-
Victims and Offenders, 3:289312, 2008Copyright Taylor &
Francis Group, LLCISSN: 1556-4886 print/1556-4991 onlineDOI:
10.1080/15564880802143397
289
UVAO1556-48861556-4991Victims and Offenders, Vol. 3, No. 2-3,
April 2008: pp. 141Victims and OffendersBullying: Short-Term and
Long-Term Effects, and the Importance of Defiance Theory in
Explanation and PreventionBullyingM. M. Ttofi and D. P.
Farrington
Maria M. Ttofi and David P. Farrington
Institute of Criminology, University of Cambridge, Cambridge,
UK
Abstract: Bullying is often followed by short-term and long-term
undesirable psycho-social consequences. Both victims and
perpetrators of bullying tend to have high num-bers of physical and
psychological symptoms. In order to prevent bullying and
itsaversive results, it is important to formulate and test theories
of bullying. This articleinvestigates the usefulness of defiance
theory in the explanation of the bullying of sib-lings in families
and peers in schools. Questionnaires were completed by 182
childrenaged 11 to 12 in ten primary schools in Nicosia, Cyprus. We
followed a vignette-basedmethodology to investigate childrens
defiant behavior. Children were given a hypo-thetical scenarioin
which the perpetrator is sanctioned by the parentsand werethen
asked questions that aimed to investigate defiant or compliant
reactions to thesanctions imposed. The type of child in the
vignette was experimentally manipulatedso that children could make
inferences regarding his/her intentionality of wrongdoing.The
results indicate that defiance theory is useful in explaining
bullying behavior. Themain implication from our research is that
defiance theory can assist teachers andpractitioners in
implementing whole-school restorative justice approaches to
reducebullying in schools.
Keywords: bullying, parental bonding, fairness, unacknowledged
shame, defiance,anti-bullying programs
INTRODUCTION
Bullying affects about 1 in 5 school-aged children in many
different countries(Glew, Rivara, & Feudtner, 2000).
Involvement in bullying (as perpetrators orvictims) has negative
effects on the physical and psychological health of
We would like to thank Mr. Lakis Koumi, a dedicated educator,
for his assistance inorganizing visits to schools and data
collection.Address correspondence to David P. Farrington, Institute
of Criminology, SidgwickAvenue, Cambridge, CB3 9DT, UK. E-mail:
[email protected]
-
290 M. M. Ttofi and D. P. Farrington
children and on their future psychosocial adjustment as adults.
This articlehas two aims: to present an evidence-based review of
the effects of bullyingand to describe an empirical test of the
usefulness of defiance theory inexplaining bullying and promoting
theoretically-grounded interventions toreduce bullying. In carrying
out the literature review on the effects of bully-ing, we searched
several databases (e.g. Google Scholar, PsycInfo, Embase)using key
words including bullying combined with consequences,
impact,effects, health, emotional impact, or victims.
The Impact of Being Involved in BullyingThere is a strong link
between involvement in bullying and depressive
symptomatology (Bosworth, Espelage, & Simon, 1999; Van der
Wal, De Wit, &Hirasing, 2003). In their follow-up study,
Kumpulainen, Rasanen, andHenttonen (1999) concluded that children
involved in bullyingas bullies orvictimsshowed significantly more
psychiatric symptoms than other chil-dren, and were psychologically
disturbed according to the Childrens Depres-sion Inventory. Also,
Sourander, Helstela, Helenious, and Piha (2000)reported that the
identification of a child as either a bully or a victim was
sig-nificantly correlated with referral for psychosocial services.
In a study of thelink between bullying and psychosocial adjustment
in primary school, Glew,Fan, Katon, Rivara, and Kernic (2005) found
that bullies-only and victims-only were more likely than bystanders
to feel sad most days. Moreover, theyreported that all three groups
involved in bullying (bullies, victims, and bully-victims) were
significantly more likely than bystanders to feel unsafe
atschool.
In a school-based survey with 16,410 Finnish adolescents aged 14
to 16,Kaltiala-Heino, Rimpela, Marttunen, Rimpela and Rantanen
(1999)showed that adolescents who were being bullied and those who
were bul-lies were at an increased risk of depression and severe
suicidal ideation.Among girls, Kaltiala-Heino, Rimpela, Rantanen,
and Rimpela (2000)found that eating disorders were associated with
involvement in bullying.Also, in their study with 1718 Korean
middle school students, Kim, Koh,and Leventhal (2005) concluded
that, among female students, all three bul-lying groupsvictims,
bullies, and bully-victimshad increased suicidalideation.
