You are on page 1of 10

AGGRESSIVE BEHAVIOR

Volume 33, pages 467476 (2007)

Does Empathy Predict Adolescents Bullying


and Defending Behavior?
Gianluca Gini, Paolo Albiero, Beatrice Benelli, and Gianmarco Altoe

Department of Developmental and Social Psychology, University of Padua, Padova, Italy

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Through structural equation modeling, this study tested a path of relations in which different levels of empathic responsiveness were
posited to be differently associated to bullying and defending behavior. Three hundred and eighteen Italian adolescents (142 girls
and 176 boys; mean age 5 13.2 years) completed the Daviss Interpersonal Reactivity Index [Davis, 1983] for empathy and the
Participant role scales [Salmivalli et al., 1996] for bullying and defending behavior. The results revealed that the model tted the
data adequately, but only in the case of boys. As hypothesized, low levels of empathic responsiveness were associated to students
involvement in bullying others. In contrast, empathy was positively associated with actively helping victimized schoolmates.
However, the estimates algorithm did not reach convergence with girls data. The current ndings conrm and extend the literature
on the relation between empathy, prosociality and aggressive behavior. Educational implications are also discussed. Aggr. Behav.
33:467476, 2007. r 2007 Wiley-Liss, Inc.

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :

Keywords: bullying; prosocial behavior; empathy; participant role approach; structural equation modeling

Violence between peers in schools is a widespread students in the peer group, that is to say those who
phenomenon that worries psychologists, teachers often behave as defenders of their victimized school-
and families in many countries around the world mates [e.g., Craig et al., 2000; Pepler and Craig,
[e.g., Baldry, 2004; Gini, 2004, 2006a; Rigby and 1995], and tested the hypothesis of a positive
Slee, 1993; Scheithauer et al., 2006; Smith et al., relation between their empathic responsiveness and
1999 for a review]. One of the most pervasive forms their helping behavior.
of school violence is bullying, which has been
dened as a repeated aggressive behavior perpe-
trated by a bully, or a group of bullies, who EMPATHY, SOCIAL AND AGGRESSIVE
systematically victimizes a weaker peer [Olweus, BEHAVIOR
1993; Perry et al., 1990]. Moreover, school bullying Empathy is generally dened as sharing another
is characterized by intentionality and can be either persons emotional state [Eisenberg and Strayer,
direct (physical bullying or verbal harassment) or 1987]. Current approaches describe dispositional
indirect (social exclusion, malicious rumor spread- empathy as a multidimensional construct that has
ing, withdrawal of friendship, etc.) [Bjorkqvist, both cognitive and affective/emotional components
1994; Crick and Grotpeter, 1995; Olweus, 1993]. [Davis, 1994]. The cognitive component [in parti-
This study focused on a specic characteristic that cular, perspective taking (PT)] reects the ability
has been proposed as a possible moderator of to identify with and understand other peoples
aggressive and social behavior: empathic respon- perspective, whereas the emotional component [in
siveness. Drawing on the literature on this particular
topic, which will be discussed below, we tested the
hypothesis of a negative relation between empathy Correspondence to: Gianluca Gini, Department of Developmental
and bullying behavior in Italian adolescents. More- and Social Psychology, University of Padua, via Venezia, 8, 35131,
over, since the most recent literature on school Padova, Italy. E-mail: gianluca.gini@unipd.it
bullying has stressed the social nature of bullying, Received 20 December 2005; Accepted 24 July 2006
following the participant role approach [Salmivalli Published online 5 April 2007 in Wiley InterScience (www.
et al., 1996], we considered also the most prosocial interscience.wiley.com). DOI: 10.1002/ab.20204.

r 2007 Wiley-Liss, Inc.


