Vous êtes sur la page 1sur 13

AGGRESSIVE BEHAVIOR

Volume 40, pages 56–68 (2014)

Moral Disengagement Among Children and Youth:


A Meta‐Analytic Review of Links to Aggressive Behavior
Gianluca Gini1*, Tiziana Pozzoli1, and Shelley Hymel2
1
Department of Developmental and Social Psychology, University of Padua, Padova, Italy
2
Faculty of Education, University of British Columbia, Vancouver, Canada
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
A growing body of research has demonstrated consistent links between Bandura’s theory of moral disengagement and aggressive
behavior in adults. The present meta‐analysis was conducted to summarize the existing literature on the relation between moral
disengagement and different types of aggressive behavior among school‐age children and adolescents. Twenty‐seven independent
samples with a total of 17,776 participants (aged 8–18 years) were included in the meta‐analysis. Results indicated a positive
overall effect (r ¼.28, 95% CI [.23, .32]), supporting the hypothesis that moral disengagement is a significant correlate of
aggressive behavior among children and youth. Analyses of a priori moderators revealed that effect sizes were larger for
adolescents as compared to children, for studies that used a revised version of the original Bandura scale, and for studies with
shared method variance. Effect sizes did not vary as a function of type of aggressive behavior, gender, or publication status. Results
are discussed within the extant literature on moral disengagement and future directions are proposed. Aggr. Behav. 40:56–68,
2014. © 2013 Wiley Periodicals, Inc.
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..

Keywords: aggression; bullying; cyberbullying; moral disengagement; meta‐analysis

INTRODUCTION Helfenfinger, 2010; Menesini, Nocentini, & Camodeca,


2013; Tisak, Tisak, & Goldstein, 2006).
Aggressive behavior toward peers during childhood
The present study focused on one particular type of
and adolescence has been studied for decades (Dodge,
moral reasoning, moral disengagement (henceforth MD),
Coie, & Lynam, 2006) and has been shown to be a
as described in Bandura’s social cognitive theory of
significant correlate, both concurrently and longitudinally,
moral agency (1986, 1990). Specifically, this study
of poor health and maladjustment in both perpetrators and
reports on the first meta‐analytic synthesis of develop-
victims (e.g., Card, Stucky, Sawalani, & Little, 2008; Gini
mental research on the relation between MD and
& Pozzoli, 2009, 2013; Ttofi, Farrington, Losel, &
aggressive behavior in school‐age children and youth.
Loeber, 2011). Although the majority of aggressive
We also explored the factors that might moderate the
children display temporary or desisting aggressive
effect of MD on aggression, including the type of
behavior, about 10% of the general population are
aggressive behavior, participant characteristics (age,
persistently aggressive over the years and can follow a
gender), and methodological features of the studies
deviant “career” path (Pepler, Jiang, Craig, & Connolly,
conducted to date.
2008). Personal correlates and risk factors for youth
aggressive behavior include positive attitudes toward Social Cognitive Theory of Moral Behavior
(Carney & Merrell, 2001; Rigby & Slee, 1993) and high
Bandura (1986, 1990, 1991) focused on moral reason-
self‐efficacy for (Andreou & Metallidou, 2004) the use of
ing and its relation to social behavior in an attempt to
aggression, low empathy (e.g., Gini, Albiero, Benelli, &
Altoè, 2007a; Jolliffe & Farrington, 2011), and high
masculinity (Gini & Pozzoli, 2006). Individual differ- 
Correspondence to: Gianluca Gini, Department of Developmental and
ences in aggression have also been attributed to biases in Social Psychology, University of Padua, via Venezia 8, Padova 35131,
morality, with aggressive behavior linked to distorted Italy. E‐mail: gianluca.gini@unipd.it
moral reasoning that helps to minimize guilt (Arsenio & Received 10 November 2012; Accepted 24 July 2013
Lemerise, 2004; Caravita, Gini, & Pozzoli, 2012; DOI: 10.1002/ab.21502
Hymel, Schonert‐Reichl, Bonanno, Vaillancourt, & Published online 13 September 2013 in Wiley Online Library
Rocke Henderson, 2010; Malti, Gasser, & Gutzwiller‐ (wileyonlinelibrary.com).

© 2013 Wiley Periodicals, Inc.


Moral Disengagement and Aggressive Behavior 57

explain how “good” people can behave “badly”. With developed for adolescents, and a longer scale consisting
age, children develop standards of right and wrong that of 32 items is used with adults (Caprara et al., 1995). The
serve as guides for their conduct. Through this self‐ scale has also been adapted to specific populations (i.e.,
regulatory process, individuals usually act in ways that American minority youth; Pelton, Gound, Forehand, &
give them satisfaction and a sense of self‐worth, and tend Brody, 2004) and it is sometimes used in revised versions
to avoid behaviors that violate their moral standards including subsets of the original items (e.g., Ando,
in order to prevent or minimize self‐condemnation. Asakura, & Simons‐Morton, 2005; Barchia & Bussey,
However, Bandura (2002) argued that the self‐regulation 2011). The scale has shown good reliability (a ¼ .81;
of behavior involves more than just moral reasoning, and Bandura et al., 1996), and is by far the most commonly
that moral reasoning is linked to moral behavior through used measure of MD across countries.
a series of self‐regulatory mechanisms through which A few studies have used different scales of MD.
moral agency is exercised. Moreover, the development of Hymel, Rocke‐Henderson, and Bonanno (2005) also
self‐regulation does not create an invariant control utilized self‐reports to assess MD in children and youth,
system within a person, and there are many psychological although their survey focused specifically on MD
and social processes by which self‐sanctions can be regarding peer bullying. The original 18 items of the
disengaged. Selective activation and disengagement scale were identified “post hoc” from a larger survey
of internal control permit different types of conduct— about bullying as reflecting the four broad categories of
sometimes very negative—with the same moral standards. MD outlined by Bandura (2002). However, factor
Specifically, Bandura described eight mechanisms, analytic results failed to distinguish the four different
clustered into four broad categories through which moral types of MD and instead yielded a single, 13‐item scale
control can be disengaged (see Hymel et al., 2010, for a tapping overall MD with regard to bullying (a ¼ .81).
more detailed discussion). The first, cognitive restructur- Nevertheless, this measure has been adopted by other
ing, operates by framing the behavior itself in a positive researchers (Almeida, Correia, Marinho, & Garcia, 2012;
light, by (i) portraying immoral conduct as warranted Vaillancourt et al., 2006). Although the Bandura and
(moral justification); (ii) contrasting a negative act with Hymel et al. scales show a significant degree of
worse conduct (advantageous comparison); or (iii) using conceptual overlap, a recent study considering a subset
language which palliates the condemned act, thus of items from each scale indicated a moderate association
diminishing its severity (euphemistic labeling). The between the two (r ¼ .51; Ribeaud & Eisner, 2010), with
second set of disengagement strategies operates by the Bandura scale tapping a broader range of MD beliefs
obscuring or minimizing one’s agentive role in the harm and the Hymel et al. scale tapping a more restricted set of
caused (displacement or diffusion of responsibility). The MD beliefs about peer bullying.1
third set of strategies operates by minimizing, disregard-
Moral Disengagement and Different Types of
ing or distorting the consequences of one’s action,
Aggressive Behavior
allowing individuals to distance themselves from the
harm caused or to emphasize positive rather than Starting from early age, individuals who morally
negative outcomes (minimizing or misconstruing con- disengage may perceive some types of antisocial
sequences). Finally, negative feelings can be avoided by behavior as reasonable or justified, at least under some
stripping the recipients of detrimental acts of human circumstances, even if they have internalized moral rules
qualities (dehumanization) or considering aggression as that prohibit such behavior. Indeed, research has shown
provoked by the victim (attribution of blame). These that children and youth who endorse these mechanisms
mechanisms can lead to aggressive behaviors through a are more likely to engage in both general aggression (e.g.,
process of MD, that is a partial gap between the “abstract” Bandura et al., 1996; Caprara et al., 1995) and peer
personal idea of moral behavior and the individual’s real bullying (e.g., Gini, 2006; Hymel et al., 2005). Impor-
life behavior. In this way, the individual protects him/ tantly, the link between MD and aggressive behavior
herself from negative feelings, such as guilt or shame, remains significant even after other predictors of such
that usually follow immoral conduct (Bandura, 1991). behavior, such as aggression efficacy, rule perception, or
parenting, are controlled (e.g., Barchia & Bussey, 2011;
Measures of Moral Disengagement
Caravita & Gini, 2010; Pelton et al., 2004). Interestingly,
Bandura et al. were the first to develop self‐report scales
to measure proneness to MD (Bandura, Barbaranelli, 1
Caprara, & Pastorelli, 1996; Caprara, Pastorelli, & Menesini et al. (2003) have investigated MD by assessing children’s
attributions of morally engaged emotions (guilt, shame) versus disengaged
Bandura, 1995). A short version (14 items) of his moral emotions (pride, indifference) in response to hypothetical moral trans-
disengagement scale has been adapted for use with gressions. In this meta‐analysis we did not include that study because its
elementary school children, a 24‐item version has been methodology differs importantly from those considered here.