Turning to bullies specifically, Salmon (1998), in a sample of
children aged12 to 17, found that those who bullied others tended
to have higher scores ondepression. Similarly, Forero, McLellan,
Rissel and Bauman (1999), in a sur-vey of 3918 Australian school
children, concluded that bullies had high scoreson psychosomatic
symptoms. Forero and colleagues also reported that bullieswere
significantly more likely to dislike school and to think that
school wasnot a nice place to be.
-
Bullying 291
Bullying others has also been identified as a risk factor for
other types ofantisocial behavior (such as excessive drinking and
substance use; Kaltiala-Heino et al., 2000) and later offending
(Farrington, 1993; Sourander et al.,2006). For instance, in a
survey of 4811 Dutch primary school children aged9 to 13, Van der
Wal et al. (2003) found that bulliesirrespective of gendermuch more
often reported delinquent behavior such as shoplifting. In
follow-upstudies of bullies in Norway, Olweus (1997) discovered
that, of those originallyidentified as bullies in the sixth through
the ninth grades, 70% were convictedof at least one crime by age
24.
The Impact of Being BulliedExposure to bullying influences
childrens physical and emotional health
in many ways. For example, children who are bullied tend to show
more anxi-ety than others (Salmon, 1998). Similarly, Bond, Carlin,
Thomas, Rubin, andPatton (2001), in a survey of 2680 secondary
school students, found that vic-timization at age 13 significantly
predicted self- reported symptoms of anxietyor depression at age
14. In a review of existing literature, Salmon, James,Cassidy and
Javaloyes (2000) found that being bullied was frequently a
factorinfluencing the referral of adolescents to psychiatric
services, with depressionbeing diagnosed in over 70% of cases.
Children who are bullied show symptoms of depression and
suicidal ide-ation. In a cross-sectional study of 2766 elementary
school children aged 9 to 12years, Fekkes, Pijpers, and
Verloove-Vanhorick (2004) showed that victims ofbullying, compared
to bullies, bully-victims and children who were notinvolved, had
significantly higher chances of depression and
psychosomaticsymptoms (e.g. headaches, sleeping problems, abdominal
pain, bed-wetting,feeling tired). Kaltiala-Heino et al. (1999) in
Finland discovered that victimsof bullying had an increased risk of
depression and/or suicide and that depres-sion was equally likely
to occur among those who were bullied and those whowere bullies.
For primary school children, Van der Wal et al. (2003)
concludedthat depression and suicidal ideation were common outcomes
of beingbulliedirrespective of genderand that these associations
were stronger forindirect than for direct bullying.
Children who are bullied tend to show many psychosomatic
symptoms. Ina large-scale international comparison (including
123,227 students aged11, 13 and 15 in 28 countries) on bullying and
health among adolescents, Dueet al. (2005) discovered that those
who were bullied showed more evidence of12 different physical and
psychological symptoms (i.e. headache, stomach ache,backache,
dizziness, bad temper, feeling nervous, feeling low, difficulties
ingetting to sleep, morning tiredness, feeling left out,
loneliness, helplessness).Similarly, in a study of a representative
sample of 15,686 students in grades 6through 10 in public and
private American schools (Nansel, Overpeck, Pilla,
-
292 M. M. Ttofi and D. P. Farrington
Ruan, Simons-Morton, & Scheidt, 2001), bullied students had
poorer socialand emotional adjustment, greater difficulty in making
friends, poorer rela-tionships with classmates, and greater
loneliness. Also, in a study of 344 kin-dergarten children aged 5
to 7, Perren and Alsaker (2006) concluded thatvictims were more
submissive, more withdrawn, more isolated, less coopera-tive, less
sociable, had fewer leadership skills, and frequently had no
play-mates.
Bullying has long-term detrimental effects on victims.
Victimization atschool often results in long-term social, emotional
and psychological effects(Duncan, 1999a; Parker & Asher, 1987;
Sharp, 1995). In a study of 276 adultmembers of the British
Stammering Association concerning school experi-ences related to
bullying and its long-term effects, Hugh-Jones and Smith(1999)
found that the majority of respondents had experienced bullying
atschool. Of those who were bullied, the majority reported
immediate negativeeffects and 46% reported undesirable long-term
effects.
Jantzer, Hoover and Narloch (2006) collected data on 170 college
studentsto assess retrospective perceptions of school-age bullying
experiences. Theyconcluded that rates of reported victimization at
school were positively corre-lated with adult levels of shyness.
Also, they found a statistically significantnegative relationship
between retrospectively reported victimization and cur-rent
friendship quality and trust. Similarly, Gilmartin (1987) found
that menwho had been victimized at school often reported
difficulties in trust and inti-macy in opposite-sex relationships
as adults, and Dietz (1994) replicated theseresults for both
sexes.
Olweus (1993) discovered that boys who were victims of bullying
at schoolbetween ages 13 and 16 were, at age 23, more likely to
show depressive ten-dencies and continued to have poor self-esteem.