468 Gini et al.

particular, empathic concern (EC)] is characterized Specically, empathy can inhibit or reduce
by the tendency to experience feelings of concern or aggressive behavior through two different moderat-
sympathy toward others [Davis, 1994]. ing mechanisms. The rst is related to the cogni-
At rst, Feshbach [1978, 1987] stressed the multi- tive component of empathy and acts through the
dimensional nature of empathy. In particular, both individuals role taking ability [Davis, 1994]:
the cognitive and the emotional components of the more a person is able to appreciate other
empathy coexist in her model, but the cognitive peoples perspective, the more he or she can under-
abilities (i.e., the ability to recognize others emo- stand and tolerate the position of others, thus
tions and the role taking ability) are considered as making the adoption of aggressive behavior less
prerequisites of empathy. In other words, according likely [Feshbach, 1978]. More precisely, role-taking
to Feshbach, being able to recognize the emotions of ability should allow a detached analysis of other
another individual and to take his/her own point of peoples reasons and motivations, so that their
view is necessary, but not sufcient, to empathize actions can be better understood and accepted.
with others feelings. For example, we could make The second mechanism, instead, deals with the
the case of an individual who has good social- affective components of empathy, through which
cognitive abilities in terms of, but not limited to, PT aggressors can experience the victims pain and
or theory of mind [e.g., Gini, 2006b; Sutton and inhibit their own aggressive behavior to avoid the
Keogh, 2000], but lacks the emotional ability to emotional stress caused by the situation or reduce
participate in others emotion. the victims suffering [Batson et al., 1989; Eisenberg
Following the important contribution given and Fabes, 1998]. Both cognitive [Kaukiainen et al.,
by Feshbach to the literature on empathy, other 1999] and emotional [Albiero and Lo Coco, 2001;
authors have proposed a multidimensional ap- Mehrabian, 1997] components of empathy mitigate
proach to the study of empathy [e.g., Davis, 1994; aggressive behavior and violence.
Hoffman, 2001; Strayer, 1993; Strayer and Roberts,
1997]. Despite some differences between their
BULLYING AND EMPATHY
models, all those authors agree that the different
components of empathy can be considered and The traditional stereotype of bullies describes
measured separately, but they have to be put these children as not academically bright, anxious,
together and considered jointly to understand a insecure, and prone to resort to violence to solve
complex and articulated construct such as empathy conicts, this being the only response mechanism
is [Hoffman, 2001]. available to them [Olweus, 1993]. Moreover,
The relation between empathy and aggressive other potential deciencies have been identied
behavior has been extensively studied in childhood in their social information-processing [e.g., Crick
and adolescence [Miller and Eisenberg, 1988]. Em- and Dodge, 1999] and social problem-solving [e.g.,
pathic responsiveness, in fact, usually induces indivi- Crick and Dodge, 1994]. However, Sutton et al.
duals to moderate their aggressive behavior, in that [1999] argued that at least some bullies are socially
highly empathic individuals are able to emotionally competent and have superior theory of mind
anticipate the negative outcomes produced by their skills. These children have good levels of social
own conduct toward another person [Hoffman, 2000]. intelligence and are well able to understand
Recently, a meta-analysis by Jolliffe and Farrington others mental states, even though their theory
[2004] has summarized these results, conrming the of mind seems to be purely instrumental and
positive relation between antisocial behaviors and used in a Machiavellian way for personal advan-
low levels of empathy. This association is stronger in tages [Andreou, 2004, 2006; Sutton and Keogh,
adolescents and young adults. Furthermore, high 2000].
levels of empathic responsiveness enhance prosocial What bullies may lack, therefore, are empathic
behavior [e.g., Davis, 1994; Eisenberg and skills or, in other words, the ability to appreciate the
Fabes, 1998; Hoffman, 2001] and are positively emotional consequences of their behaviors on other
related to a decrease in aggressive or other peoples feelings and share and empathize with the
externalizing behaviors [e.g., Bjorkqvist et al., feelings of others [Arsenio and Lemerise, 2001;
2000; Kaukiainen et al., 1999; Mehrabian, 1997; Eisenberg and Fabes, 1998]. Other researchers also
Miller and Eisenberg, 1988]. Kaukiainen et al. agree with the idea of bullies as individuals
[1999] found that higher levels of PT were associated characterized by a kind of cold cognition who
with low physical, verbal and indirect aggression in fail to understand others feelings [e.g., Bjorkqvist
preadolescents. et al., 2000; Randall, 1997] and have suggested that

Aggr. Behav. DOI 10.1002/ab


Bullying and Empathy 469

if a victim displays distress, this only serves to levels of students helping behavior toward victi-
reinforce the bullies behavior [Davis, 1994]. In other mized schoolmates.
words, according to Sutton et al. [1999] view of the In their study on empathy and bullying, Endresen
skilled manipulator, bullies understand others and Olweus [2001] used self-report measures to
emotions, but they do not share them, as they are, assess bullying behavior. However, owing to the
being characterized by a sort of theory of nasty nature of the topic under study, this may represent a
minds. problem, because participants self-evaluation of
Notwithstanding this recent emphasis on bullies bullying behavior may be biased by their level of
lack of empathy, there has been very little examina- empathy: lower levels of empathic responsiveness
tion of these childrens empathic responsiveness. might give rise to lower abilities to recognize ones
Endresen and Olweus [2001], for example, have own aggressive acts towards others. In our study
studied the relation between empathy, measured students pro-bullying and helping behaviors have
through an ad hoc developed short scale, and been assessed through peer nomination, which
bullying behavior, self-reported by Norwegian represents a valid and reliable method to evaluate
preadolescents. The authors found a negative social behavior within the peer-group context,
weak relation between empathy and bullying because it benets from the independent judgments
(Pearson correlation coefcient 5 0.15). A larger of all classmates [e.g., Huesmann et al., 1994; Perry
negative correlation was found between empathy et al., 1988].
and positive attitudes to bullying (r 5 0.40). To this respect, it is important to note that,
A possible limitation of this study, however, in the growing literature on participant roles,
was the use of self-report measures for all constructs, two different approaches have been used. A rst
especially for active bullying behavior, which approach considers participant roles as typologies,
might have led to social desirability biases. that is, children are classied into different groups
A different methodology was employed by Warden (i.e., bullies, victims, defenders, etc.) accordingly
and Mackinnon [2003], who compared empathic to their highest participant role score. A second way
responsiveness of peer-nominated bullies, victims to deal with nominations in participant roles is to
and prosocial children and found that prosocial consider them as dimensions, that is, continuous
children were more empathic than bullies. However, variables along which childrens typical behavior is
when gender was covaried out, the difference measured. Researchers have adopted frequently
between bullies and prosocial children disappeared. both approaches within the same empirical
Interestingly, the victim group did not differ work [e.g., Gini, 2006b; Salmivalli et al., 1996] or,
signicantly from other pupils on measures of alternatively, have used participant roles scores as
empathy. independent dimensions only [e.g., Salmivalli
and Voeten, 2004]. Coherently with this second
line of research, in this study, we adopted a
THE PRESENT STUDY
dimensional approach and participant role scores
The aim of this work was to study the relation were used as a continuous measure rather than a
between empathy and peer-evaluated bullying and categorical one.
helping behavior, measured through the participant As discussed above, most theoretical models about
role scales [Salmivalli et al., 1996], in a sample of empathy adopted a multidimensional approach to
Italian adolescents. The roles we were interested in the study of this construct [Davis, 1994; Hoffman,
were bullies and their followers on the one side, and 2001; Strayer, 1993], whereas, as far as the assess-
defenders on the other. The rst two roles were ment of empathy through the interpersonal reactiv-
assumed to share high levels of aggression and low ity index (IRI) is concerned, two different positions
levels of empathy. The last role was assumed to be emerged in the literature. Some authors have
little aggressive and highly empathic. Drawing on assessed the different dimensions of empathy sepa-
the previous literature on the inuences of different rately (i.e., a separate score for each subscales)
levels of empathic responsiveness on aggressive and [e.g., Davis, 1983, 1994], whereas in other recent
prosocial behavior [e.g., Eisenberg and Fabes, 1998; studies [e.g., Burkard and Knox, 2004; Burke, 2001;
Hoffman, 2001; Jolliffe and Farrington, 2004], we Moriarty et al., 2001; Webster, 2002] a global
aimed to test a model that posits a negative path empathy index has been calculated (i.e., the summa-
from empathy to bullying behavior and predicts that tion of all subscale scores). The latter choice is
higher levels of empathy are related to lower consistent with the second-order hierarchical model
engagement in bullying behavior and to higher validated by Cliffordson [2001, 2002], who argued