Aggr. Behav.
58 Gini et al.

MD has been shown to be a significant correlate of these significant decline, (ii) a normative group (44.5%) with
behaviors in juvenile delinquents samples (Hodgdon, initially moderate levels that later declined, (iii) a “later
2010; Kiriakidis, 2008; Shulman, Cauffmann, Piquero, desistent” group (6.9%) that started with initially high‐
& Fagan, 2011), representing extremely violent individ- medium levels followed by a significant increase from
uals, as well as community samples, thus confirming that ages 14 to 16 and an even steeper decline from ages 16 to
MD mechanisms operate within the “normal” range of 20, and (iv) a “chronic” group (10.7%) that maintained
psychological functioning (Bandura, 1986). constant medium‐high levels of MD. Importantly, youth
Of recent interest is the degree to which MD is who maintained high levels of MD were more likely to
associated with cyberbullying, defined as aggressive engage in aggressive acts in later adolescence. At least
behavior perpetrated via information and communication one study has reported age differences in mean levels of
technologies, such as the Internet and mobile phones MD favoring older students (e.g., Barchia & Bussey,
(Smith et al., 2008). Several authors have suggested that 2011), whereas others reported no significant age differ-
MD might be less evident with cyberbullying, albeit for ences (e.g., Bandura et al., 1996; Pornari & Wood, 2010).
different reasons. Pornari and Wood (2010), for example, However, these studies did not assess directly whether
suggest that online aggression may not demand the same age‐groups differed in the link between MD and
level of rationalization and justification as traditional aggressive behavior. This meta‐analysis tested whether
aggression because youngsters might consider cyberbul- the relation between MD and aggression varied as a
lying as less serious than traditional forms of aggression function of age, comparing studies of children versus
and “the anonymity, the distance from the victim, and the adolescents. Consistent with Bandura’s idea that moral
consequences of the harmful act do not cause so many disengagement develops over time as a result from
negative feelings (e.g., guilt, shame, self‐condemnation), behaving in contrast to internal moral values, it was
and reduce the chance of empathizing with the victim” (p. expected that the relation between MD and aggression
89). Others (e.g., Bauman, 2010; Perren & Sticca, 2011) would be stronger in adolescence as compared to
argue that the inability to observe the immediate reaction childhood.
of the victim may allow the aggressor to minimize the Regarding sex differences, higher levels of self‐
impact of the negative behavior; this would make MD exonerating mechanisms have consistently been found
less necessary. Indeed, the “online disinhibition effect” in male as compared to female samples from different
(Suler, 2004), which refers to a loosening of social/moral cultural contexts, even after controlling for other
restrictions and inhibitions during online interaction that demographic variables, such as ethnicity or socio‐
would otherwise be present in face‐to‐face interaction, economic status (Bandura et al., 1996; Obermann,
itself can represent a variation of MD, allowing the 2011; Yadava, Sharma, & Gandhi, 2001). Less clear is
individuals to behave in ways that are contrary to their whether the magnitude of the relation between MD and
moral code. Drawing upon this literature, our first aim aggression varies across boys and girls. Some authors
was to evaluate the strength of the association between have suggested stronger links for boys (Bussman, 2007;
MD and any form of peer‐directed aggressive behavior Paciello et al., 2008), others report the reverse (Yadava
among school‐age children and youth. Of additional et al., 2001), and still others find no moderating effect of
interest was whether the magnitude of this effect varied as gender (Gini, 2006; Obermann, 2011). Accordingly, the
a function of the behavior considered (e.g., aggression vs. present meta‐analysis tested directly whether sex
bullying vs. cyberbullying). moderated the link between MD and aggressive behavior.
Discrepancies in findings across studies as a function
Testing Potential Moderators
of methodological differences were also considered.
Three categories of potential moderators were hypoth- First, because different scales to measure MD exist, we
esized to influence the relationship between MD and tested whether effect sizes varied as a function of the type
various forms of aggressive behavior. First considered of instrument, by comparing studies that used the original
are participant characteristics, specifically age and sex. scale devised by Bandura, studies that used a revised
Previous longitudinal research by Paciello, Fida, Tra- shortened version of that scale, and studies that employed
montano, Lupinetti, and Caprara (2008) examined other scales. Second, although MD is always assessed
stability and change in MD and its relation to aggressive through self‐reports, studies differ in their assessments of
behavior among 366 Italian adolescents, followed at four children’s aggressive behavior, with self, peers, and
time points from 14 to 20 years. Although generally MD adults (teachers, parents) used as sources of information.
appeared to decline with age, especially between ages 14 As demonstrated in a previous meta‐analysis by Hawker
and 16, four distinct trajectories were identified: (i) non‐ and Boulton (2000) on the relations between peer
disengaged adolescents (37.9% of the sample) who victimization and psychosocial adjustment, use of the
initially showed low levels of MD followed by a same informant for both constructs can inflate the

