Journal of Adolescence 1999, 22, 461±466

Article No. jado.1999.0240, available online at http://www.idealibrary.com on

Concurrent and longitudinal links between friendship and

peer victimization: implications for befriending interventions
MICHAEL J. BOULTON, MARK TRUEMAN, CAM CHAU, CAROLINE WHITEHAND AND
KISHORI AMATYA

This short-term longitudinal study examined the associations between peer reports of
victimization and self-reported friendship among early adolescents. It was carried out
to test the view that friendship provides protection against victimization (called the
``friendship protection hypothesis''). Data were collected at two points within a school
year, separated by 6 months. At Time 1, 170 pupils provided data, and 158 of these did
so again at Time 2. Three sets of ®ndings were consistent with the friendship
protection hypothesis. First, at Time 1 those early adolescents who had a reciprocated
best friend in their home class received signi®cantly fewer peer nominations for
victimization than did classmates without a reciprocated best friend. Second, those
early adolescents that did not have best friend at either Time 1 or Time 2 showed the
highest increase in victimization over the course of the study, whereas those that did
have a best friend at both Times 1 and 2 showed the highest falls in victimization.
Third, a hierarchical multiple regression analysis indicated that a decrease in con¯ict
and betrayal reported to characterize the participants' best friendship was associated
with falls in victimization. The implications of these results for anti-bullying initiatives
based on ``befriending'' principles were discussed.
# 1999 The Association for Professionals in Services for Adolescents

Introduction

Studies have shown that many early adolescents are victimized by their peers and that such
experiences are associated with psychological disturbances including low self-esteem,
depression, loneliness and anxiety (Olweus, 1993; Boulton and Smith, 1994; Craig, 1998;
Hawker and Boulton, 1998). These ®ndings have prompted the development of anti-bullying
initiatives that have taken a wide variety of different forms (see Smith and Sharp, 1994;
Ross, 1996). One recent approach can be considered under the heading of ``befriending''.
The essence of this type of intervention is to train pupils to provide caring support for peers
that are victimized in the hope that this will alleviate their distress and enable them to better
resist victimization in the future. To date, most published accounts of interventions of this
type have described small-scale studies (see Cowie and Sharp, 1996) but it appears to be
growing in popularity.
Indirect support for such an approach comes from the theoretical (Hartup, 1996) and
empirical (Hodges et al., 1997; Boulton and Chau, 1998) literatures which consider how
school pupils might bene®t from having good quality friendships. However, much less
research has been carried out to determine the actual associations between friendship and
victimization. One robust ®nding from North America is that early adolescents without
friends are more likely to be victimized than those who do have friends (Bukowski et al.,

Reprint requests and correspondence should be addressed to M. J. Boulton, Department of Psychology, Keele
University, Keele, Staffordshire ST5 5BG, U.K.

0140-1971/99/040461+06 $30
00/0 # 1999 The Association for Professionals in Services for Adolescents
462 M. J. Boulton et al.

1995; Hodges et al., 1997). It is tempting to take this as evidence for the hypothesis that
friendship protects against victimization (called the ``friendship protection hypothesis''
henceforward). However, we caution against doing so since it is possible that this ®nding
merely re¯ects pupils' tendencies to select non-victimized peers as their friends. A better test
of the friendship protection hypothesis would come from a longitudinal investigation in
which the association between changes in having versus not having a friendship and changes
in victimization are examined. To the best of our knowledge, no such study has been reported
in the literature. Thus, at present there is scant direct evidence, and none outside North
America, that having a friend reduce the likelihood that early adolescents will experience
victimization. If this could be established, it would provide a stronger rationale for the use of
befriending as an anti-bullying intervention.
It is not only having versus not having a friend that is relevant to the friendship protection
hypothesis. In addition, Hartup (1996) also drew attention to two other friendship
variablesÐthe identity of friends, and friendship qualityÐand claimed that both can
in¯uence social adaptation in complex ways. As far as we are able to tell, researchers have
not considered either of these variables in relation to the friendship protection hypothesis.
The present study will extend knowledge by investigating the degree of con¯ict and betrayal
within early adolescents' friendships, and the association between this and victimization.
Many psychologists accept that con¯ict and betrayal, along with positive features (such as
help and support), make up the two key friendship quality dimensions (see Hartup, 1996).
We think a focus on con¯ict and betrayal is important because one way that peers protect
their friends from victimization appears to be by retaliating on their behalf, thus making
future attacks less likely (Boulton and Chau, 1998). Thus, we suggest that friends may be less
likely to help each other in this way if their friendship is characterized by high levels of
con¯ict and betrayal.
Building on the arguments presented in the preceding paragraphs, the aims of the present
study were as follows. First, to try to replicate in the U.K. the ®nding from North America
that friendless early adolescents are more likely to be victimized than befriended early
adolescents. Second, to provide the ®rst investigation of the association between changes in
having/not having a reciprocated best friend and changes in victimization. Third, again for
the ®rst time, to examine the association between changes in degree of con¯ict and betrayal
within friendship and changes in victimization.