Aggr. Behav. DOI 10.1002/ab


470 Gini et al.

that empathy consists of a single global dimension, (0.4%) were excluded from the analyses because of
in which EC plays the central role. missing data in their questionnaires.
Consistent with this second line of research, in
this study we used a single index calculated as the
Measures
sum of the scores measuring the two dimensions
traditionally considered most relevant for empathy Participant role. The participant role scales
(i.e., PT and EC). Besides, the model used in our [Salmivalli et al., 1996] were used to collect peer
study for data analysis allowed us to verify whether nominations. Our participants were asked to nomi-
empathy, measured through a global index, is nate up to ve classmates who tted each of the
related to bullying behavior and, at the same time, behavioral descriptions of bullying situations that
to establish (through the standardized coefcients) compose the questionnaire. For each nominated
the role each component (the cognitive and the classmate, participants were also asked to indicate
affective) has in the denition of empathy in whether he or she sometimes (scored as 1) or
the model. often (scored as 2) shows that behavior. The
Finally, we wanted to test for possible gender nominations each person received per item were
differences in the participant role scales and in summed and divided by the number of nominators,
empathy. Consistent with the literature about yielding a score from (0) not nominated by any of
participant roles in bullying [e.g., Salmivalli et al., the nominators to (2) nominated as often by every
1996; Sutton and Smith, 1999], we expected that nominator. Following this, scores were summed
boys would be nominated in the pro-bullying scales across items to yield an overall bullying/defending
more often than girls. In contrast, girls were score per person.
expected to be nominated as defenders more often In their original work, Salmivalli et al. [1996]
than boys. Gender differences were also expected in identied three different aggressive roles: the bully
the levels of participants empathic responsiveness, (four items, e.g., Starts bullying), the bullys
because the literature on empathy [see Davis, 1994; assistant (two items, e.g., Helps the bully, maybe
Lennon and Eisenberg, 1987] usually describes girls by catching or holding the victim) and the bullys
as more empathic than boys. Given such expected reinforcer (ve items, e.g., Laughs at people getting
gender differences in all the measures used in bullied). However, several studies in different
the study, it was reasonable to hypothesize that also countries have shown high levels of intercorrelation
the relationships between empathy and bullying or among the three aggressive scales and a general
defending tested in our model could be different for pro-bullying factor emerged consistently [e.g., Me-
boys and girls. To test this hypothesis we computed nesini et al., 2003; Salmivalli et al., 1998; Sutton and
the structural equation modeling (SEM) for boys Smith, 1999]. This factor was also found in our
and girls, separately. sample. For this reason, in this study we created a
single pro-bullying scale, merging the items of the
three aggressive scales (intercorrelations among the
METHOD three scales ranged from 0.69 to 0.87). In contrast,
the ve items for defender behaviors (e.g., Stick up
Participants
for the victim) loaded on an independent factor,
A total of 318 Italian adolescents (142 girls and which was kept as a separate scale. Reliability
176 boys) participated in the study. They were coefcients for the two scales obtained in our sample
attending 7th and 8th grades (2nd and 3rd grades in were 0.94 for the pro-bullying scale and 0.76 for the
Italian middle schools) in a middle school situated in defender scale.
a middle-sized town in the North of Italy. A total of Interpersonal reactivity index. IRI [Davis,
14 school classes participated (mean class size 5 23 1980, 1983]. The 28-item questionnaire consists
students, range: 2025). The school serves a low- of four 7-item subscales (Fantasy-empathy, PT,
middle class population. At the time of the study, EC, Personal distress), measuring separate but
the mean age of the sample was 13.2 years intercorrelated components of empathy. As sug-
(SD 5 0.53, range: 1214). In terms of racial/ethnic gested by Burkard and Knox [2004], a review of the
background, the sample was all Caucasian (100%). subscale items indicates that the EC and the PT
All participants received school and parental per- subscales correspond more directly with the con-
mission to participate before the collection of the ceptual denition of empathy described above. In
data. None of the parents denied permission for contrast, items from the Fantasy and Personal
their children to participate. Few participants Distress subscales do not seem to assess concep-