Matsui, Tsuzuki, Kakuyanaand Onglateo (1996) also reported
long-term effects of school bullying onJapanese students, but only
for those who had low self-esteem and highdepression prior to
victimization; apparently, victimization amplified theseeffects,
which is why these authors characterized the whole process as
avicious cycle.
The Impact on Bully-VictimsNot surprisingly, those who are both
bullies and victims tend to have poor
psychosocial adjustment. For instance, Troy and Sroufe (1987)
showed thatpreschool children who displayed an anxious-avoidant
pattern of attachmenttended to be at greater risk of being
classified as bully-victims. Glew et al.(2005) found that victims
and bully-victims were most likely to report feelingthat they did
not belong at school. Nansel et al. (2001) discovered that
bully-victims demonstrated poorer adjustment across both
social/emotional dimen-sions and problem behaviors; the specific
relationships between bullying/being
-
Bullying 293
bullied and psychosocial adjustment were similar across all age
and sexgroups. Kumpulainen et al. (1999) found that, among 8-year
old children, themost psychologically disturbed group were those
who were both bullies andvictims.
Kaltiala-Heino et al. (1999) also concluded that depression was
most com-mon among those students who were classified as
bully-victims (comparedwith bullies-only and victims only).
Similarly, Kim et al. (2005) showedthat, compared with students who
were not involved with school bullying,bully-victims reported more
suicidal or self-injurious behaviors and suicidalideation. Also,
Kaltiala-Heino et al. (2000) found that anxiety, depression
andpsychosomatic symptoms (i.e. neck and shoulder pain, low back
pain, stomachache, feeling tensed or nervous, irritation or
tantrums, difficulties to getasleep or waking up at night, headache
and fatigue) were most frequentamong bully-victims.
As mentioned above, bully-victims tend to report psychosomatic
symptoms.In a survey of 1639 primary school children, Wolke, Woods,
Bloomfield, andKarstadt (2001) found that direct bully-victims (and
direct victims) were mostlikely to have high psychosomatic health
problems (e.g. poor appetite, worriesabout going to school)
compared to other comparison groups of bullies andvictims.
Kaltiala-Heino et al. (2000) reported that, among boys, eating
disorderswere especially associated with bully-victims. Forero et
al. (1999) found thatbully-victims were significantly more likely
to be boys, experienced frequent andhigh scores on psychosomatic
symptoms, reported being alone, and were currentsmokers. In this
survey, bully-victims had the greatest number of psychologicaland
psychosomatic symptoms compared to bullies-only and
victims-only.
Existing work indicates that bully-victims also tend to have
problembehaviors and attitudes in favor of aggression. For
instance, Andreou (2004),in a study of 186 primary school children
in central Greece, found that bully/victims (compared to bullies
and victims) were worse on Lack of Faith inHuman Nature and overall
Machiavellianism. Bully-victims were similar tovictims on
Self-efficacy for Assertion and similar to bullies on Self-efficacy
forAggression. In a study on the association between bullying and
behavior prob-lems among primary school children, Wolke, Woods,
Bloomfield, and Karstadt(2000) concluded that all children involved
in direct bullying (either as bullies,victims or bully/victims) had
significantly greater total behavior problems,hyperactivity,
conduct problems, peer problem scores, and lower prosocialbehavior
scores compared to those not involved in bullying. In all cases,
fordirect bullying, the mean scores were highest for
bully-victims.
THE PRESENT RESEARCH
Because involvement in bullyingeither as a bully, victim or
bully/victimisoften followed by negative life outcomes (Rigby,
2003), it is important to
-
294 M. M. Ttofi and D. P. Farrington
develop and test theories of bullying, and this is the main aim
of the presentresearch. Specifically, we focus on the extent to
which defiance theory (Sherman,1993) might be useful in explaining
bullying, and especially in explaining thelink between family
factors and bullying. Both sibling bullying and peer bully-ing are
studied; it is known that these two types of bullying are linked
(Duncan,1999b; Wiesner, Capaldi & Patterson, 2003; Wolke &
Samara, 2004). To thebest of our knowledge, this article is the
first attempt to apply defiance theoryto bullying. The theory has
been tested on other topics such as self-reportedoffending
(Freeman, Liosis & David, 2006), marital violence (Sherman,
1995),public responses to police requests (Mastrofski, Snipes &
Supina, 1996), andgang membership (Brownfield, 2006).
Defiance TheoryShermans (1993) theory aims to explain the
conditions under which pun-
ishment increases offending because of defiant reactions to the
sanctionsimposed. The theory explains how the perception of
punishment as fair canlead to compliance and conformity.