Method

Participants
The early adolescents were drawn from ®nal year classes (selected at the head teachers'
discretion) at ®ve junior schools in the U.K. The schools were selected on a convenience
basis. Parental and child consent was obtained. At Time 1 there were 170 participants (equal
numbers of girls and boys) with a mean age of 11
3 years. At Time 2, 85 girls and 73 boys
(n=158) from the original sample again provided data.

Measures
Victimization was indexed by means of peer nominations. Participants were shown class list
containing all members of their class, and they were asked to indicate who was victimized in
 Friendship and victimization 463

each of four different ways: ``Gets hit and kicked by other kids for no reason,'' ``Gets called
nasty names by other kids for no reason,'' ``Gets left out of games and things by other kids for
no reason,'' and ``Gets called nasty things behind their backs by other kids for no reason.''
Participants were told that they could give as many or as few nominations as they wanted.
The four items tap the different manifestations of victimizationÐphysical, verbal, indirect
and relationÐidenti®ed in prior research (BjoÈrkqvist et al., 1992; Crick and Grotpeter,
1995). These items were found to be highly correlated (mean r=0
77, all p50
01) and had
excellent internal reliability (alpha=0
093 at Time 1). Consequently, a Victim score was
computed for each child as the mean percentage of classmates that nominated them as a
victim across the four separate items. Other researchers (e.g. Egan and Perry, 1998) have also
found that in some samples of around the same age as those tested here, different types of
victimization are highly correlated.
To determine whether or not each participant had a reciprocated best friend, we modi®ed
the procedure of Parker and Asher (1993). Each pupil was asked to say who in their class was
their ``very best friend''. Only participants who reciprocated each other's nominations were
deemed to meet this criterion.
The degree of con¯ict and betrayal characteristic of each participant's best friendship
(regardless of whether it was reciprocated or not) was assessed by means of the 7-item
subscale of the same name as in Parker and Asher's (1993) Friendship Quality
Questionnaire. Here, some items were slightly modi®ed because they contained terms that
caused confusion in pilot testing (e.g. ``my best friend and I bug each other a lot'' was
changed to ``my best friend and I get on each others' nerves a lot''). Each participant was
asked to respond to the items with reference to their friend in the class, for example ``x (best
friend) and I get angry with each other'', and ``x (best friend) says mean things about me to
other kids''. In the present study, each item was followed by a 5-point response option
anchored by ``doesn't happen at all'' and ``happens all the time''. Each item was scored from
one to ®ve so that a high score indicated greater con¯ict and betrayal. The items had good
internal reliability (alpha=0
075 at Time 1), and so the Con¯ict and Betrayal score was
computed as the mean across them all.

Procedure
Data were collected in individual interviews in a quiet location away from other pupils. At
the outset, participants were informed that they were not being tested, that there were no
right or wrong answers, and that what they said would not be shared with anybody from their
school. Time 1 data were collected during November and December, and Time 2 data during
May and June of the same school year.

Results

Victimization in participants with and without a reciprocated best friend

At Time 1 signi®cantly more victim nominations were received by participants without a
reciprocated best friend (mean=32
8, S.D.=18
1) than by those with a reciprocated best
friend (mean=25
3, S.D.=15
4), t(168)=2
78, p50
01.
464 M. J. Boulton et al.

