83(404),
1198–1202. https://doi.org/10.1080/01621459.1988.10478722.

15
A. Bravo et al. Journal of School Psychology 101 (2023) 101252

Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing the number of components in a normal mixture. Biometrika, 88(3), 767–778.
Ma, T.-L., Meter, D. J., Chen, W.-T., & Lee, Y. (2019). Defending behavior of peer victimization in school and cyber context during childhood and adolescence: A meta-
analytic review of individual and peer-relational characteristics. Psychological Bulletin, 145(9), 891–928. https://doi.org/10.1037/bul0000205.
Malamut, S. T., Trach, J., Garandeau, C. F., & Salmivalli, C. (2021). Examining the potential mental health costs of defending victims of bullying: A longitudinal
analysis. Research on Child and Adolescent Psychopathology, 49(9), 1197–1210. https://doi.org/10.1007/s10802-021-00822-z.
Mardia, K. V. (1970). Measures of multivariate skewness and kurtosis with applications. Biometrika, 57(3), 519–530. https://doi.org/10.2307/2334770.
Masyn, K. E. (2013). Latent class analysis and finite mixture modeling. In T. D. Little (Ed.), The Oxford handbook of quantitative methods (pp. 551–611). Oxford
University Press.
Mayeux, L., & Cillessen, A. H. N. (2008). It’s not just being popular, it’s knowing it, too: The role of self-perceptions of status in the associations between peer status
and aggression. Social Development, 17(4), 871–888. https://doi.org/10.1111/j.1467-9507.2008.00474.x.
Mazzone, A., Camodeca, M., & Salmivalli, C. (2018). Stability and change of outsider behavior in school bullying: The role of shame and guilt in a longitudinal
perspective. The Journal of Early Adolescence, 38(2), 164–177. https://doi.org/10.1177/0272431616659560.
McNeish, D., Stapleton, L. M., & Silverman, R. D. (2017). On the unnecessary ubiquity of hierarchical linear modeling. Psychological Methods, 22(1), 114–140. https://
doi.org/10.1037/met0000078.
Meter, D. J., & Card, N. A. (2015). Effects of defending: The longitudinal relations among peer-perceived defending of victimized peers, victimization, and liking.
Social Development, 24(4), 734–747. https://doi.org/10.1111/sode.12129.
Morin, A. J. S., McLarnon, M. J. W., & Litalien, D. (2020). Mixture modeling for organizational behavior research. In Y. Griep, & S. D. Hansen (Eds.), Handbook on the
temporal dynamics of organizational behavior (pp. 351–379). Edward Elgar Publishing. https://doi.org/10.4337/9781788974387.00031.
Muthén, B. O. (2004). Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In D. Kaplan (Ed.), Handbook of quantitative
methodology for the social sciences (pp. 345–368). Sage Publications, Inc.
Muthén, B. O., & Shedden, K. (1999). Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics, 55(2), 463–469. https://doi.org/10.1111/j.
0006-341X.1999.00463.x.
Muthén, L. K., & Muthén, B. O. (2017). Mplus user’s guide (8th ed.). Muthén & Muthén.
Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo
simulation study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569. https://doi.org/10.1080/10705510701575396.
van der Ploeg, R., Kretschmer, T., Salmivalli, C., & Veenstra, R. (2017). Defending victims: What does it take to intervene in bullying and how is it rewarded by peers?
Journal of School Psychology, 65, 1–10. https://doi.org/10.1016/j.jsp.2017.06.002.
Pozzoli, T., & Gini, G. (2021). Longitudinal relations between students’ social status and their roles in bullying: The mediating role of self-perceived social status.
Journal of School Violence, 20(1), 76–88. https://doi.org/10.1080/15388220.2020.1850462.
Pozzoli, T., Gini, G., & Thornberg, R. (2016). Bullying and defending behavior: The role of explicit and implicit moral cognition. Journal of School Psychology, 59,
67–81. https://doi.org/10.1016/j.jsp.2016.09.005.
Pronk, J., Olthof, T., Aleva, E. A., Meulen, M., Vermande, M. M., & Goossens, F. A. (2020). Longitudinal associations between adolescents’ bullying-related indirect
defending, outsider behavior, and peer-group status. Journal of Research on Adolescence, 30(S1), 87–99. https://doi.org/10.1111/jora.12450.
Ram, N., & Grimm, K. J. (2009). Methods and measures: Growth mixture modeling: A method for identifying differences in longitudinal change among unobserved
groups. International Journal of Behavioral Development, 33(6), 565–576. https://doi.org/10.1177/0165025409343765.
Rambaran, J. A., Pozzoli, T., & Gini, G. (2022). Socio-cognitive processes and peer-network influences in defending and bystanding. Journal of Youth and Adolescence,
51(11), 2077–2091. https://doi.org/10.1007/s10964-022-01643-z.
Romera, E. M., Bravo, A., Ortega-Ruiz, R., & Veenstra, R. (2019). Differences in perceived popularity and social preference between bullying roles and class norms.
PLoS One, 14(10), Article e0223499. https://doi.org/10.1371/journal.pone.0223499.