Conversely, sanctions that are perceivedas unfair can lead to
defiant reactions that increase future offending. Sher-man (1993,
p. 459) differentiates between two types of defiance. Direct
defi-ance is a crime against a sanctioning agency whereas indirect
defiance isdefined as the displaced just deserts committed against
a target vicariouslyrepresenting the sanctioning agent provoking
the anger. The latter can beseen as a form of retaliation.
The focal point of defiance theory is the impact that sanctions
can have onfuture defiance of the law and future deterrence of
lawbreaking. The theory isbased on the notion that punishment does
not necessarily control offending.Similar criminal sanctions can
have different effects on different offenders(Sherman, 1993, p.
449). Given the widely varying results across a range ofsanction
studies, we should try to understand the conditions under which
eachtype of criminal sanction reduces, increases, or has no effect
on future crimes(Sherman, 1993, p. 445). In doing so, we should not
only pay attention to thesanction per se (e.g. severity, length
etc.), but also to the way in which it isdelivered. The sanctioning
style is linked to the legitimacy of the authority fig-ure and,
sequentially, to possible defiance of the law. Sherman (2000, p.
7)indicates how different types of sanctioning can interact with
different kindsof citizen personalities in ways that predict
different rates of repeat offending.
Based on defiance theory, sanctions can lead to defiant
reactions underfour conditions (Sherman, 1993, p. 460):
1. The offender defines the criminal sanction as unfair,
2. The offender is poorly bonded to or alienated from the
sanctioning agentor the community the agent represents,
-
Bullying 295
3. The offender defines the sanction as stigmatizing and
rejecting a person,not a lawbreaking act, and
4. The offender denies or refuses to acknowledge the shame the
sanction hasactually caused him to suffer.
Applicability to BullyingSherman (1993, p. 466) considered that
defiance theory could be applied
much more widely than to explain the effect of criminal
sanctions on offend-ing. This is reasonable given that the
individual learns what it means to besanctioned for wrongdoing in
the early years of life. Parents and teachersmight be the first
persons to be perceived as sanctioning agents by children. Infact,
it could be argued that the quality of sanctioning they offer is a
crucialelement of each persons socialization process.
The present article aims to assess the usefulness of defiance
theory inexplaining bullying, both of siblings in families and of
peers in schools. Pastresearch has reported a link between bullying
and defiance, especially in dis-cussing Oppositional Defiant
Disorder (ODD). This link is understandable,since ODD is a pattern
of negative (e.g. deliberately doing things that annoyother people,
blaming others for own mistakes), disobedient (defying or
notcomplying with grownups rules or requests) and hostile behaviors
(e.g. losingtemper) (McMahon & Frick, 2005, p. 478). The
Kokkinos and Panayiotou(2004, p. 528) survey supports the link
between ODD and bullying: they foundthat high ODD students tended
to be bullies.
Figure 1 offers a schematic presentation of defiance theory as
applied tobullying behavior. In line with defiance theory, the
following hypotheses aretested:
Figure 1: Applying Defiance Theory to Sibling and Peer Bullying;
Hypothesized Model.
Unacknowledged ShameSibling
Bullying
Mother/Father BondingDefiance to the
sanctionimposed
Perceptions of FairnessPeer
Bullying
Type of child
Note: The dashed lines indicate interrelationships of constructs
that are not directly proposed in defiance theory.
-
296 M. M. Ttofi and D. P. Farrington
1. Unacknowledged shame, parental bonding, perceived fairness
and thetype of the child (good or bad; see later) will be directly
related to defiance.
2. Defiance will be directly related to sibling and peer
bullying.
3. The type of the child will be directly related to perceptions
of fairness.
Building upon previous research, it is hypothesized that
parental bonding will bepositively related to perceptions of
fairness (Kochanska and Murray, 2000) andnegatively related to
unacknowledged shame (Ahmed and Braithwaite, 2004).
METHOD
Sample and ProcedureQuestionnaires were given to 182 male and
female primary school
students aged 11 to 12 attending the sixth grade in Nicosia
primary schools.Ten schools were randomly selected based on the
registers of primary schoolsoffered online by the Ministry of
Education and Culture of the Republic ofCyprus. The survey received
ethical approval from the Ministry. The studentscompleted the
questionnaire after the researchers obtained consent from boththe
children and their parents. No parents denied permission for their
chil-dren to participate in the survey and no child refused to
participate. The ques-tionnaire was self-completed during one
teaching period. The researcherworked beforehand with the children
in another teaching period and helpedthem practice completing
measurement scales based on exercises written onthe blackboard.
This was important in producing valid responses.