Aggr. Behav. DOI 10.1002/ab


Bullying and Empathy 471

tually shared and recognized aspects of empathy. person being bullied to defend himself or herself. It
For this reason and consistent with other studies is also bullying when a child or young person is
[e.g., Burkard and Knox, 2004], in this study we teased repeatedly in a nasty way. But, it is not
used only two subscales of the IRI: (a) the PT bullying when two children or young people of
subscale was used to assess the cognitive component about the same strength have the odd ght or
of empathy, that is the ability to adopt the viewpoint quarrel.
of others in everyday life (e.g., I sometimes try to
understand my friends better by imagining how
things look from their perspective); (b) the EC RESULTS
subscale was used to measure the emotional aspect Descriptive Results
of empathy, as it assesses feelings of sympathy and
concern for the less fortunate (e.g., I often have First, we analyzed gender differences in the two
tender, concerned feelings for people less fortunate participant role scales and the two IRI subscales.
than myself). Participants were asked to indicate Boys (M 5 0.36, SD 5 .35) scored higher than girls
how well each item described them on a 5-point (M 5 0.10, SD 5 .13) in the pro-bullying scale,
scale, from 1 (not at all) to 5 (extremely). t(316) 5 8.10, Po.001. In contrast, girls scores
The IRI was validated on the Italian population (M 5 0.91, SD 5 .57) in the defender scale were
by Albiero et al. [2006], who conrmed the factorial higher than boys scores (M 5 0.35, SD 5 .35),
model hypothesized by Davis and found sufcient t(316) 5 10.80, Po.001. Gender differences were
reliability coefcients for all the scales (range: also found for the IRI subscales. Girls (M 5 3.30,
0.630.75). In this study similar coefcients were SD 5 .55) reported higher PT scores than
found (0.69 and 0.73 for the PT and the EC, boys (M 5 2.94, SD 5 .51), t(316) 5 5.98, Po.001.
respectively). The same result was found for the EC sub-
scale, where girls (M 5 3.66, SD 5 .53) scored higher
than boys (M 5 3.32, SD 5 .52), t(316) 5 5.92,
Procedure
Po.001.
The measures were administered in a counter- In Table I, the means and standard deviations
balanced order at class-level by two researchers who for all variables and the intercorrelation coefcients
were not acquainted with the students. Before are reported for boys and girls separately. Girls
administration of the Participant role scales, parti- correlations are in parenthesis. For both groups of
cipants were provided with the denition of bullying participants, the two peer-evaluated behavioral
behavior used in previous Italian studies on school measures (pro-bullying and defending behavior
bullying [Gini, 2006b]: We say a child or young scales) were not correlated signicantly with each
person is being bullied, or picked on when another other. This conrmed the independency of the two
child or young person, or a group of children or constructs.
young people, say nasty and unpleasant things to Within the male group, a signicant negative
him or her. It is also bullying when a child or young association between pro-bullying scores and empa-
person is hit, kicked, threatened, locked inside a thy emerged for the perspective-taking subscale
room, sent nasty notes, when no one ever talks to (Pearson r 5 0.19, N 5 176, Po.01) and the EC
them and things like that. These things can happen subscale (r 5 0.28, N 5 176, Po.001). In contrast,
frequently and it is difcult for the child or young a positive association emerged between the latter

TABLE I. Descriptive Statistics and Intercorrelations for all Measures

1 2 3 4

1. Pro-bullying scale
2. Defender scale .146 (.020)
3. Perspective taking .191 (.178) .102 (.162)
4. Empathic concern .276 ( .063) .220 (.168) .428 (.403)
Boys: M (SD) 0.36 (.35) 0.35 (.35) 2.94 (.51) 3.32 (.52)
Girls: M (SD) 0.10 (.13) 0.91 (.57) 3.30 (.55) 3.66 (.53)

Notes: In each cell, the rst correlation coefcient refers to boys and the correlation within parenthesis refers to girls.
Po.05, **Po.01, ***Po.001.
In each cell, correlations in italics signicantly differ from each other at .01 level (Bonferroni correction).