Associations between changes in friendship status and changes in victimization

Participants were classi®ed into one of four groups on the basis of whether or not they
maintained their status of having/not having a reciprocated best friend at Times 1 and
2 (called ``friendship status''). One group had a reciprocated best friend at both Times 1 and
2 (called ``Friend at Times 1 and 2'', n=25), another had a friend at Time 1 but not Time 2
(called ``Friend at Time 1 only'' n=40), a third did not have a friend at Time 1 but did so at
Time 2 (called ``Friend at Time 2 only'' n=43) and the fourth did not have a friend at either
Time 1 or Time 2 (called ``No friend'' n=46). This grouping was used as an independent
variable, along with sex, in a 4 (friendship status)62 (sex) analysis of variance (ANOVA). The
dependent variable was change in Victim score from Time 1 to Time 2, and it was made up of
the standardized residuals from regressing participants' Time 2 Victim scores on to their Time
1 Victim scores. The main effect of friendship status was highly signi®cant, F(3, 146)=9
37,
p50
001. The ``Friend at Time 2 only'' group (mean=0
23, S.D.=1
20) and especially the
``No friends'' group (mean=0
62, S.D.=0
76) showed an increase in victimization over the
duration of the study. The ``Friend at Time 1 only'' group (mean=70
14, S.D.=0
056) and
especially the ``Friend at Times 1 and 2'' group (mean=70
43, S.D.=0
94) were victimized
less over the course of the study. Follow-up Duncan's Multiple Range tests indicated that
(i) the ``No friends'' group showed a signi®cantly greater increase in victimization than all of
the other groups and (ii) that the ``Friend at Times 1 and 2'' group showed a signi®cantly
greater fall in victimization than the ``No friends'' group and the ``Friend at Time 2 only''
group. Neither the sex main effect nor the interaction effect was signi®cant in the analysis of
variance. To help interpret the results of the ANOVA, the mean Victim scores for the four
friendship status subgroups are shown in Table 1.

Predicting changes in victimization from changes in friendship con¯ict and

betrayal
A hierarchical multiple regression analysis was used to determine the extent to which change
in victimization could be predicted from change in Con¯ict and Betrayal score. Sex was
entered at step 1, change in Con¯ict and Betrayal score was entered at step 2, and the sex 6
change in con¯ict and Betrayal score entered on the ®nal step. The overall model was
signi®cant [F(3, 154)=2
70, p50
05]. Of greatest interest was the ®nding that change in
Con¯ict and Betrayal score signi®cantly predicted changes in victimization, even after the
effect of sex had been controlled (F change=7
85, p50
01; R2 change=0
05). The positive
beta for this variable (0
22) indicates that a decrease in Con¯ict and Betrayal score was
associated with a fall in victimization. The interaction effect was not signi®cant, indicating
that this effect held for both girls and boys.

Table 1 Mean (and standard deviation) Victim scores for the four friendship groups at
Time 1 and Time 2
Friendship group Time 1 Time 2
Friend at Times 1 and 2 24
6 (13
3) 20
8 (13
9)
Friend at Time 1 only 28
0 (15
6) 26
6 (17
1)
Friend at Time 2 only 37
9 (20
2) 39
8 (23
5)
No friend 29
4 (17
6) 31
8 (19
4)
 Friendship and victimization 465

Discussion

In line with earlier North American studies (see Introduction), those participants with
reciprocated best friend received signi®cantly lower concurrent Victim scores than their peers
who did not have such a relationship. While this ®nding is consistent with the friendship
protection hypothesis, other explanations are also possible (see below). The original
contributions of this investigation stem from the longitudinal analyses. Those participants
who had a reciprocated best friend at both Times 1 and 2 showed a fall in victimization,
whereas those participants who lacked a friend at both Times 1 and 2 showed an increase in
victimization. Importantly, our data also help rule out the possibility that the association
between friendship and victimization found in the concurrent analyses is due to individuals
choosing as friends peers who are low in victimization. If this was true, we would not have
found that those who only had a reciprocated friend at Time 2 but not at Time 1 showed a
(slight) increase in victimization, and that those who had a reciprocated friend at Time 1
only showed a (slight) decrease in victimization (see Table 1).
The latter two results also suggest that there may be an inoculating effect of having an
earlier reciprocated friendship that persists for some time even when that friendship is lost,
and that an earlier lack of reciprocated friendship carries a future risk of victimization even
when a such a friendship is acquired later. These results lead us to suggest that there may be a
process of labelling with regard to peer friendships, and that this means that any protection
conveyed by friendships or risk associated with their absence takes time to accrue.
Our study also points to the importance of friendship quality since decreases in friendship
con¯ict and betrayal predicted a decline in peer victimization. Whether this result can be
explained in terms of friends being more likely to help partners avoid victimization when
their friendship is characterized by low levels of con¯ict and betrayal, or because friends who
share this type of friendship do not victimize one another, or because of some other
mechanism, remains an important issue for future research to address.
This study, by adopting a longitudinal design, has provided the strongest support to date
for the friendship protection hypothesis. However, even with such a design it would be
premature to imply causal relationships between friendship and victimization since it remains
possible that the skills used to form and maintain friendship are also the skills employed to
avoid being bullied. Unravelling causal developmental sequences remains a dif®cult but
worthwhile challenge for researchers working in this area.
It is pertinent to highlight several weaknesses and limitations with the present study. With
respect to friendship, we only indexed one aspect of friendship quality (i.e. con¯ict and
betrayal). It is not unreasonable to suggest that positive aspects of friendship quality, such as
help and support, may have at least as much relevance to the friendship protection
hypothesis as negative aspects. Additionally, we did not investigate whether children who
has a reciprocated best friend at Times 1 and 2 maintained the same friend or a formed a best
friendship with a different classmates. Again, intuition might suggest that enduring
friendships offer most protection against victimization. Finally, in the present study, due to
the high degree of correlation between our four victimization items, it was not possible to
investigate if friendship variables offer different levels or types of protection against the
various manifestations of peer victimization. For example, it seems plausible to suggest that
having a best friend may offer more protection against deliberate social exclusion than
against physical assaults, since the former does not require the friend to directly confront the
aggressor. In contrast, children may be less likely to stand up for friends who are being
466 M. J. Boulton et al.