Romera, E. M., Carmona-Rojas, M., Ortega-Ruiz, R., & Camacho, A. (2022). Bidirectional association between normative adjustment and bullying perpetration in
adolescence: A prospective longitudinal study. Revista de Psicodidáctica, 27(2), 132–140. https://doi.org/10.1016/j.psicod.2022.03.001.
Romera, E. M., Casas, J. A., Gómez-Ortiz, O., & Ortega-Ruiz, R. (2019). Moral domain as a risk and protective factor against bullying. An integrating perspective
review on the complexity of morality. Aggression and Violent Behavior, 45, 75–82. https://doi.org/10.1016/j.avb.2018.07.005.
Romera, E. M., Jiménez, C., Bravo, A., & Ortega-Ruiz, R. (2021). Social status and friendship in peer victimization trajectories. International Journal of Clinical and
Health Psychology, 21(1), 100191. https://doi.org/10.1016/j.ijchp.2020.07.003.
Rudman, L. A., & Glick, P. (2021). The social psychology of gender: How power and intimacy shape gender relations. Guilford Publications.
Saarento, S., Boulton, A. J., & Salmivalli, C. (2015). Reducing bullying and victimization: Student- and classroom-level mechanisms of change. Journal of Abnormal
Child Psychology, 43(1), 61–76. https://doi.org/10.1007/s10802-013-9841-x.
Salmivalli, C. (2010). Bullying and the peer group: A review. Aggression and Violent Behavior, 15(2), 112–120. https://doi.org/10.1016/j.avb.2009.08.007.
Salmivalli, C., Lagerspetz, K., Björkqvist, K., Österman, K., & Kaukiainen, A. (1996). Bullying as a group process: Participant roles and their relations to social status
within the group. Aggressive Behavior, 22(1), 1–15. https://doi.org/10.1002/(SICI)1098-2337(1996)22:1<1::AID-AB1>3.0.CO;2-T.
Salmivalli, C., & Voeten, M. (2004). Connections between attitudes, group norms, and behaviour in bullying situations. International Journal of Behavioral Development,
28(3), 246–258. https://doi.org/10.1080/01650250344000488.
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66(4), 507–514. https://doi.org/10.
1007/BF02296192.
Satorra, A., & Bentler, P. M. (2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika, 75(2), 243–248. https://doi.org/10.1007/
s11336-009-9135-y.
Sijtsema, J. J., Rambaran, J. A., Caravita, S. C. S., & Gini, G. (2014). Friendship selection and influence in bullying and defending: Effects of moral disengagement.
Developmental Psychology, 50(8), 2093–2104. https://doi.org/10.1037/a0037145.
Smith, P. K. (2016). Bullying: Definition, types, causes, consequences and intervention. Social and Personality Psychology Compass, 10(9), 519–532. https://doi.org/10.
1111/spc3.12266.
Thornberg, R., Wänström, L., Hong, J. S., & Espelage, D. L. (2017). Classroom relationship qualities and social-cognitive correlates of defending and passive
bystanding in school bullying in Sweden: A multilevel analysis. Journal of School Psychology, 63, 49–62. https://doi.org/10.1016/j.jsp.2017.03.002.
Troop-Gordon, W., Frosch, C. A., Wienke Totura, C. M., Bailey, A. N., Jackson, J. D., & Dvorak, R. D. (2019). Predicting the development of pro-bullying bystander
behavior: A short-term longitudinal analysis. Journal of School Psychology, 77, 77–89. https://doi.org/10.1016/j.jsp.2019.10.004.
Veenstra, R., & Laninga-Wijnen, L. (2023). The prominence of peer interactions, relationships, and networks in adolescence and early adulthood. In L. Crockett,
G. Carlo, & J. Schulenberg (Eds.), APA handbook of adolescent and young adult development (pp. 225–241). American Psychological Association.
Vermunt, J. K. (2010). Latent class modeling with covariates: Two improved three-step approaches. Political Analysis, 18, 450–469.
Vermunt, J. K., & Magidson, J. (2002). Latent class cluster analysis. In J. A. Hagenaars, & A. L. McCutcheon (Eds.), Applied latent class analysis (pp. 89–106). Cambridge
University Press.
de Vries, E., Kaufman, T. M. L., Veenstra, R., Laninga-Wijnen, L., & Huitsing, G. (2021). Bullying and victimization trajectories in the first years of secondary
education: Implications for status and affection. Journal of Youth and Adolescence, 50(10), 1995–2006. https://doi.org/10.1007/s10964-020-01385-w.
Wickrama, K., Lee, T. K., O’Neal, C. W., & Lorenz, F. (2016). Higher-order growth curves and mixture modeling with Mplus. A Practical Guide (1st Edition). Routledge.
https://doi.org/10.4324/9781315642741.
World Health Organization. (2014). Health for the world’s adolescents. WHO Document Production Services: A second chance in the second decade.
Zhou, Y., Zheng, H., Liang, Y., Wang, J., Han, R., & Liu, Z. (2022). Joint developmental trajectories of bullying and victimization from childhood to adolescence: A
parallel-process latent class growth analysis. Journal of Interpersonal Violence, 37(3–4), NP1759–NP1783. https://doi.org/10.1177/0886260520933054.

16

View publication stats

You might also like