British Journal of Developmental Psychology (2011), 29, 110–123
C 2010 The British Psychological Society
On-line ostracism affects children differently from
adolescents and adults
Dominic Abrams∗ , Mario Weick, Dominique Thomas,
Hazel Colbe and Keith M. Franklin
Centre for the Study of Group Processes, Department of Psychology,
University of Kent, Canterbury, UK
This research examines adults’, and for the first time, children’s and adolescents’
reaction to being ostracized and included, using an on-line game, ‘Cyberball’ with
same and opposite sex players. Ostracism strongly threatened four primary needs
(esteem, belonging, meaning, and control) and lowered mood among 8- to 9-yearolds, 13- to 14-year-olds, and adults. However, it did so in different ways. Ostracism
threatened self-esteem needs more among 8- to 9-year-olds than older participants.
Among 13- to 14-year-olds, ostracism threatened belonging more than other needs.
Belonging was threatened most when ostracism was participants’ first experience in the
game. Moreover, when participants had been included beforehand, ostracism threatened
meaning needs most strongly. Gender of other players had no effect. Practical and
developmental implications for social inclusion and on-line experiences among children
and young people are discussed.
There are multiple reasons why people may be excluded from social relationships
(Abrams & Christian, 2007; Abrams, Hogg, & Marques, 2005). Psychologically, an
especially salient manifestation of exclusion is ostracism – being ignored and excluded
from participating in social activity (Williams & Zadro, 2005). This research examines
the previously untested questions of whether and how inclusion and ostracism affects
children and adolescents differently from adults.
Williams (2007) proposes that four fundamental needs are reflexively affected by ostracism: belonging, self-esteem, control, and meaningful existence. When the satisfaction
of these needs is reduced by ostracism, the immediate effect should be a more negative
mood. These needs are interconnected but have distinct psychological characteristics.
In adults, the need to belong (Baumeister & Leary, 1995) is affected strongly by ostracism
(Zadro, Williams, & Richardson, 2005). Self-esteem, an important component of positive
mental health, is threatened by ostracism if victims infer something about them is wrong
or socially devalued. Self-esteem is linked to belonging because it can serve as an index
of being accepted and valued by others (Leary, Tambor, Terdal, & Downs, 1995).
∗ Correspondence should be addressed to Professor Dominic Abrams, Centre for the Study of Group Processes, Department
of Psychology, University of Kent, Canterbury, Kent CT2 7NP, UK (e-mail: email@example.com).
Ostracism: children, adolescents, adults
Ostracism undermines the sense of meaningful existence by making victims feel
invisible to others (Williams, 2007). Feelings of uncertainty are uncomfortable and may
prompt a more active search to define and categorize oneself (e.g. Hogg, 2007). People’s
sense of control is also weakened when they are given the ‘silent treatment’ (Williams,
Shore, & Grahe, 1998), potentially creating feelings of hopelessness or helplessness
that may prompt strategies to regain control (Bandura, 2000; Metalsky, Halberstadt, &
Williams (2007) theorizes that need threats and mood are ‘hard wired’ reflexive
responses to ostracism that, in a subsequent reflective stage, are linked to different
coping responses. Esteem and belonging threats are associated with prosocial responses
whereas control and meaning threats relate to aggression and antisocial behaviour
(Warburton, Williams, & Cairns, 2006), possibly being implicated in school shootings
(Leary, Kowalski, Smith, & Phillips, 2003).
Primarily using student participants, researchers have explored different modes of
ostracism, ranging from being left out of a ball-tossing game, to being ignored on a train
(Zadro et al., 2005). Direct ostracism from Internet chat rooms affects the four needs
(Williams et al., 2002), as does impersonal on-line ostracism when the ball-tossing game
is played as a computer game labelled ‘Cyberball’ (Williams, Cheung, & Choi, 2000;
Williams & Jarvis, 2006), even when people feel they are playing a computer rather than
other people (Zadro, Williams, & Richardson, 2004), and regardless of whether they are
playing with in-group or with despised out-group members (Gonsolkorale & Williams,
2007; Williams et al., 2000).
However, ostracism can occur even during preschool years (Crick, Casas, & Ku,
1999). Over and Carpenter (2009) showed indirectly that 5-year-old children are sensitive
to ostracism. Children who were primed with videos in which one shape ostracized
another subsequently imitated the actions of a model more closely than did children in
a control condition.