Aggr. Behav. DOI 10.1002/ab


472 Gini et al.

and the defender scale (r 5 0.22, Po.01). In the case a value that implies non-signicant departure from
of girls, instead, lower coefcients of correlations normality ( 1.96ozo1.96) [Mardia, 1970].
were found and only two correlation indexes were
signicant at the 5% level. For both groups, PT and
EC were moderately intercorrelated with each other Testing the model
(r 5 0.43 and 0.40, for boys and girls, respectively). The scores for the two Participant role behaviors
In each cell, boys and girls correlation coefcients were included in the SEM as observed dependents.
were compared. This comparison yielded two The two IRI subscales were introduced as
signicant results: (a) the correlation between pro- observed independents. Empathy was the latent
bullying scores and PT scores, and (b) the correla- variable.
tion between pro-bullying scores and EC scores were Two SEM analyses were performed to test the
signicantly different (Po.01, Bonferroni corrected) hypothesized model in the male and the female
between boys and girls. group. In the case of boys, the model provided
an excellent t to the data: a non-signicant w2
The SEM analysis (2, n 5 176) 5 1.17, P 5 .56, w2/dfr2; GFI 5 1.00,
AGFI 5 .98, CFI 5 1.00, RMSEAo.001 (90%
The posited structural model was tested on the condence interval: 0.000.13) were obtained. In
covariance matrix for boys and girls separately using this model, empathic responsiveness was posit-
the LISREL 8.54 Program [Joreskog and Sorbom, ively associated to defending behavior (standardized
1993]. The Maximum Likelihood method of estima- coefcient 5 0.29) and negatively associated to pro-
tion was used. Various t indices were used to assess bullying behavior ( 0.37). Defending and pro-
the t of the model: the w2, an index dependent on bullying behaviors were not signicantly intercorre-
sample size that measures the extent to which the lated (Fig. 1). The whole model accounted for 14%
overall model predicts the observed covariance; of the variance for pro-bullying behavior and 9% of
the goodness-of-t index (GFI) and the adjusted the variance for defending behavior.
GFI (AGFI), which are thought of as proportions After tting the theoretical model on boys, we
comparing the value of the tting criterion for the tested the same model on girls. However, in this case
model with the value of the tting criterion when no the estimates algorithm did not reach convergence.
model ts the data; the comparative t index (CFI), This result suggests that the model ts well with
which measures the adequacy of the model if boys, whereas it cannot be validated with girls
compared with the null model. This index is probably because of the very low intercorrelations
independent of sample size and performs well both among the variables involved.
on large and small samples. Finally, the Root Mean Because pro-bullying and defending scores were
Square Error of Approximation (RMSEA) was also non-normally distributed, in both boys and girls, we
used. This is an absolute t index measuring also estimated the model using Generally Weighted
approximation of parameter estimates to true Least-Squares, an alternative method of estimation
parameters in the population. This index takes suggested by Joreskog and Sorbom [1993]. Findings
explicitly the parsimony of the model into account presented in the text were conrmed. The estimates
(i.e., the number of parameters xed versus the algorithm reached convergence only in boys. In this
number of parameters free to be estimated). group, both parameters and t indexes were almost
To evaluate the t of the model, the following equal to those obtained with the Maximum Like-
criteria are considered commonly. If the model is lihood method.
correct, the w2 test statistic should be non-signi-
cant. A GFI, an AGFI and a CFI above .90 indicate
a close t of the model to the data [Hu and Bentler, Perspective
1995]. A RMSEA of .05 or less also indicates a close Taking
0.41
PRO-BULLYING
BEHAVIOR
0.86

t [Browne and Cudeck, 1993]. Moreover, since the -0.37

validity of the likelihood ratio w2 test is based in part EMPATHY

on the assumption of multivariate normality of the 0.29


0.27 DEFENDING
observed data, the issue of normality was consid- BEHAVIOR
0.91

ered. For this purpose, Mardias measure of relative Empathic Concern

multivariate kurtosis was obtained via the PRELIS Fig. 1. Parameter estimates of the veried model. Note: The numbers
program [Joreskog and Sorbom, 1993]. For the refer to standardized structural coefcients. All coefcients are
present sample the multivariate kurtosis was 1.272, signicantly above the Po.05 level.