Aggr. Behav.
Moral Disengagement and Aggressive Behavior 59

magnitude of the measured effects due to shared method behavior from a broader measure of externalizing
variance. Accordingly, this meta‐analysis tested whether problems at the age of 15 and this study was then
this feature accounted for differences in effect sizes included in the present meta‐analysis. Second, eligible
across studies. studies were required to have enough quantitative
Finally, publication bias is a threat to any meta‐analytic information to calculate effect sizes. Therefore, studies
review, with concern that unpublished studies are more based on interviews or open‐ended questions were
likely to have smaller or non‐significant results and less excluded. Third, study participants were school‐age
likely to be included in a meta‐analysis than published children or adolescents from the community, with studies
studies, yielding estimated effect sizes larger than those involving clinical samples or incarcerated offenders, and
that actually exist. To reduce publication bias, efforts studies of adults excluded. Finally, both published reports
were made to include as many unpublished studies as (i.e., journal articles) and unpublished studies (e.g.,
possible (e.g., dissertations, conference papers). To check conference papers, doctoral theses) were considered. In
whether a significant difference existed between pub- the latter case, data were obtained from the principal
lished and unpublished studies in the reported effect investigator or his/her supervisor. When multiple reports
sizes, we also tested for the moderating effect of (e.g., a conference paper or dissertation and a published
publication status. article) presented results from the same sample, only one
effect size was used in the meta‐analysis. Using these
inclusion criteria, the final sample of the current meta‐
METHODS
analysis included 27 studies; 12 examined the relation
Literature Search between MD and general aggression, 11 considered MD
and bullying and four considered MD and cyberbullying
Multiple methods were used to identify potentially
(see Table I).
eligible studies. First, computer literature searches from
All studies were coded independently by the first and
the year each database started until March 2012 were
the second author, using an a priori coding scheme,
conducted using PsychInfo, Educational Research
recording authors and year of publication, the type and
Information Center, Scopus and Google Scholar with
form of MD and aggression measures used (self‐report
“moral disengagement,” “aggressive behavior,” “aggres-
vs. peer/adult reports), sample size, national setting, and
sion,” “bullying,” “school violence,” “antisocial behav-
demographic characteristics of participants (age, gender).
ior” used as keywords. Second, recent review articles and
Inter‐rater agreement was found to be very good; all
book chapters on aggressive behavior, bullying, or
Cohen’s kappas exceeded .92. Discrepancies were
morality in children were reviewed for relevant citations.
resolved by discussion.
Third, reference sections of the collected articles were
searched for relevant earlier references (i.e., “backward Data Analysis
search” procedure). Finally, authors were contacted
Pearson’s r was used as the effect size metric, because
directly to obtain other relevant studies. With unpub-
almost all studies provided zero‐order correlation
lished studies (conference papers, dissertations), princi-
coefficients between the constructs of interest. In three
pal investigators were contacted to ask for ad hoc analysis
cases (Bacchini, Amodeo, Ciardi, Valerio, & Vitelli,
(if no response was received, a second e‐mail was sent
1998; Del Bove, Caprara, Pastorelli, & Paciello, 2008;
2–3 months after the first). A total of 70 potentially
Hymel et al., 2005), the effect size was calculated from
relevant journal articles, chapters, conference and
the comparison between a group of aggressive children
dissertation abstracts were reviewed.
and a control (non‐aggressive) group, by first calculating
the standardized mean difference and then converting it
Inclusion Criteria
into r (for details see Borenstein, Hedges, Higgins, &
The most basic requirement for inclusion in the Rothstein, 2009, p. 48; Card, 2011, p. 100–101). Most
present meta‐analysis was consideration of measures of studies did not report data separately for boys and girls.
Bandura’s MD mechanisms and any form of aggressive Given that one aim of the study was to test the possible
behavior, or bullying, or cyberaggression/cyberbullying, moderating role of gender, authors were contacted and
including self‐report questionnaires, as well as peer‐, asked for ad‐hoc analyses. This resulted in 45 correlation
parent‐, or teacher‐reports. Studies were excluded if the coefficients disaggregated by gender (Hyde et al.’s study
aggression items were part of a wider measure (e.g., a only provided boys’ effect size). In four cases for a given
scale measuring externalizing problems) and a separate study we had two independent effect sizes for each
effect size was not available. In one case (Hyde, Shaw, & gender (e.g., for primary school boys/girls and for middle
Moilanen, 2010), the original author was able to school boys/girls). However, the very small numbers of
calculate, upon request, the effect size for aggressive these subgroups did not allow for a more detailed

Aggr. Behav.
60 Gini et al.

TABLE I. Summary of Studies Included in the Meta‐Analysis


Shared
Sample Size Age National Method Effect
Authors (Year) (% of Girls) Range Setting Behavior Measure Variance Size: r

Almeida, Correia, Marinho, 499 (47.1%) 11–18 Spain Cyberbullying, SR Yes .28
and Garcia (in press)
Ando et al., (2005) 2,301 (49.8%) 12–15 Japan Bullying, SR Yes .25
Bacchini et al. (1998) 169 (46.9%) 9–14 Italy Bullying, SR Yes .16
Bandura et al. (1996) 799 (45.2%) 10–15 Italy Aggression, SR, PN, TR, PR Mixed .27
Barchia and Bussey (2011) 1,285 (53.8%) 12–15 Australia Aggression, SR Yes .27
Bauman (2010) 190 (54.2%) 10–14 United States Cyberbullying, SR Yes .32
Bussey and Quinn (2012) 1,152 (37.2%) 12–17 Australia Aggression Yes .47
Bussman (2008, study 2) 136 (52.9%) 9–12 United States Aggression, PN No .16
Caprara et al. (1995) 706 (43.6%) 8–14 Italy Aggression, SR, PN, TR Mixed .20
Caravita and Gini (2010) 538 (46.6%) 9–15 Italy Bullying, PN No .18
Caravita, Gini, and Pozzoli, (2011) 879 (47.4%) 8–15 Italy Bullying, PN No .20
Del Bove et al. (2008) 475 (45.1%) 11–18 Italy Aggression, SR Yes .22
Fitzpatrick and Bussey (2012) 708 (57.1%) 12–16 Australia Bullying, SR Yes .31
Gini (2006) 581 (49.2%) 8–11 Italy Bullying, PN No .22
Gini et al. (2007b) 1,084 (50.9%) 15–17 Italy Bullying, PN No .27
Gini, Pozzoli, and Hauser (2011) 719 (48.5%) 9–13 Italy Bullying, PN No .13
Hyde et al. (2010) 257 (0%) 15 United States Aggression, PR No .20
Hymel et al. (2005) 468 (43%) 13–16 Canada Bullying, SR Yes .59
Menesini, Fonzi, and Vannucci (1999) 652 (48.2%) 8–14 Italy Bullying, PN No .14
Obermann (2011) 677 (47.6%) 11–14 Denmark Bullying, SR, PN Mixed .22
Paciello et al. (2008) 349 (53.3%) 12–14 Italy Aggression, PN No .17
Pelton et al. (2004) 245 (49.4%) 9–14 United States Aggression, SR, TR, PR Mixed .13
Perren and Sticca (2011) 480 (48.9%) 12–18 Switzerland Cyber/Bullying, SR Yes .42
Pornari and Wood (2010) 359 (53%) 12–14 UK Cyber/Aggression, SR Yes .40
Qingquan, Zongkui, Fan, and Lei (2009) 1,578 (48%) 9–11 China Aggression No .23
Stevens and Hardy (in press) 290 (60.3%) 13–18 Samoa Aggression Yes .54
Yadava et al. (2001) 200 (50%) 15–17 India Aggression Yes .30

Note. Measures of moral disengagement were all self‐reports. SR, self‐report; PN, peer nominations, TR, teacher‐report; PR, parent‐report.

analyses and effect sizes were thus pooled by gender use of correlation coefficients can result in problematic
group. In order to avoid violation of the assumption of error formulation, the correlation coefficient for each
independence, mean effect sizes for the total sample were study was converted to the Fisher’s z scale, and all
calculated for those studies reporting multiple effect sizes analyses were performed using the transformed values
(e.g., two or more informants for the same behavior) (Lipsey & Wilson, 2000; Rosenthal, 1991). Then, the
(Becker, 2000; Borenstein et al., 2009). resulting summary effect and its confidence interval were
Outlying effect sizes and sample sizes were identified converted back to correlations for ease of interpretation.
on the basis of standardized z values larger than 3.29 or A 95% confidence interval (CI) was computed around
smaller than 3.29 (e.g., Tabachnick & Fidell, 2001). No each mean effect size. Confidence intervals not including
outliers were detected for effect size, but there was one zero were interpreted as indicating a statistically
study with an outlying sample size (Ando et al., 2005). detectable result favoring the association between MD
We winsorized this number of participants (i.e., reduced and aggressive behavior.
to the next largest sample size, following Barnett & Heterogeneity was assessed using the Q statistic
Lewis, 1994; Lipsey & Wilson, 2000), resulting in an (which is distributed as x2 with df ¼ k  1, where k
N of 1,578. represents the number of effect sizes; Lipsey &
Data from individual studies were pooled (with Wilson, 2000), evaluating whether the pooled studies
comprehensive meta‐analysis program—v.2.2) using a represented a homogeneous distribution of effect sizes.
random effects model. To account for variations in Significant heterogeneity indicates that variations in
sample size, which influences precision with larger effect sizes are likely due to sources other than sampling
samples yielding more precise estimates than smaller error (e.g., study characteristics). Also reported is the I2
samples, each study was weighted by the inverse of its statistic, indicating the proportion of observed variance
variance (Hedges & Olkin, 1985). Moreover, because the that reflects real differences in effect size (Higgins,