physically victimized, since doing so may place them at risk of victimization themselves. It
would be useful if all of the issues raised in this paragraph were addressed in future studies.
In conclusion, friends, by de®nition, share a warm and helping reciprocal relationship. As
such, and more than any other members of the peer group, friends may be highly motivated
to help each other resist victimization or deal with its after effects should it occur.
Investigations of the mechanisms through which friendship helps protect pupils from
victimization are just beginning (Boulton and Chau, 1998). Such studies can help facilitate
the development of optimally effective befriending interventions. Speci®cally, practitioners
could then ensure that the friendships they encourage pupils to adopt have the charac-
teristics most likely to convey protection from victimization.

Acknowledgement

I am indebted to the staff and pupils from the participating schools, and to the Keele
University sabbatical system which provided time for the ®rst two authors to prepare this
report.

Reference
BjoÈrkqvist, K., Lagerspetz, K. M. J. and Kaukiainen, A. (1992). Do girls manipulate and boys ®ght?
Developmental trends to regard to direct and indirect aggression. Aggressive Behavior, 18, 117±127.
Boulton, M, J. and Chau, C. (1998). Friendship moderates the harmful effects of peer victimization.
Manuscript submitted for publication.
Boulton, M. J. and Smith, P. K. (1994). Bully/victim problems in middle-school children: stability, self-
perceived competence, peer perceptions and peer acceptance. British Journal of Developmental
Psychology, 12, 315±329.
Bukowski, W. M., Sippola, L. K. and Boivin, M. (1995, March). Friendship protects `at risk' children
from victimization from peers. In The role of friendship in children's developmental risk and resilience,
Price, J. M. (chair), Symposium at the biennial meeting of the Society for Research in Child
Development, Indianapolis, IN.
Craig, W. M. (1998). The relationship among bullying, victimization, depression, anxiety, and
aggression in elementary school children. Personality and Individual Differences, 24, 123±130.
Crick, N. and Grotpeter, J. K. (1995). Relational aggression, gender and social-psychological
adjustment. Child Development, 66, 710±722.
Cowie, H. and Sharp, S. (Eds) (1996). Peer Counselling in Schools: a time to listen. London: David Fulton.
Hartup, W. (1996). The company they keep: friendship and their developmental signi®cance. Child
Development, 67, 1±13.
Hawker, D. S. and Boulton, M. J. (1998). Peer victimization: cause and consequence of psychosocial
maladjustment. Manuscript submitted for publication.
Hodges, E.V. E., Malone, M. J. and Perry, D. G. (1997). Individual risk and social risk as interacting
determinants of victimization in the peer group. Developmental Psychology, 33, 1032±1039.
Olweus, D. (1993). Bullying at School: what we know and what we can do. Oxford, England: Blackwell.
Parker, J. G. and Asher, S. R. (1993). Friendship and friendship quality in middle childhood: links with
peer group acceptance and feelings of loneliness and social dissatisfaction. Developmental
Psychology, 29, 611±621.
Ross, D. M. (1996). Childhood Bullying and Teasing: what school personnel, other professionals, and parents
can do. Alexandria, VA: American Counselling Association.
Smith, P. K. and Sharp, S. (Eds) (1994). School Bullying: insights and perspectives. London: Routledge.