Aggr. Behav. DOI 10.1002/ab


Bullying and Empathy 473

DISCUSSION hypothesis, in that among adolescent boys the level


of empathic responsiveness emerged to be positively
The aim of our study was to test a model that associated to the helping behavior of defenders. This
described the different relations between empathic does not mean that being highly empathic is a
responsiveness and participation in bullying and sufcient requisite for prosocial behavior; it simply
defending behavior in a sample of Italian adoles- means that male prosocial individuals are character-
cents. Our ndings substantially corroborate the ized by the same high levels of empathy as prosocial
assumptions on which we based the study [Arsenio girls usually are.
and Lemerise, 2001; Eisenberg and Fabes, 1998] and More generally, we are aware that empathy is not
lend support to the validity of the conceptual model the only dispositional trait that could predict the
tested, at least in the male sample, in that boys development of bullying and prosocial behavior.
bullying behavior was signicantly associated with Several personal, educational and social aspects
low levels of emphatic responsiveness. In this must also be considered to understand the pathways
respect, it is worth noting that the empathyaggres- that bring children to assume different positive and
sion negative relation that emerged in our study was negative roles with peers. Nevertheless, the results
slightly higher than has been found in the literature of this study indicate that empathy may play an
[e.g., Endresen and Olweus, 2001; Miller and important role in this issue.
Eisenberg, 1988]. A possible reason for this stronger The application of the SEM, which represents a
relation may be because of the fact that this nding more powerful alternative to multiple regression, to
refers to a group of adolescent boys only. Another the study of this phenomenon also proved effective.
possible explanation may be ascribed to the use of The SEM approach, in fact, allows testing models
a peer-nomination questionnaire to assess bullying overall rather than estimating coefcients individu-
behavior, instead of the more frequently used self- ally, and testing models with multiple dependents.
report measures. Furthermore, avoiding the use of self-report mea-
In contrast, as expected the prosocial behavior of sures in both tasks by using different types of
victim defenders was positively associated to high measures in the model (i.e., a self-report measure
levels of empathy. Those students who stand up for to assess empathy and peer nominations for the
the victim and usually intervene in a bullying participant role behaviors) enhanced the validity of
situation to defend their victimized schoolmates the model and allowed us to be more condent in
not only are competent in social cognition, i.e., interpreting our ndings. The relations found in the
theory of mind and social information processing structural model cannot be because of a general self-
[e.g., Camodeca and Goossens, 2005], or moral serving attribution bias in the case of defenders or
cognition [Gini, 2006b], but also have high levels of a lack of awareness about their behavior in the case
empathic responsiveness. This result is consistent of bullies.
with the literature on the development of prosocial However, some limitations have to be acknowl-
behavior, which has repeatedly found a positive edged. The fact that the model was validated only
relation between empathy and helping behavior with a sample of Italian adolescent boys limits the
from infancy to adolescence [Davis, 1994; Eisenberg generalizability of the present ndings to the male
and Fabes, 1998; Hoffman, 2001]. adolescent population. Our current data do not
The fact that positive relation between empathy allow us to clearly understand why this was the case.
and helping behavior emerged clearly in a sample of Some tentative explanations, however, may provide
boys is also noteworthy, because it contributes to possible directions for future studies.
some extent to the debate on the role of empathic In the broad literature on empathy, several studies
responsiveness in prosocial behavior. Some authors found a stronger negative relation between empathy
[Warden and Mckinnon, 2003], in fact, have noted and aggressive behavior for boys than for girls.
that a common problem in the studies in this eld Miller and Eisenberg [1988], in their meta-analytic
is the confounding role of participants gender, review of the literature on this topic, found
because girls are usually both more empathic and a negative relation between empathy and aggression
more prosocial than boys. Those authors hypothe- in both genders, with the association more consistent
size that empathy may guide prosocial behavior in males. Furthermore, in a study by Kaukiainen
among girls but not among boys and that prosocial et al. [1999], empathy negatively correlated with
boys are no more empathic than boys who are direct but not with indirect forms of aggression.
bullies [Warden and Mckinnon, 2003; p 381]. Finally, Albiero and Lo Coco [2001] have studied the
However, our ndings seem to disconrm this relation between empathy and aggression in Italian

Aggr. Behav. DOI 10.1002/ab


474 Gini et al.

children aged 610 years. Their results showed a students personal characteristics, such as self-
negative association only for boys. esteem and self-efcacy beliefs, that may be pre-
From a methodological point of view, the way in dictive of active defending or passive bystanding
which the participant role scales were devised may behavior.
have contributed partly to our results. Those scales, Despite these limitations, this study may have
similarly to other peer-nomination instruments, some important implications for anti-bullying
measure direct forms of bullying, which are more school projects. In particular, by demonstrating that
typical of boys than of girls, and therefore the empathy is positively associated to defending
distribution of roles scores is often different between behavior and negatively associated to pro-bullying
girls and boys [Bjorkqvist, 1994]. In other words, it behavior, our ndings suggest the relevance of
is often difcult to obtain satisfactory distributions integrating the traditional curricular model, typi-
of scores for pro-bullying behaviors in girls. This is a cally used in anti-bullying intervention programs
recognized limitation of this type of questionnaire [e.g., Gini, 2004; Sharp and Smith, 1994], with
because of different conceptual and methodological specic activities about emotions and socialization
reasons, the analysis of which goes beyond the scope to foster students emotional intelligence. We
of this paper [see Salmivalli et al., 1996]. In our suggest that, through the methodology used by
study, this problem may have had a negative impact teachers in these projects (literary and video stimuli,
on the estimation of the model with girls data. In role-play and group discussions), the themes of
future research, we will attempt to replicate our personal experiences, feelings of bullies and victims,
study with a larger sample of girls and to use moral thoughts of bystanders should be introduced
different instruments to measure bullying behavior, in classroom activities.
which allow to collect information on relational or
social forms of bullying. ACKNOWLEDGMENTS
Furthermore, the use of different scales for the
measurement of empathy will be necessary. The The authors wish to thank the school personnel
validity of the IRI, in fact, has been recently and the participants for their important collabora-
criticized [Jolliffe and Farrington, 2006b]. According tion. Moreover, we are grateful to Erika Cuccarollo
to those authors, the IRI does not really measure and Elena Della Bella for their help in data
cognitive empathy (i.e., the ability to understand collection. Finally, we thank Francesca Cristante
others emotions) but simply PT (i.e., the broad and Lucia Ronconi for their comments on data
ability to take anothers perspective). In our case, analyses.
having a valid measure of cognitive empathy is
essential for understanding the relationship between
empathy and bullying, since both affective and REFERENCES
cognitive empathy would be expected to provide a
Albiero P, Lo Coco A. 2001. Designing a method to assess empathy
unique contribution to this relationship. A valid in Italian children. In: Bohart A, Stipek D, (eds). Constructive
alternative to the IRI is the Basic Empathy Scale and Destructive Behavior. Implications for Family, School &
[Jolliffe and Farrington, 2006b], which already Society, Washington, DC: American Psychological Association,
proved to be effective in studying empathy in bullies pp 205223.
[Jolliffe and Farrington, 2006a]. In future research, Albiero P, Ingoglia S, Lo Coco A. 2006. Contributo alladattamento
italiano dellInterpersonal Reactivity Index. [A Contribution to
we will seek to replicate and extend the current
the Italian validation of the Interpersonal Reactivity Index].
ndings by assessing the relation between bullying TPM Testing Psicometria Metodologia 13:107125.
behavior and empathy, as measured by the Basic Andreou E. 2004. Bully/victim problems and their association with
Empathy Scale, which is now under validation with Machiavellianism and self-efcacy in Greek primary school
the Italian population. children. Br J Educ Psychol 74:297309.
Andreou E. 2006. Social preference, perceived popularity and social
Finally, the question of why defenders do inter-
intelligence. Relations to overt and relational aggression. Sch
vene in a bullying situation cannot be answered by Psychol Int 27:339351.
simply demonstrating that defenders are higher in Arsenio WF, Lemerise EA. 2001. Varieties of childhood bullying:
empathy than their schoolmates. We need to Values, emotion processes and social competence. Soc Dev 10:
improve our knowledge on this group of children 5973.
Baldry AC. 2004. Bullying in School: A Psycho-Social Approach.
and to better understand what differentiates them
Rome: Carlo Amore.
from those schoolmates who do not stick up for Batson CD, Batson JG, Griftt CA, Barrientos S, Brandt JR,
the victim and stand aside, that is the outsiders. Sprengelmeyer P, Bayly MJ. 1989. Negative-state relief and the
In future research, therefore, we aim at analyzing empathy-altruism hypothesis. J Pers Soc Psychol 56:922933.