Aggr. Behav.
Moral Disengagement and Aggressive Behavior 61

Thompson, Deeks, & Altman, 2003). Moderator analy- correlation can be interpreted as absence of publication
ses were conducted to examine this variability. bias (Begg & Mazumdar, 1994).
Even though we were able to include several
unpublished studies, we evaluated the potential “publi-
cation bias” in different ways. We computed the “fail‐safe RESULTS
N” (Nfs) according to the method proposed by Orwin
Association Between MD and Aggressive
(1983), which is more conservative than the traditional
Behavior
Rosenthal’s Nfs (Rosenthal, 1978, 1979). Orwin’s Nfs
determines the number of additional studies yielding null The 27 studies examining the association between
results that would be needed to reduce meta‐analytic MD and aggressive behavior reported data on 17,776
results to a negligible result of .05 (Durlak & participants aged 8–18. The distribution of effect sizes is
Lipsey, 1991). We also inspected the funnel plot, which presented in Figure 1. Under the random effects model,
displays effect sizes plotted against the sample size, the mean effect size for the association between MD and
standard error, or some other measure of the precision of aggressive behavior was r ¼ .28, which was significantly
the estimate. An unbiased sample of studies would different from zero (Z ¼ 11.06, P <.001), with a 95% CI
ideally show a cloud of data points that is symmetric ranging from .23 to .32. The Nfs of null results needed
around the population effect size (Field & Gillett, 2010). to overturn this significant result suggested that there
Moreover, the association between the effect sizes and would need to be at least 127 studies with a null effect
the variances of these effects was analyzed by rank size added to the analysis before the cumulative effect
correlation with use of the Kendall’s t method. If small would become negligible. In order to evaluate the
studies with negative results were less likely to be existence of publication bias, a funnel plot of standard
published, the correlation between variance and effect errors plotted against effect sizes was developed. The
size would be high. Conversely, lack of significant funnel plot indicated no systematic publication bias, with

Study Statistics for each study Correlation and 95% CI


Lower Upper
Correlation limit limit Z-Value p-Value
Almeida et al. (in press) 0,28 0,20 0,36 6,41 0,00
Ando et al. (2005) 0,25 0,21 0,28 12,09 0,00
Bacchini et al. (1998) 0,16 0,01 0,30 2,09 0,04
Bandura et al. (1996) 0,27 0,21 0,34 7,90 0,00
Barchia & Bussey (2011) 0,27 0,22 0,32 9,99 0,00
Bauman (2010) 0,32 0,19 0,44 4,54 0,00
Bussey & Quinn (2012) 0,48 0,43 0,52 17,51 0,00
Bussman (2007, study 2) 0,16 -0,01 0,32 1,84 0,07
Caprara et al. (1995) 0,20 0,13 0,27 5,30 0,00
Caravita & Gini (2010) 0,18 0,10 0,26 4,21 0,00
Caravita et al. (2011) 0,20 0,11 0,29 4,13 0,00
Del Bove et al. (2008) 0,22 0,14 0,30 4,97 0,00
Fitzpatrick & Bussey (2012) 0,31 0,25 0,38 8,64 0,00
Gini (2006) 0,22 0,15 0,30 5,48 0,00
Gini et al. (2007b) 0,27 0,22 0,33 9,17 0,00
Gini, Pozzoli, Hauser (2011) 0,13 0,05 0,21 3,32 0,00
Hyde et al. (2010) 0,20 0,08 0,31 3,23 0,00
Hymel et al. (2005) 0,59 0,53 0,64 16,06 0,00
Menesini et al. (1999) 0,14 0,06 0,21 3,49 0,00
Obermann (2011) 0,22 0,15 0,30 5,86 0,00
Paciello et al. (2008) 0,17 0,06 0,27 3,14 0,00
Pelton et al. (2004) 0,13 0,00 0,25 2,03 0,04
Perren & Sticca (2011) 0,42 0,34 0,49 9,70 0,00
Pornari & Wood (2010) 0,40 0,31 0,49 7,87 0,00
Qingquan et al. (2009) 0,23 0,18 0,28 9,29 0,00
Stevens & Hardy (in press) 0,54 0,46 0,62 10,64 0,00
Yadava et al. (2001) 0,30 0,16 0,42 4,28 0,00
0,28 0,26 0,29 37,45 0,00
-1,00 -0,50 0,00 0,50 1,00

Favours A Favours B

Fig. 1. Forest plot for random‐effects meta‐analysis of the association between moral disengagement and aggressive behavior. Note. Studies are
represented by symbols whose area is proportional to the study’s weight in the analysis.

Aggr. Behav.
62 Gini et al.

TABLE II. Tests of Categorical Moderators


95% Confidence Interval

Study Characteristics Between‐Group Effect (Qb) Effect Size (r) Lower Limit Upper Limit k N

Type of behavior 2.63


Aggression .27 .20 .34 12 7,472
Bullying .25 .17 .32 11 8,776
Cyberbullying .31 .27 .36 4 1,528
Gender 0.001
Boys .26 .21 .31 23 5,704
Girls .26 .21 .31 22 5,267
Age group 13.47
Children (8–11 years) .18 .14 .22 10 4,201
Adolescents (12–18 years) .31 .26 .36 22 12,326
Type of MD scale 62.80
Bandura’s original .24 .19 .28 16 8,734
Bandura‐revised .31 .23 .38 8 7,939
Others .36 .07 .60 3 1,103
Shared method variance 14.89
Yes .35 .28 .42 13 8,576
No .20 .17 .23 10 6,773
Mixed .24 .15 .33 4 2,427
Publication status 0.14
Published .27 .21 .33 19 11,516
Not published .29 .20 .37 8 6,260
Note. k indicates the number of independent subsamples; N indicates the number of participants.

P <.05.

P <.01.

P <.001.

studies distributed symmetrically about the mean effect interpretability of moderation results, especially in cases
size. This was confirmed by the Egger’s test, which of multiple categories, we report contrasts of correlations
yielded a statistically non‐significant P‐value of 0.48 (with their 95% CIs) between different study categories
(one‐tailed), and by the rank correlation Kendall’s (Bonett, 2008).2
t ¼ .06, P ¼.66. First, we analyzed the existence of any difference in
effect size as a function of the type of behavior measured.
Only four of the studies reported data for cyberbullying
Moderator Effects
and two of them (Perren & Sticca, 2011; Pornari &
The test of homogeneity of variance revealed signifi- Wood, 2010) reported data for both traditional
cant heterogeneity across studies, Q ¼ 271.05, P <.001, aggression/bullying and cyberbullying from the same
I2 ¼ 90.41%. Therefore, mixed effects moderator analy- sample. To avoid dependence of data, we first ran a
ses were conducted to examine the association between moderation analysis excluding the effect sizes about
study characteristics and their effect sizes. The mixed cyberbullying and comparing the 12 studies that
effects model assumes that the variability in effect sizes measured aggression with the 11 studies that measured
consists of systematic variance (that can be statistically bullying. The mean effect size in these two subgroups
modeled), sampling error, and an additional, unexplain- was very similar (r ¼ .27 and .25, respectively).
able random component. Using the random effects model Subsequently, the subgroup of studies assessing cyber-
to combine studies within subgroups in the moderator bullying was included in the analysis (with the exclusion
analyses, a mixed effects model typically allows for of relative MD‐aggression/bullying effect size), yielding
population parameters to vary across studies, reducing a non‐significant between‐group difference: Q(2) ¼ 2.63,
the probability of Type I error, and is usually regarded as P ¼.27. The contrast between bullying and cyberbully-
a more rigorous meta‐analytical model than a fixed effect ing effect sizes yielded a value of .06, 95% CI
model only (Borenstein et al., 2009; Hedges & Vevea,
1998). The results of the moderator analyses are
summarized in Table II. Furthermore, to enhance 2
We thank an anonymous reviewer for this suggestion.