Measuring BullyingSeveral surveys, mainly based on Family
Systems Theory (Minuchin,
1974), indicate that there is a strong link between
intra-familial and extra-familial peer relationships (Duncan,
1999a; Garcia, Shaw, Winslow & Yaggi,2000; Wolke & Samara,
2004). Building upon previous research, the question-naire included
items concerning bullying of both siblings and peers. The
itemsreferred to bullying in a physical/direct and a
psychological/indirect way. Thisis the only part of the
questionnaire in which items from a previous instru-ment were used
(the Negative Family Interactions Scale; Simonelli, Mullis&
Rohde, 2005). The previous instrument concerned familial
victimizationand the relevant items referred to psychological,
physical and sexual aggres-sion experienced by the respondent. Only
items from the first two types ofaggression were included in the
present survey and were changed to ask aboutbullying rather than
being victimized. The prevalence of bullying was definedaccording
to committing the act three or more times in the previous
sevenmonths, because bullying by definition involves repeated acts.
The -value for
-
Bullying 297
the overall sibling bullying scale was .89 and for the overall
peer bullyingscale was .90.
As in many non-English speaking cultures, capturing terms like
bullyingand bully in the Greek language is very difficult. Yet, the
terminology used in aquestionnaire can affect the results of the
survey. Smith, Cowie, Olafsson & Lie-fooghe, (2002, p. 1121)
give a nice example on how similar terms (e.g. bullying,teasing,
harassment, abuse) are associated with different connotations and
con-texts and may be understood differently by persons answering
questionnaires.An alternative to using global terms such as
bullying in questionnaire surveys isto ask for information on
particular acts (Smith et al., 2002, p. 1131) and this iswhat we
have done. Other researchers have also commented on how difficult
it isto render the term bullying in the Greek language as well as
on the negativeconsequences (with regard to admission rates) that
could follow from this diffi-culty (Kalliotis, 2000, p. 49;
Pateraki & Houndoumadi, 2001, p. 174).
Measuring Parental BondingWithin this survey, a parental bonding
instrument was constructed. Since
we did not use an existing instrument, exploratory factor
analyses were per-formed on the data from the 182 students. In all
cases, principal componentanalyses with varimax rotation and
eigenvalues greater than 1 were used,and a loading of 0.4 was taken
as the criterion for deciding whether an itemwould be retained in
the construction of the scale. Cronbach suggested that, ifseveral
factors exist, the -value should be calculated separately for
itemsrelating to different factors (Field, 2005, p. 668), and
therefore was calcu-lated separately for each subscale of the
questionnaire.
In the questionnaire, 50 questions were used to assess bonding,
but only42 questions were retained after the factor analyses. For
both mother andfather scores, the four factors were labelled
as:
1. childs perceptions of maternal/paternal effort;
2. communication;
3. psychological autonomy; and
4. emotional support.
Cronbachs for the total bonding score was .86 and .88 for mother
and fatherbonding respectively.
Measuring Fairness and DefianceA vignette-based methodology was
used to measure defiant and compliant
reactions to the sanctions imposed. The questionnaire included a
vignette inwhich the transgressor (a child) causes harm to a
sibling. In order to avoid any
-
298 M. M. Ttofi and D. P. Farrington
gender bias, female students were given stories in which both
the transgres-sor and the victim were female and similarly for male
students. The incidentdescribed in the vignette was ambiguous; the
transgressors responsibility forthe wrongdoing was not clearly
defined. Several surveys indicate that violentchildren are more
likely than non-violent ones to attribute negative intentionsto
others in ambiguous situations (Astor, 1994, p. 1055).
Sherman (1993, pp. 460461) offered some conditions under which
sanc-tions could be perceived as unfair. Following one of the
theory stipulationsthat the sanction must be excessivethe parents
in the vignette insulted thechild in the presence of the childs
friends. Children were asked to indicatehow fair they thought the
parents behavior was.
After these questions, children were asked to imagine that this
incident hap-pened to them, and they were asked to indicate what
they would have done.Children could choose one of four types of
response towards the sanctioningagents decision. In line with
defiance theory, two types of response specifieddirect defiance (I
would start yelling at my parents to defend myself) and indi-rect
defiance (I would do what my parents say but later on I would teach
mybrother/sister a good lesson when my parents are not watching).
However, giventhe young age of the children, it seemed plausible to
anticipate thatat least insome caseschildren would not defy their
parents mainly because they per-ceived themselves as less powerful
and not because they agreed with the sanc-tions imposed. Therefore,
we examined two other forms of reaction towards thesanctions
imposed, namely committed compliance (I would do what my parentssay
and feel that they are right) and situational compliance (I would
do whatmy parents say even though they are not right). The two
types of compliancewere defined according to Kochanska, Aksan and
Koenig (1995, p. 1753) who dis-tinguished between internally
driven, self-regulated committed complianceand situational
compliance when the child, though essentially cooperative,
nev-ertheless lacks sincere commitment and requires parental
sustained control.