Aggr. Behav. DOI 10.1002/ab


Bullying and Empathy 475

Bjorkqvist K. 1994. Sex differences in physical, verbal, and indirect Gini G. 2006b. Social cognition and moral cognition in bullying:
aggression. A review of recent research. Sex Roles 30:177188. Whats wrong? Aggr Behav 32:528539.
Bjorkqvist K, Osterman K, Kaukiainen A. 2000. Social intelligence Hoffman ML. 2000. Empathy and Moral Development. Cam-
empathy 5 aggression? Aggr Violent Behav 5:191200. bridge: Cambridge University Press.
Browne MW, Cudeck R. 1993. Alternative ways of assessing model Hoffman ML. 2001. Toward a comprehensive empathy-based theory
t. In: Bollen KA, Long JS, (eds). Testing Structural Equation of prosocial moral development. In: Bohart A, Stipek D, (eds).
Models, Newbury Park: Sage Publications, pp 136162. Constructive & Destructive Behavior: Implications for Family,
Burkard AW, Knox S. 2004. Effect of therapist color-blindness on School, & Society, Washington, DC: American Psychological
empathy and attributions in cross-cultural counseling. J Couns Association, pp 6186.
Psychol 51:387397. Hu L, Bentler PM. 1995. Evaluating model t. In: Hoyle RH, (ed).
Burke DM. 2001. Empathy in sexual offending and nonoffending Structural Equation Modeling: Concepts, Issues and Applica-
adolescent males. J Interpers Violence 16:222233. tions, London: Sage, pp 7699.
Camodeca M, Goossens A. 2005. Aggression, social cognitions, Huesmann LR, Eron LD, Guerra NG, Crawshaw VB. 1994.
anger and sadness in bullies and victims. J Child Psychol Measuring childrens aggression with teachers predictions of
Psychiatry 46:186197. peer nominations. Psychol Assess 6:329336.
Cliffordson C. 2001. Parents judgments and students self-judg- Jolliffe D, Farrington DP. 2004. Empathy and offending:
ments of empathy. The structure of empathy and agreement of A systematic review and meta-analysis. Aggr Violent Behav 9:
judgments based on the Interpersonal Reactivity Index (IRI). Eur 441476.
J Psychol Assess 17:3647. Jolliffe D, Farrington DP. 2006a. Examining the relationship between
Cliffordson C. 2002. The hierarchical structure of empathy: low empathy and bullying. Aggr Behav, accepted, 32:540550.
Dimensional organization and relations to social functioning. Jolliffe D, Farrington DP. 2006b. Development and validation of the
Scand J Psychol 43:4959. Basic Empathy Scale. J Adolesc 29:589611.
Craig WM, Pepler DJ, Atlas R. 2000. Observations of bullying in the Joreskog KA, Sorbom D. 1993. LISREL 8: Structural Equation
playground and in the classroom. Sch Psychol Int 21:2236. Modelling with the SIMPLIS Command Language. Chicago:
Crick NR, Dodge KA. 1994. A review and reformulation of social Scientic Software.
information-processing mechanisms in childrens social adjust- Kaukiainen A, Bjorkqvist K, Lagerspetz K, Osterman K, Salmivalli
ment. Psychol Bull 115:74101. C, Rothberg S, Ahlbom A. 1999. The relationships between social
Crick NR, Dodge KA. 1999. Superiority is in the eye of the intelligence, empathy, and three types of aggression. Aggr Behav
beholder: A comment of Sutton, Smith and Swettenham. Soc Dev 25:8189.
8:128131. Lennon R, Eisenberg N. 1987. Gender and age differences in
Crick NR, Grotpeter JK. 1995. Relational aggression, gender, and empathy and sympathy. In: Eisenberg N, Strayer J, (eds).
social-psychological adjustment. Child Dev 66:710722. Empathy and Its Development, New York: Cambridge
Davis MH. 1980. A multidimensional approach to individual University Press, pp 195217.
differences in empathy. JSAS Catalog Selected Documents Mardia KV. 1970. Measures of multivariate skewness and kurtosis
Psychol 10:85. with applications. Biometrika 57:519530.
Davis MH. 1983. Measuring individual differences in empathy: Mehrabian A. 1997. Relations among personality scales of
Evidence for a multidimensional approach. J Pers Soc Psychol 44: aggression, violence, and empathy: Validational evidence
113126. bearing on the Risk of Eruptive Violence scale. Aggr Behav 23:
Davis MH. 1994. Empathy, A Social Psychological Approach. 433445.
USA: Wm. C. Brown Communications, Inc. Menesini E, Sanchez V, Fonzi A, Ortega R, Costabile A, Lo Feudo
Eisenberg N, Fabes F. 1998. Prosocial development. In: Eisenberg N, G. 2003. Moral emotions and bullying: A cross-national
(ed). Handbook of Child Psychology, Volume 3: Social, comparison of differences between bullies, victims and outsiders.
Emotional and Personality Development, 5th edition, New Aggr Behav 29:515530.
York: Wiley, pp 701778. Miller PA, Eisenberg N. 1988. The relationship of empathy to
Eisenberg N, Strayer J. 1987. Empathy and Its Development. New aggressive and externalizing/antisocial behavior. Psychol Bull
York: Cambridge University Press. 103:324344.
Endresen IM, Olweus D. 2001. Self-reported empathy in Norwegian Moriarty N, Stough C, Tidmarsh P, Eger D, Dennison S. 2001.
adolescents: Sex differences, age trends, and relationship to Decits in emotional intelligence underlying adolescent sex
bullying. In: Bohart A, Stipek D, (eds). Constructive & offending. J Adolesc 24:743751.
Destructive Behavior: Implications for Family, School, Olweus D. 1993. Bullying at School. What We Know and What We
& Society, Washington, DC: American Psychological Associa- can Do. Oxford, UK: Blackwell.
tion, pp 147165. Pepler DJ, Craig WM. 1995. A peek behind the fence: Naturalistic
Feshbach N. 1978. Studies in empathic behavior in children. In: observations of aggressive children with remote audiovisual
Maher B, (ed). Progress in Experimental Personality Research, recording. Dev Psychol 31:548553.
Vol 8. New York: Academic Press, pp 147. Perry DG, Kusel SJ, Perry LC. 1988. Victims of peer aggression. Dev
Feshbach N. 1987. Parental empathy and children adjustment Psychol 24:807814.
/maladjustment. In: Eisenberg N, Strayer J, (eds). Empathy Perry DG, Williard JC, Perry LC. 1990. Peers perceptions of the
and Its Development. New York: Cambridge University Press, consequences that victimized children provide aggressors. Child
pp 271291. Dev 61:13101325.
Gini G. 2004. Bullying in Italian schools: An overview of Randall P. 1997. Adult Bullying: Perpetrators and Victims.
intervention programmes. Sch Psychol Int 25:106116. London: Routledge.
Gini G. 2006a. Bullying as a social process: The role of group Rigby K, Slee PT. 1993. Childrens attitudes towards victims.
membership in students perception of inter-group aggression at In: Tattum DP, (ed). Understanding and Managing Bullying,
school. J Sch Psychol 44:5165. London: Heinemann, pp 119135.