Aggr. Behav.
Moral Disengagement and Aggressive Behavior 63

[.15, .02]. The contrast between aggression and 95% CI [.07, .23], whereas the difference between
cyberbullying yielded a value of .04 [.13, .04]. the former and studies with mixed method was .11
A second moderator analysis compared the effect sizes [.004, .22]. Finally, mean effect sizes did not vary as a
observed for boys versus girls. Interestingly, the effect of function of publication status (Q(1) ¼ 0.14, P ¼.71;
sex was not significant, with the effect sizes from the two contrast: .02, [.12, .09]).
sex groups being identical. Potential age differences
were tested by comparing effect sizes observed for
DISCUSSION
children (i.e., 8‐ to 11‐year olds) versus adolescents (i.e.,
12‐ to 18‐year olds). Four studies provided two A meta‐analytic review of 27 independent studies was
independent effect sizes each, one for children and one conducted to summarize current research on the link
for adolescents. Two studies (Bacchini et al., 1998; between MD and aggressive behavior in children and
Bauman, 2010) were excluded because separate effect adolescents and to test whether differences in the reported
sizes for the two age‐groups were not available. The effects can be explained by the type of aggressive
analysis revealed a significant difference between the two behavior considered, characteristics of the participants,
age‐groups favoring adolescents: Q(1) ¼ 13.47, P <.001 and/or methodological features of the studies. Of interest
(contrast: .13, 95% CI [.19, .06]). Because some was computation and interpretation of effect sizes in
studies included a wider range of adolescent ages (until order to determine whether the magnitude of the effect
18 years) compared to others, and Paciello et al. (2008) represents something psychologically important. In
have reported changes in MD particularly during middle‐ doing so, one possibility is to compare a computed
adolescence (14–16‐years), a sensitivity analysis was effect size with “standard” cut‐off criteria. For example,
performed. Exclusion of the samples with participants Cohen (1992) proposed conventional values as bench-
older than 15 (Bussey & Quinn, 2012; Del Bove marks for what are considered to be “small,” “medium,”
et al., 2008; Fitzpatrick & Bussey, 2012; Gini, Albiero, and “large” effects (r: .1, .3, and .5, respectively). More
Benelli, Matricardi, & Pozzoli, 2007b; Perren & recently, based on empirical findings, Hemphill (2003)
Sticca, 2011; Stevens & Hardy, 2013; Yadava et al., recommended a reconceptualization of effect sizes in
2001) did not change this result (Q(1) ¼ 4.48, P ¼.03, psychological research, in which r ¼ .1 is “small,” r ¼ .2
with adolescents’ effect size r ¼ .27). In sum, the is “medium,” and r ¼ .3 is “large” (see also Huang, 2011).
associations observed between aggression/bullying and These benchmarks, however, have been criticized
MD did not vary as a function of gender but were stronger because they are purely conventional, and somewhat
among adolescents, as compared to children. arbitrary, whereas practical and clinical importance
Moderation analysis by type of MD scale yielded a depends on the situation researchers are dealing with
significant between‐group difference: Q(2) ¼ 62.80, (e.g., Kline, 2004; Thompson, 2002). A preferable
P <.001. The contrast between studies that used the solution is to put one effect size into a meaningful
original scale devised by Bandura and those that used a context, comparing it to other effects that have been
revised version of that scale yielded a value of .07, 95% reported within the same literature and are commonly
CI [.16, .02], indicating that the effect tended to be considered important. Such an approach is especially
slightly higher in the latter studies. The contrast between needed when dealing with a multi‐causal phenomenon
studies that used the original scale and those that used a such as aggressive behavior, where one should not expect
different MD scale yielded a value of .12 [.36, .17]. any single factor to explain much of the variance
Finally, the contrast between studies that used a revised (Anderson et al., 2010).
version of Bandura’s scale and those that used a different The composite effect size yielded by the present meta‐
scale yielded a value of .05 [.30, .25]. analysis was small‐to‐medium according to Cohen’s
In order to consider the possible effect of shared criteria, and medium‐to‐large according to Hemphill’s
method variance, we distinguished studies with shared criteria. A qualitative comparison of our findings with
method variance, studies with no shared method available meta‐analyses on individual risk‐factors and
variance, and studies that employed both self‐reports correlates of aggressive behavior in children and
and other informants to measure aggression (“mixed” adolescents indicates that the MD‐aggressive behavior
studies). The analysis revealed a significant between‐ link is equal—in absolute value—to the association
group heterogeneity: Q(2) ¼ 14.89, P ¼.001. Studies between other‐related cognitions (i.e., children’s
with shared method variance reported significantly thoughts, beliefs, or attitudes about others, including
higher effect sizes relative to those that employed normative beliefs about others, empathy, and perspective
different informants (Q(1) ¼ 14.73, P <.001). The taking) and bullying (r ¼ .27; Cook, Williams, Guerra,
contrast between studies with shared method variance Kim, & Sadek, 2010). Moreover, the current result is
and studies without this problem yielded a value of .15, larger than the associations reported between hostile

Aggr. Behav.
64 Gini et al.

attributions and aggression (r ¼ .17; Orobio de Castro, tudinal studies (Barchia & Bussey, 2011; Hyde et al.,
Veerman, Koops, Bosch, & Monshouwer, 2002), be- 2010; Paciello et al., 2008) that differed considerably in
tween emotion knowledge and externalizing problems design. As a result, we did not feel that the studies were
(r ¼ .17; Trentacosta & Fine, 2010), and larger than sufficiently comparable to include analyses comparing
effect sizes previously reported for other individual longitudinal and cross‐sectional studies. Efforts to tease
predictors of bullying (Cook et al., 2010), such as social apart variations in moral development and aggression as
competence (r ¼ .15), self‐related cognitions (e.g., self‐ a function of age, time, or cohort may be a particularly
esteem, self‐efficacy, r ¼ .09), social problem‐solving fruitful focus in future research.
(r ¼ .18), and academic performance (r ¼ .18). Notably, effect sizes did not significantly vary as a
In sum, the link between MD and aggressive behavior function of type of aggressive behavior considered
is both statistically significant and practically (aggression vs. bullying vs. cyberbullying). In the case of
meaningful. cyberbullying, however, the estimated correlation with
As expected, significant heterogeneity across effect MD was slightly higher than that observed for traditional
sizes was also observed, and some significant a priori aggression/bullying. The contrasts of correlations sug-
moderators were identified. Estimated effect size was gest that cyberbullying may have slightly stronger links
higher for older participants than for children, indicating with MD, given the CI just hugs zero. However, this
the existence of a developmental change in the link analysis was limited by the small number of studies that
between MD and aggressive behavior. As noted in the examined the relation between MD and cyberbullying,
introduction, Paciello et al. (2008) documented different which were limited to adolescent samples and were
developmental trajectories of MD during adolescence characterized by low precision that resulted in a quite
and found that higher levels of MD increased risk for wide confidence interval. Future studies are certainly
youth aggression, although such changes were not needed to further explore the issue of moral justifications
observed for children. The stronger relation observed in online/virtual aggressive relationships.
between MD and aggression among adolescents relative Results of the present meta‐analysis also revealed that
to children is also consistent with Bandura’s description the correlation between MD and aggressive behavior did
of MD as a gradual process: “disengagement practices not differ significantly across boys and girls. Even though
will not instantly transform considerate people into cruel absolute levels of moral disengagement and aggressive
ones. Rather, the change is achieved by progressive behavior are often higher in boys, the relation between
disengagement of self‐censure. Initially, individuals the two variables is identical. These results should be
perform mildly harmful acts they can tolerate with viewed with caution, however, given that the small
some discomfort. […] The continuing interplay between number of studies in each cell did not allow for a more
moral thought, affect, action and its social reception is thorough analysis and we cannot rule out the possibility
personally transformative” (Bandura, 2002, p. 110). In that gender differences do not exist at different age‐
other words, one would expect disengaged justifications levels. Results of the present meta‐analysis are important
and moral transgressions to reinforce each other over in guiding the design of future studies testing specific
time (Bandura et al., 1996; Gibbs, Potter, & Goldstein, hypotheses regarding sex differences.
1995). For chronically disengaged adolescents, however, Regarding the possible effect of methodological
MD could represent a “strategy of adaptation that is differences, the link between MD and aggression was
embedded into a system of beliefs about the self and moderated by type of MD scale, with the mean effect
others and leads to perceive aggression and violence as being slightly larger when MD was measured through a
appropriate means to pursue one’s own goals” (Paciello revised version of Bandura’s scale as compared to the
et al., 2008, p. 1302). Understanding the nature and original scale. This difference may be due to the fact that,
mechanisms underlying this developmental shift is an at least in some cases, the revised scale was designed to
important focus in future research. To this end, it retain items more explicitly related to disengagement for
becomes important to develop techniques to assess MD aggressive acts. In addition, as expected, shared method
in children below 8 years of age, in order to study when variance was found to be a significant moderator, with
such distortions emerge and their relation to moral somewhat larger effect sizes observed when the same
development (e.g., the emergence of the distinction informant (participating child/youth) evaluated both MD
between moral and social‐conventional rules) and with and aggressive behavior. Unfortunately, the very small
environmental factors, such as parenting or early number of studies using adult informants precluded any
experiences with peers. Finally, future research would further comparisons. Finally, effect sizes did not differ for
benefit from comparisons of developmental changes as published versus unpublished studies, supporting the
documented in cross‐sectional versus longitudinal re- validity of the current meta‐analytic estimations, which
search. The present meta‐analysis included three longi- were not inflated by publication bias.