Apart from the above four options, children were asked the
extent to whichthey would feel like giving their sibling a good
lesson (indirect defiance) andthe extent to which they would feel
like yelling at the parents to defend them-selves (direct
defiance). Given the sensitivity of the topic, it seemed
reasonableto worry about the possibility of some children
indicating what they would havedone based not on their real
emotions but on what they assumed they should do(social
desirability bias). In order to avoid this problem, we included the
abovetwo questions to detect childrens preference for direct and
indirect defiance.These last two items were used in the path
analysis (see later).
The Experimental Manipulation of Type of ChildThe wrongdoer of
the story was presented to the students (in a short
story along with relevant cartoons) before the actual vignette.
In half of the
-
Bullying 299
cases, the hero/heroine was described in a way that indicates
intentionality ofwrongdoing (bad child), while in the other half
the hero/heroine was pre-sented in positive terms so that
intentionality of wrongdoing could not beinferred (good child).
Students were asked to put themselves in the place ofthe child, so
that they would have different ideas about the childs
intentional-ity in regard to the wrongdoing. According to Sherman
(1993, pp. 460461)sanctions that appear to be undeserved are
related to perceptions of fairness.Therefore, it follows that the
sanctioning of the good child should be rated asmore unfair than
the sanctioning of the bad child.
Other researchers (Martin & Ross, 1996) found that, although
parentsconsidered that sibling physical aggression was a serious
transgression, theybelieved that mitigated aggression (including
provocation or reciprocity orlack of aggressive intent) was more
excusable. Therefore, they intervened lessoften to prohibit
mitigated than non-mitigated aggression, even when theaggression
was severe. Moreover, these researchers found that children
alsobelieved that mitigated aggression deserved less punishment.
Within thepresent survey, it seemed interesting to investigate
whether defiance levelswould vary according to the type of the
offender and his/her intentionality inregard to the
wrong-doing.
Special attention was paid to the administration of the two
types of ques-tionnaire. The questionnaires were separated in two
piles, one for boys andone for girls. Within each pile, the two
types of questionnaire (one with thegood child and one with the bad
child) were placed alternately. In this way,we could ensure that,
within each school, half of the students (chosen at ran-dom) would
receive one type of questionnaire and the other half would
receivethe other type of questionnaire. This alternate allocation
was equivalent tothe random assignment of students to experimental
conditions.
Measuring Unacknowledged ShameFollowing defiance theory, the
questionnaire included items on shame
management. Children were asked questions about whether they
would beashamed of themselves if they committed the act described
in the vignette.The questions were relevant to the possibility of a
child not acknowledgingshame and, instead, engaging in
neutralization techniques in order to displaceshame and to justify
his/her wrongdoing. Questions on neutralization tech-niques were
constructed based on the work of Sykes and Matza (2003). Thefactor
analysis of the relevant scores yielded two sub-scales (denial of
respon-sibility and condemnation of the condemners) with
eigenvalues greaterthan 1 and loadings greater than 0.4. Example
items were So what! He/shealways cries like that, but its not as if
he/she really means it (denial ofresponsibility) and My parents
always take it out on me (condemnation ofthe condemners). The
Cronbachs for the total scale was .86.
-
300 M. M. Ttofi and D. P. Farrington
RESULTS
Prevalence of BullyingTable 1 shows the bullying items, the
prevalence of different types of sib-
ling and peer bullying, and the significance of gender
differences in the totalbullying scores. The mean scores were
derived by scoring each item 1 (never),2 (once or twice), 3 (three
to six times), or 4 (seven times or more in the previ-ous seven
months) and adding up over all the items. The physical and
psycho-logical scores for sibling bullying (r=.75, p < .001) and
peer bullying (r = .73,p < .001) were highly correlated,
justifying the use of total bullying scores insubsequent
analyses.
On the total sibling bullying score, boys scored higher than
girls, and thisdifference was statistically significant [t(172) =
2.15, p = .033]. Similarly, onthe total peer bullying score, boys
scored higher than girls, and this difference
Table 1: Prevalence of sibling and peer bullying.