Aggr. Behav. DOI 10.1002/ab


476 Gini et al.

Salmivalli C, Lagerspetz K, Bjorkqvist K, Osterman K, Kaukiainen Strayer J. 1993. Childrens concordant emotions and cognitions in
A. 1996. Bullying as a group process: Participant roles and their response to observed emotions. Child Dev 64:188201.
relations to social status within the group. Aggr Behav 22:115. Strayer J, Roberts W. 1997. Facial and verbal measures of childrens
Salmivalli C, Lappalainen M, Lagerspetz KMJ. 1998. Stability and emotions and empathy. Int J Behav Dev 20:627649.
change of behavior in connection with bullying in schools: A two- Sutton J, Keogh E. 2000. Social competition in school: Relationships
year follow-up. Aggr Behav 24:205218. with bullying, Machiavellianism and personality. Br J Educ
Salmivalli C, Voeten M. 2004. Connections between attitudes, group Psychol 70:443456.
norms, and behaviors associated with bullying in schools. Int Sutton J, Smith PK. 1999. Bullying as a group process: An
J Behav Dev 28:246258. adaptation of the Participant Role approach. Aggr Behav 25:
Scheithauer H, Hayer T, Petermann F, Jugert G. 2006. Physical, 97111.
verbal, and relational forms of bullying among German students: Sutton J, Smith PK, Swettenham J. 1999. Bullying and theory of
Age trends, gender differences, and correlates. Aggr Behav 32: mind: A critique of the social skills decit view of anti-social
261275. behaviour. Soc Dev 8:117134.
Sharp S, Smith PK. 1994. Tackling Bullying in Your School: Warden D, Mackinnon S. 2003. Prosocial children, bullies and
A Practical Handbook for Teachers. London: Routledge. victims: An investigation of their sociometric status, empathy and
Smith PK, Morita Y, Junger-Tas J, Olweus D, Catalano R, Slee P, social competence. Br J Dev Psychol 21:367385.
(eds). 1999. The Nature of School Bullying: A Cross National Webster SD. 2002. Assessing victim empathy in sexual offenders
Perspective. London: Routledge. using the victim letter task. Sex Abuse: J Res Treat 14:281300.

Aggr. Behav. DOI 10.1002/ab

You might also like