Aggr. Behav.
Moral Disengagement and Aggressive Behavior 65

Limitations and Future Directions observed effect sizes and should be used to interpret
those effects, it is important that measurement studies as
Despite the strengths of the present meta‐analysis, it
well as substantive studies systematically report reliabil-
also presents limitations due to the characteristics and
ity coefficients for their samples. A lack of data also
quality of the primary studies. One limitation deals with
prevented the consideration of additional potential
the limited capacity to measure single mechanisms in a
moderators. For example, the samples generally included
reliable and valid manner (e.g., the four broad categories
participants from a variety of SES and ethnic/racial
of MD strategies). Indeed, studies to date have treated
groups, yet no studies reported the association separately
MD as a unidimensional construct (see Pozzoli, Gini, &
for the different groups. Moreover, although the positive
Vieno, 2012, for an exception) and we know nothing
association between MD and aggressive behavior is
about whether different MD mechanisms act differently
established, in reviewing studies for this meta‐analysis,
in aggressive behavior nor whether their relative
the lack of research investigating moderators of the
importance varies as a function of age. Conceptually,
association between MD and aggression was readily
this may suggest that the psychological function of the
apparent. Although some studies have shown that the
MD process—to free the individual from self‐censure
link between MD and aggressive behavior is significant
and potential guilt—is more important than the specific
even after the role of other variables are accounted for
strategy used to achieve the individual’s self‐serving goal
(e.g., Barchia & Bussey, 2011; Caravita & Gini, 2010;
(e.g., justifying the negative behavior enacted vs.
Pelton et al., 2004), little is known about how moral
blaming the victim). Still, a better understanding of
disengagement interacts with other individual risk
whether and under what circumstances these mecha-
factors, particularly personality characteristics that
nisms can be differently activated may have both
make some youth more likely to engage in antisocial
theoretical and practical implications. In this regard,
conduct, such as aggression. Overall, the individual and
future research may benefit from consideration of new
contextual factors that may buffer or exacerbate the
research paradigms and methodologies, in both labora-
relation of MD and aggression remains unclear. Given
tory and field studies. Future research would also benefit
the current findings, it is time to move from “main effect”
from further consideration of other ways to conceptualize
studies, aimed at establishing a relation between MD
and measure MD. Particularly promising here are studies
and aggressive behavior, to “interaction effect” studies,
examining how morally disengaged emotions, such as
testing specific hypotheses and more complex patterns of
pride or indifference (instead of guilt or shame) following
relations.
an aggressive act influence subsequent behavior. Efforts
In conclusion, this study presents the first meta‐
to expand and integrate the two approaches into a
analytic synthesis of the research on the relation between
coherent theoretical model of MD would be welcomed.
MD and aggressive behavior in school‐age children and
Another limitation is that the directionality of the
adolescents. Our results showed that MD can be
association between MD and aggressive behavior is not
considered one major correlate of aggressive behavior,
clear, and bidirectionality might be the rule instead of the
and that this relation is moderated by age and shared
exception. Indeed, MD mechanisms are likely to
method variance. It is clear that additional methodologi-
influence aggressive behavior over time (Hyde et al.,
cally strong studies are needed to have a more complete
2010; Paciello et al., 2008), but their activation may
understanding of these factors, especially in relation to
also be made easier by repeated immoral acts (e.g.,
developmental processes and contextual influences on
Bandura, 1990) and frequent exposure to an aggressive
MD, that will provide useful information for prevention
environment can alter children’s evaluation of moral
and intervention efforts.
transgression (Ardila‐Rey, Killen, & Brenick, 2009).
Unfortunately, this domain of research is still largely
dominated by correlational studies that preclude definite REFERENCES
conclusions about direction of effects and causality in
general. Intervention studies investigating whether References marked with an asterisk indicate
changing proneness to MD also changes children’s studies included in the meta‐analysis
use of aggression could help to clarify the issue of

directionality. Almeida, A., Correia, I., Marinho, S., & Garcia, D. (2012). Virtual but not
Our analyses were limited by the methodological less real: A study of cyberbullying and its relations to moral
disengagement and empathy. In Q. Li, D. Cross, & P. K. Smith
quality of the studies available and we were not able to
(Eds.), Cyberbullying in the global playground: Research from
account for the reliability of the measures used in international perspectives. (pp. 223–244) Chichester: Wiley‐Blackwell.
the studies because such information was not available Anderson, C. A., Shibuya, A., Ihori, N., Swing, E. L., Bushman, B. J.,
in many cases. Because score reliability influences Sakamoto, A., Rothstein, H. R., & Saleem, M. (2010). Violent video

Aggr. Behav.
66 Gini et al.

game effects on aggression, empathy, and prosocial behavior in Eastern Caprara, G. V., Pastorelli, C., & Bandura, A. (1995). Measuring age
and Western countries: A meta‐analytic review. Psychological Bulletin, differences in moral disengagement. Eta Evolutiva, 51, 18–29.

136, 151–173. doi: 10.1037/a0018251 Caravita, S. C. S., & Gini, G. (2010 , March). Rule perception or moral

Ando, M., Asakura, T., & Simons‐Morton, B. (2005). Psychosocial disengagement? Associations of moral cognition with bullying and
influences on physical, verbal, and indirect bullying among Japanese defending in adolescence. Paper presented at the SRA Biennial
early adolescents. Journal of Early Adolescence, 25, 268–297. doi: Meeting. Philadelphia, USA.

10.1177/0272431605276933 Caravita, S. C. S., Gini, G., & Pozzoli, T. (2011 , March, April). Moral
Andreou, E., & Metallidou, P. (2004). The relationship of academic and disengagement for victimization in bullying: Individual or contextual
social cognition to behavior in bullying situations among Greek processes? Paper presented at the SRCD Biennial Meeting. Montreal,
primary school children. Educational Psychology, 24, 27–41. doi: Canada.
10.1080/0144341032000146421 Caravita, S. C. S., Gini, G., & Pozzoli, T. (2012). Main and moderated
Ardila‐Rey, A., Killen, M., & Brenick, A. (2009). Moral reasoning in effects of moral cognition and status on bullying and defending.
violent contexts: Displaced and non‐displaced Colombian children’s Aggressive Behavior, 38, 456–468.
evaluations of moral transgressions, retaliation, and reconciliation. Social Card, N. A. (2011). Applied meta‐analysis for social science research. New
Development, 18, 181–209. doi: 10.1111/j.1467‐9507.2008.00483.x York: Guilford Press.
Arsenio, W. F., & Lemerise, E. A. (2004 , June). Aggression and moral Card, N. A., Stucky, B. D., Sawalani, G. M., & Little, T. D. (2008). Direct
development: Towards an integration of the social information and indirect aggression during childhood and adolescence: A meta‐
processing and moral domain models. Poster presented at meeting analytic review of gender differences, intercorrelations, and relations to
of Jean Piaget Society, Toronto, Canada. maladjustment. Child Development, 79, 1185–1229. doi: 10.1111/