Items on Bullying
% Sibling Bullying
%Peer Bullying
Boys Girls Boys Girls
PhysicalPushed, shoved or pulled him/her 30.8 16.5 30.8
9.9Scratched and/or pinched him/her, but not for a joke 29.7 13.2
16.5 4.4Hit him/her on the face 14.5 7.7 18.7 5.5Threatened him/her
with, or used, a knife or any
other sharp object6.6 1.1 4.4 1.1
Threw things on him/her (e.g. shoe, pen, rubber) 30.0 20.0 25.3
13.2Pulled his/her hair 30.0 19.8 15.4 8.8
Mean Physical Bullying 11.2 9.4 10.2 7.9p-value **
***PsychologicalMade fun of him/her in a hurtful way 49.5 49.5 28.9
18.7Screamed at him/her 59.4 62.7 42.2 27.0Treated him/her like
he/she was stupid 40.7 20.9 24.5 17.6Turned other kids against
him/her NA NA 19.8 9.9Called him/her names or cursed at him/her
40.7 31.9 40.0 23.3Excluded him/her from games (did not let him/her
play) 26.4 14.3 18.7 5.5Mean Psychological Bullying 11.8 11.1 11.9
10.3p-value ns **Mean Total Bullying 22.9 20.5 22.1 18.1p-value *
***
Notes: ns = nonsignificant; * p < .05; ** p < .01; *** p
< .001Prevalence=% committing the act against the same person
three or more times in the previ-ous seven months.NA = not
applicable; in the sibling bullying scale, the wording of this item
was turned othersiblings against him/her. Because some children
came from families of only one or two sib-lings, this item was
omitted from the sibling bullying scale.
-
Bullying 301
was statistically significant [t(167) = 3.76, p < .0001].
Regarding differenttypes of bullying within the family, boys scored
higher than girls in both phys-ical and psychological bullying, but
the gender difference was significant onlyfor physical bullying.
Regarding different types of bullying within the school,boys scored
higher than girls in both physical and psychological bullying,
andboth differences were statistically significant.
Path Analysis ModelThis survey was theory-oriented: our aim was
to test the theoretical con-
structs of defiance theory and the interrelationships among
these constructsas proposed by the theory. This is why we conducted
path analysis usingAMOS 5.0.1 structural equation modelling
software, which is a useful tool fortheory testing. Maximum
likelihood estimation was used for the estimation ofthe model.
As a first step in this analysis, a hypothesized model was
developed in linewith the basic postulates of defiance theory
(Figure 1). Figure 2 represents thefinal path model that the sample
data fitted best. In Figure 2, all paths from
Figure 2: Applying Defiance Theory to Sibling and Peer Bullying;
Final Model.
-
302 M. M. Ttofi and D. P. Farrington
father bonding to other variables are omitted. Bonding towards
the sanction-ing agent (in this case, the parents) is an important
theoretical construct ofthe theory. Yet, the path coefficients from
father bonding to other observedvariables were very small and not
statistically significant, indicating that thecontribution of this
factor to the model was negligible. Thus, father bondingwas omitted
from the model. This did not affect the goodness-of-fit
indices.Moreover, we omitted all the path coefficients from mother
bonding to othervariables when these estimates were small and not
statistically significant.According to Schumacker and Lomax (2004,
p. 71) the statistical significanceof a path coefficient is an
intuitive indicator that should guide the researcher toomit some
paths from the model. At the same time, however, we acknowledgethat
the statistical significance of the estimates is dependent on
sample size.
Where paths were not clearly specified by defiance theory, we
chose themaccording to the best-fitting model. For example, it was
not clear a prioriwhether unacknowledged shame led to fairness,
whether fairness led to unac-knowledged shame, or whether there was
no path between the two constructs(see Figure 1). The assumption
that unacknowledged shame led to fairnessproduced the best-fitting
model (see Figure 2).
The first step in evaluating the results of the path analysis
was the examina-tion of the fit criteria, to determine whether the
hypothesized model fitted thesample data. Nine fit indices were
taken into consideration. For the total sample,the chi-square (x2 =
22.73), which is a global fit measure, was not statistically
sig-nificant (p = .16), indicating that the data were not
significantly different from themodel. The x2/df. ratio indicated a
good fit as well, since its value of 1.33 was lessthan 2.5. The
RMSEA was .04; according to Hu and Bentler (1999), its valueshould
be equal or less than .06. The NFI, the IFI, the TLI, the CFI and
the GFIall indicated a very good fit of the model (.94; .99; .97;
.99; .94 respectively).
When constructing structural equation models, 100 to 150
subjects is con-sidered to be the minimum satisfactory sample size
(Schumacker and Lomax,2004, p. 49). Moreover, both in SEM and path
analysis, the statistical signifi-cance of the path coefficients is
dependent on sample size. However, it wasconsidered useful to
examine the model for boys and girls separately (91 indi-viduals
each). Table 2 shows the path coefficients, and generally suggests
thatdirect defiance had greater effects on bullying for boys.
Because of the smallsample size for each gender separately, the
results for boys and girls should beregarded as tentative. The main
finding is that the data fitted the model(shown in figure 2) very
well. The path coefficients will be discussed below.
Type of Child, Perceptions of Fairness, Unacknowledged Shame,
and DefianceBased on the survey design, the type of the child
should affect perceptions
of fairness which, in turn, should influence defiant or
compliant reactions.