Bacchini, D., Amodeo, A. L., Ciardi, A., Valerio, P., & Vitelli, R. (1998). j.1467‐8624.2008.01184.x
La relazione vittima‐prepotente: stabilità del fenomeno e ricorso a Carney, A. G., & Merrell, K. W. (2001). Perspectives on understanding and
meccanismi di disimpegno morale. Scienze dell’Interazione, 5, 29–46. preventing an international problem. School Psychology International,
Bandura, A. (1986). Social foundations of thought and action: A social 22, 364–382. doi: 10.1177/0143034301223011
cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.
Bandura, A. (1990). Selective activation and disengagement of moral doi: 10.1037/0033‐2909.112.1.155
control. Journal of Social Issues, 46, 27–46. doi: 10.1111/j.1540‐ Cook, C. R., Williams, K. R., Guerra, N. G., Kim, T. E., & Sadek, S. (2010).
4560.1990.tb00270.x Predictors of bullying and victimization in childhood and adolescence.
Bandura, A. (1991). Social cognitive theory of moral thought and action. School Psychology Quarterly, 25, 65–83. doi: 10.1037/a0020149

In W. M. Kurtines, & G. L. Gewirtz (Eds.), Handbook of moral Del Bove, G., Caprara, G. V., Pastorelli, C., & Paciello, M. (2008).
behavior and development: theory, research and applications (Vol. 1, Juvenile firesetting in Italy: Relationship to aggression, psychopathol-
pp. 71–129). Hillsdale, NJ: Erlbaum. ogy, personality, self‐efficacy, and school functioning. European Child
Bandura, A. (2002). Selective moral disengagement in the exercise of & Adolescent Psychiatry, 17, 235–244. doi: 10.1007/s00787‐007‐
moral agency. Journal of Moral Education, 312, 101–119. doi: 0664‐6
10.1080/0305724022014322 Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression and antisocial

Bandura, A., Barbaranelli, C., Caprara, G. V., & Pastorelli, C. (1996). behaviour in youth. In W. Damon, & R. M. Lerner (Series, Eds.) & N.
Mechanisms of moral disengagement in the exercise of moral agency. Eisenberg (Vol. Ed.), Handbook of child psychology: Vol. 3. Social,
Journal of Personality and Social Psychology, 71, 364–374. doi: emotional, and personality development (pp. 719–788). New York: Wiley.
10.1037/0022‐3514.71.2.364 Durlak, J. A., & Lipsey, M. W. (1991). A practitioner’s guide to meta‐

Barchia, K., & Bussey, K. (2011). Individual and collective social analysis. American Journal of Community Psychology, 19, 291–332.
cognitive influences on peer aggression: Exploring the contribution of doi: 10.1007/BF00938026
aggression efficacy, moral disengagement, and collective efficacy. Field, A. P., & Gillett, R. (2010). How to do a meta‐analysis. British
Aggressive Behavior, 37, 107–120. doi: 10.1002/ab.20375 Journal of Mathematical and Statistical Psychology, 63, 665–694. doi:
Barnett, V., & Lewis, T. (1994). Outliers in statistical data (3rd ed.). New 10.1348/000711010X502733

York: Wiley. Fitzpatrick, S., Bussey, K. (2012 , March). The influence of moral

Bauman, S. (2010). Cyberbullying in a rural intermediate school: An disengagement on social bullying in dyadic very best friendships. Paper
exploratory study. Journal of Early Adolescence, 30, 803–833. doi: presented at the SRA Biennial Meeting, Vancouver, Canada.
10.1177/0272431609350927 Gibbs, J., Potter, G., & Goldstein, A. (1995). The EQUIP program:
Becker, B. J. (2000). Multivariate meta‐analysis. In H. E. A. Tinsley, & S. Teaching youth to think and act responsibly through a peer‐helping
Brown (Eds.), Handbook of applied and multivariate statistics and approach. Champaign, IL: Research Press.

mathematical modeling. San Diego: Academic Press. Gini, G. (2006). Social cognition and moral cognition in bullying: What’s
Begg, C. B., & Mazumdar, M. (1994). Operating characteristics of a rank wrong? Aggressive Behavior, 32, 528–539. doi: 10.1002/ab.20153
correlation test for publication bias. Biometrics, 50, 1088–1101. doi: Gini, G., Albiero, P., Benelli, B., & Altoè, G. (2007a). Does empathy
10.2307/2533446 predict adolescents’ bullying and defending behavior? Aggressive
Bonett, D. G. (2008). Meta‐analytic interval estimation for Pearson Behavior, 33, 467–476. doi: 10.1002/ab.20204

correlations. Psychological Methods, 13, 173–189. Gini, G., Albiero, P., Benelli, B., Matricardi, G., Pozzoli, T. (2007b ,
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). August). Association between empathy, social intelligence, moral
Introduction to meta‐analysis. Chichester, UK: Wiley. disengagement and different forms of bullying behaviour in Italian

Bussey, K., & Quinn, C. (2012 , March). Moral disengagement as the link adolescents. Paper presented at the 13th European Conference on
between the influences of empathy and perspective taking on peer Developmental Psychology, Jena, Germany.
conflict. Paper presented at the 2012 SRA Biennial Meeting, Gini, G., & Pozzoli, T. (2006). The role of masculinity in children’s
Vancouver, Canada. bullying. Sex Roles, 7/8, 585–588. doi: 10.1007/s11199‐006‐9015‐1

Bussman, J. R. (2007). Moral disengagement in children’s overt and Gini, G., & Pozzoli, T. (2009). Association between bullying and
relational aggression. Dissertation Abstracts International: Section B: psychosomatic problems: A meta‐analysis. Pediatrics, 123, 1059–
The Sciences and Engineering, 68(7‐B), 4813. 1065. doi: 10.1542/peds.2008‐1215

Aggr. Behav.
Moral Disengagement and Aggressive Behavior 67

Gini, G., Pozzoli, T. (2013). Being bullied and psychosomatic problems: A of Developmental Psychology, 31, 1–14. doi: 10.1111/j.2044‐835X.
meta‐analysis. Pediatrics. doi: 10.1542/peds.2013-0614 2011.02066.x

Gini, G., Pozzoli, T., & Hauser, M. (2011). Bullies have enhanced moral Menesini, E., Sanchez, V., Fonzi, A., Ortega, R., Costabile, A., & Lo
competence to judge relative to victims, but lack moral compassion. Feudo, G. (2003). Moral emotions and bullying: A cross‐national
Personality and Individual Differences, 50, 603–608. doi: 10.1016/j. comparison of differences between bullies, victims and outsiders.
paid.2010.12.002 Aggressive Behavior, 29, 515–530. doi: 10.1111/1467‐9507.00121

Hawker, D. S. J., & Boulton, M. J. (2000). Twenty years’ research on Obermann, M. L. (2011). Moral disengagement in self‐reported and peer‐
peer victimization and psychosocial maladjustment: A meta‐analytic nominated school bullying. Aggressive Behavior, 37, 133–144. doi:
review of cross‐sectional studies. Journal of Child Psychology and 10.1002/ab.20378
Psychiatry and Allied Disciplines, 41(4), 441–455. doi: 10.1111/1469‐ Orobio de Castro, B., Veerman, J. W., Koops, W., Bosch, J. D., &
7610.00629 Monshouwer, H. J. (2002). Hostile attribution of intent and aggressive
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta‐analysis. behavior: A meta‐analysis. Child Development, 73, 916–934. doi:
Orlando, FL: Academic Press. 10.1111/1467‐8624.00447
Hedges, L. V., & Vevea, J. L. (1998). Fixed and random‐effects models in Orwin, R. G. (1983). A fail‐safe N for effect size in meta‐analysis. Journal
meta‐analysis. Psychological Methods, 3, 486–504. doi: 10.1037/ of Educational and Behavioral Statistics, 8, 157–159. doi: 10.3102/
1082‐989X.3.4.486 10769986008002157

Hemphill, J. F. (2003). Interpreting the magnitudes of correlation Paciello, M., Fida, R., Tramontano, C., Lupinetti, C., & Caprara, G. V.
coefficients. American Psychologist, 58, 78–79. doi: 10.1037/0003‐ (2008). Stability and change of moral disengagement and its impact on
066X.58.1.78 aggression and violence in late adolescence. Child Development, 79,
Higgins, J., Thompson, S. G., Deeks, J. J., & Altman, D. G. (2003). 1288–1309. doi: 10.1111/j.1467‐8624.2008.01189.x

Measuring inconsistency in meta‐analyses. British Medical Journal, Pelton, J., Gound, M., Forehand, R., & Brody, G. (2004). The moral
327, 557–560. doi: 10.1136/bmj.327.7414.557 disengagement scale: Extension with an American minority sample.
Hodgdon, H. B. (2010). Child maltreatment and aggression: The mediating Journal of Psychopathology and Behavioral Assessment, 26, 31–39.
role of moral disengagement, emotion regulation, and emotional doi: 10.1023/B:JOBA.0000007454.34707.a5
callousness among juvenile offenders. Dissertation Abstracts Interna- Pepler, D., Jiang, D., Craig, W., & Connolly, J. (2008). Developmental
tional: Section B: The Sciences and Engineering, 70(9‐B), 5822. trajectories of bullying and associated factors. Child Development, 79,
Retrieved from http://ovidsp.ovid.com/ovidweb.cgi?T¼JS&PAGE¼ 325–338. doi: 10.1111/j.1467‐8624.2007.01128.x

reference&D¼psyc6&NEWS¼N&AN¼2010‐99061‐047. Perren, S., & Sticca, F. (2011 , March). Bullying and morality: Are there
Huang, C. (2011). Self‐concept and academic achievement: A meta‐ differences between traditional bullies and cyberbullies? Poster
analysis of longitudinal relations. Journal of School Psychology, 49, presented at the SRCD Biennial Meeting, Montreal, Quebec, Canada.

505–528. doi: 10.1016/j.jsp.2011.07.001 Pornari, C. D., & Wood, J. (2010). Peer and cyber aggression in secondary

Hyde, L., Shaw, D., & Moilanen, K. (2010). Developmental precursors of school students: The role of moral disengagement, hostile attribution
moral disengage‐ment and the role of moral disengagement in the bias, and outcome expectancies. Aggressive Behavior, 36, 81–94. doi:
development of antisocial behavior. Journal of Abnormal Child 10.1002/ab.20336
Psychology, 38, 197–209. doi: 10.1007/s10802‐009‐9358‐5 Pozzoli, T., Gini, G., & Vieno, A. (2012). Individual and class moral

Hymel, S., Rocke Henderson, N., & Bonanno, R. A. (2005). Moral disengagement in bullying among elementary school children.
disengagement: A framework for understanding bullying among Aggressive Behavior, 38, 378–388.

adolescents [Special issue]. Journal of Social Sciences, 8, 1–11. Qingquan, P., Zongkui, Z., Fan, P., & Lei, H. (2009). Moral
Hymel, S., Schonert‐Reichl, K. A., Bonanno, R. A., Vaillancourt, T., & disengagement in middle childhood: Influences on prosocial and
Rocke Henderson, N. (2010). Bullying and morality. Understanding aggressive behaviors. Proceedings of the First International Confer-
how good kids can behave badly. In S. R. Jimerson, S. M. Swearer, & ence on Information Science and Engineering (pp. 3361–3364). doi:
D. L. Espelage (Eds.), Handbook of bullying in schools. An 10.1109/ICISE.2009.760
international perspective (pp. 101–118). New York: Routledge. Ribeaud, D., & Eisner, M. (2010). Are moral disengagement, neutralization
Jolliffe, D., & Farrington, D. P. (2011). Is low empathy related to techniques and self‐serving cognitive distortions the same? Developing
bullying after controlling for individual and social background a unified scale of moral neutralization of aggression. International
variables? Journal of Adolescence, 34, 59–71. doi: 0.1016/j.adolescence. Journal of Conflict & Violence, 4, 298–315.
2010.02.001 Rigby, K., & Slee, P. T. (1993). Dimensions of interpersonal relating among
Kiriakidis, S. P. (2008). Moral disengagement: Relation to delinquency and Australian school children and their implications for psychological
independence from indices of social dysfunction. International Journal well‐being. Journal of Social Psychology, 133, 33–42. doi: 10.1080/
of Offender Therapy and Comparative Criminology, 52, 571–583. doi: 00224545.1993.9712116
10.1177/0306624X07309063 Rosenthal, R. (1978). Combining results of independent studies.
Kline, R. B. (2004). Beyond significance testing. Washington, DC: APA. Psychological Bulletin, 85, 185–193. doi: 10.1111/j.1420‐9101.2005.
Lipsey, M. W., & Wilson, D. B. (2000). Practical meta‐analysis (Vol. 49). 00917.x
Thousand Oaks, CA: SAGE Publications. Rosenthal, R. (1979). The file drawer problem and tolerance for null
Malti, T., Gasser, L., & Gutzwiller‐Helfenfinger, E. (2010). Children’s results. Psychological Bulletin, 86, 638–641. doi: 10.1037/0033‐
interpretive understanding, moral judgment, and emotion attributions: 2909.86.3.638
Relations to social behaviour. British Journal of Developmental Rosenthal, R. (1991). Meta‐analytic procedures for social research.
Psychology, 28(2), 275–292. doi: 10.1348/026151009X403838 Beverly Hills, CA: Sage.

Menesini, E., Fonzi, A., & Vannucci, M. (1999). Il disimpegno morale: la Shulman, E. P., Cauffmann, E., Piquero, A. R., & Fagan, J. (2011). Moral
legittimazione del comportamento prepotente. In Fonzi, A, (Ed.), Il disengagement among serious juvenile offenders: A longitudinal study
gioco crudele. Studi e ricerche sui correlati psicologici del bullismo of the relations between morally disengaged attitudes and offending.
(pp. 39–53). Firenze: Giunti. Developmental Psychology, 47, 1619–1632. doi: 10.1037/a0025404
Menesini, E., Nocentini, A., & Camodeca, M. (2013). Morality, values, Smith, P. K., Mahdavi, J., Carvalho, M., Fisher, S., Russell, S., & Tippett,
traditional bullying, and cyberbullying in adolescence. British Journal N. (2008). Cyberbullying: Its nature and impact in secondary school

Aggr. Behav.
68 Gini et al.

pupils. Journal of Child Psychology and Psychiatry, 49, 376–385. doi: Trentacosta, C. J., & Fine, S. E. (2010). Emotion knowledge, social
10.1111/j.1469‐7610.2007.01846.x competence, and behavior problems in childhood and adolescence: A

Stevens, D. L., & Hardy, S. A. (2013). Individual, family, and peer meta‐analytic review. Social Development, 19, 1–29. doi: 10.1111/
predictors of violence among Samoan adolescents. Youth & Society, 45, j.1467‐9507.2009.00543.x
428–449. doi: 10.1177/0044118X11424756 Ttofi, M. M., Farrington, D. P., Losel, F., & Loeber, R. (2011). The
Suler, J. (2004). The online disinhibition effect. CyberPsychology & predictive efficiency of school bullying versus later offending:
Behavior, 7, 321–326. doi: 10.1089/1094931041291295 A systematic/meta‐analytic review of longitudinal studies. Criminal
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th Behavior and Mental Health, 21, 80–89. doi: 10.1002/cbm.808
ed.) Needham Heights, MA: Allyn & Bacon. Vaillancourt, T., Hymel, S., Duku, E., Krygsman, A., Cunningham, L.,
Thompson, B. (2002). ‘Statistical’, ‘practical’, and ‘clinical’: How many Davis, C., Short, K., Cunningham, C. (2006 , July). Beyond the
kinds of significance do counselors need to consider. Journal of Dyad: An analysis of the impact of group attitudes and behavior on
Counseling & Development, 80, 64–71. doi: 10.1002/j.1556‐6678. bullying. Paper presented at the biennial meeting of the International
2002.tb00167.x Society for the Study of Behavioral Development, Melbourne
Tisak, M. S., Tisak, J., & Goldstein, S. E. (2006). Aggression, delinquency, Australia.

and morality: A social‐cognitive perspective. In M. Killen, & J. Yadava, A., Sharma, N. R., & Gandhi, A. (2001). Aggression and moral
Smetana (Eds.), Handbook of Moral Development (pp. 611–632). disengagement. Journal of Personality and Clinical Studies, 17(2),
Mahwah, NJ: Erlbaum. 95–99.

Aggr. Behav.

Vous aimerez peut-être aussi