-
Bullying 303
Table 3 presents the interrelationships of constructs for each
gender sepa-rately and for the total sample. When the perpetrator
of the vignette wasgood (thus implying no intentionality of
wrongdoing) the majority of bothboys (68.4%) and girls (64.6%)
perceived the sanctioning agents behavior asunfair. The association
between the type of the child and fairness was sta-tistically
significant for boys, girls and the total sample. Based on the
pathanalysis, the type of child had a direct effect on fairness for
boys ( = .29) andthe total sample ( = .22), but not for girls.
Based on the way the values werecoded, sanctioning of the bad child
was considered more fair than sanction-ing of the good child.
Another variable that affected the perceptions of fairness was
unacknowl-edged shame. Table 3 shows that there was an association
between percep-tions of unfairness and unacknowledged shame. In the
path analysis, theparameter estimates from unacknowledged shame to
fairness were statisti-cally significant for girls ( = .26) and the
total sample ( = .17), but not forboys ( =.07).
The association between the type of child and defiance
(dichotomized intocompliance versus defiance) was not statistically
significant for the total sam-ple or for boys and girls separately
(Table 3). Looking at Table 2, the type ofchild influenced direct
defiance only for girls (= -.20). Based on the path anal-ysis, the
type of child significantly influenced indirect defiance for the
wholesample (= -.14).
Fairness was significantly related to defiance (Table 3). The
majority ofboys (62.8%) and girls (54.0%) who perceived the
sanctioners behavior asunfair responded with defiance. In the path
analysis, fairness influenced
Table 2: Standardized path coefficients.
Variable Relationships Boys Girls Total
Direct Defiance Sibling Bullying .26* .06 .15*Indirect Defiance
Sibling Bullying .11 .23* .20*Unacknowledged Shame Sibling Bullying
.27** .26* .25***Direct Defiance Peer Bullying .38*** .05
.20*Indirect Defiance Peer Bullying .23* .23* .28***Unacknowledged
Shame Peer Bullying .04 .17 .07Type of Child Direct Defiance .02
.20* .10Fairness Direct Defiance .37*** .23* .31***Unacknowledged
Shame Direct Defiance .29** .23* .26***Type of Child Indirect
Defiance .11 .17 .14*Fairness Indirect Defiance .29** .19*
.23***Unacknowledged Shame Indirect Defiance .40*** .36***
.37***Mother Bonding Fairness .26* .04 .14*Type of Child Fairness
.29** .16 .22**Unacknowledged Shame Fairness .07 .26* .17*Mother
Bonding Unacknowledged Shame .32** .40*** .36***
Note: *p < .05; **p < .01; ***p < .001.
-
304
Tab
le 3
:In
ter-
rela
tion
ship
s o
f co
nst
ruc
ts.
Boys
Girl
sTo
tal
With
in th
e ty
pe
of c
hild
Go
od
Chi
ldBa
d C
hild
Go
od
Chi
ldBa
d C
hild
Go
od
Chi
ldBa
d C
hild
% U
nfa
ir68
.434
.064
.644
.266
.338
.7p
-va
lue
***
***
% D
efia
nc
e48
.846
.938
.335
.743
.241
.8p
-va
lue
ns
ns
ns
With
in p
erc
ep
tions
of f
airn
ess
Fair
Unf
air
Fair
Unf
air
Fair
Unf
air
% D
efia
nc
e34
.162
.815
.454
.025
.358
.1p
-va
lue
****
***
*M
ea
n M
oth
er B
on
din
g83
.574
.984
.081
.183
.778
.3p
-va
lue
**n
s**
Me
an
Fa
the
r Bo
nd
ing
70.4
65.1
68.1
64.6
69.3
64.8
p-v
alu
en
sn
s*
Me
an
of
Un
-ac
kno
wle
dg
ed
Sh
am
e21
.324
.421
.124
.921
.224
.7p
-va
lue
**
**
With
in th
e ty
pe
of r
esp
ons
eC
om
plia
nce
De
fianc
eC
om
plia
nce
De
fianc
eC
om
plia
nce
De
fianc
e
Me
an
Mo
the
r Bo
nd
ing
82.6
75.4
82.8
81.8
82.7
78.2
p-v
alu
e**
ns
*M
ea
n F
ath
er B
on
din
g69
.565
.268
.262
.968
.864
.2p
-va
lue
ns
**
Me
an
of
Un
-ac
kno
wle
dg
ed
Sh
am
e19
.026
.920
.126
.520
.126
.7p
-va
lue
***
****
*M
ea
n S
iblin
g B
ully
ing
19.3
27.1
18.4
24.1
18.8
25.8
p-v
alu
e**
***
***
*M
ea
n P
ee
r Bu
llyin
g18
.326
.317
.519
.517
.923
.4p
-va
lue
***
ns
***
No
te: n
s =
no
nsig
nifi
ca
nt;
*p
< .0
5; *
*p