Social cliques during school years create potential for frequent and unavoidable
ostracism. From around the age of 7 years children begin to grasp the social dynamics
involved in inclusion and exclusion from social groups (Abrams, Rutland, Pelletier,
& Ferrell, 2009), so they should also be sensitive to ostracism. Given the serious
psychological consequences of being excluded from social networks in childhood and
adolescence (Buhs, Ladd, & Herald, 2006; Hymel, Vaillancourt, McDougall, & Renshaw,
2002), it is important to examine whether ostracism has a similar effect on children
and adolescents as it does on adults. No previous research has addressed this issue
The rise of Internet social networking sites may increase children’s vulnerability to
ostracism. Computers are used commonly from the age of 7 years in many schools, so
it is important to establish whether children respond to cyberostracism in the same
way as adolescents and adults. Previous research on cyberbullying has focused on direct
manifestations of abuse and insults (Li, 2006) rather than the more indirect, and perhaps
common, form of exclusion through ostracism.
Valkenburg and Peter (2007) showed that Internet communication is especially
important for preadolescents’ and adolescents’ social relationships. With a sample of
794 participants aged 10–16 years, self-disclosure via Internet peaked at around the
age of 15 years, consistent with the idea that social connection and belongingness are
especially focal at this stage of development. At present, there is little clear evidence
about whether adolescents will respond to cyberostracism in the same way as adults,
and whether needs are threatened differently. Insight into these issues has implications
Dominic Abrams et al.
for theories about the psychology of social exclusion and has practical implications for
strategies to deal with social exclusion during the school years.
As well as extending the reach of Williams’ theory, a practical ambition for the
present research was to devise a version of Cyberball for use with children from middle
childhood as well as with adults. We therefore examined effects of cyberostracism among
8- to 9-year-olds, 13- to 14-year-olds, and introductory level undergraduate students aged
around 20 years. Establishing the viability of the Cyberball paradigm across this age
range offers a valuable tool for social developmental researchers. Traditional techniques
for studying peer exclusion often rely on labour-intensive methods such as evaluation
of peer networks and peer nomination techniques. As well taking substantial time,
these approaches entail significant practical, ethical, confidentiality, and data protection
hurdles. Moreover, these techniques are not very amenable to manipulating the source or
form of exclusion. The Cyberball paradigm circumvents these problems, is convenient
and engaging for the participants, and allows us to examine responses to ostracism
without referring to pre-existing relationships.
We examine four theoretically driven questions. First, are the four need states
threatened equally by ostracism among children, adolescents, and adults? This has not
been tested before. Adult research often aggregates the four needs into a single index
(e.g. van Beest & Williams, 2006), obscuring the possibility that each need is affected
distinctively, and ignoring the original conceptual distinction among the needs (Williams
et al., 2000). Cyberostracism is likely to affect need states and mood among all three age
groups, but we contend that different needs may be affected to different degrees during
Children may experience stronger esteem need threat from peer ostracism than do
adolescents and adults because children’s self-esteem may be less stable and less well
rooted in their other social relationships outside the context of the experiment. Children
may also be more likely to interpret ostracism with uncertainty, indicating that they have
done something wrong or made a mistake.
Prior research shows that early adolescents’ affective responses to peer rejection
are clearly negative (Reijntjes, Dekovic, Vermande, & Telch, 2007; Reijntjes, Stegge,
Terwogt, Kamphuis, & Telch, 2006). It seems likely that adolescents aged 13–14 may
find ostracism especially threatens their need to belong because they have a strong
focus and dependency on peer acceptance (Harris, 1995; Valkenburg & Peter, 2007).
Consistent with this idea, a recent study of cyberostracism effects among a small sample
of female adolescents and students revealed that adolescents showed stronger affective
reactions (Sebastian, Viding, Williams, & Blakemore, 2010). That study did not specifically
compare differences in levels of threat to different needs, but the effect size of ostracism
was highest for belonging need threat, in line with the present theorizing. Secondary
schools in the UK place adolescents in multiple classes whose membership changes from
lesson to lesson, depending on the subject matter. This may result in a less controllable or
stable network of social relationships compared with primary schools (8- to 9-year-olds
typically have remained with their classmates for all lessons throughout the preceding 4
University students often have romantic partners, well-established ex-school peer
networks, and opportunities to choose their group affiliations from a wide array. Thus,
age differences can be expected in the need threats following ostracism.
A second new question is whether need threat depends on whether the person has
previously been included. It is extremely surprising that this has not been explored
systematically in the social psychological literature on ostracism. We expect that, in the
Ostracism: children, adolescents, adults
absence of alternative inclusive relationships in a particular situation, ostracism during
the first encounter with other people may most strongly threaten belonging needs.
However, if inclusion has occurred first, ostracism may raise levels of uncertainty about
what is happening, threatening meaning. We test these ideas by comparing responses
when ostracism either precedes or follows inclusion – a distinctive feature of the present
research. It is also conceivable that children, adolescents, and adults might respond
differently to the sequence of ostracism versus inclusion. Although we do not have a
strong developmental hypothesis regarding this issue, the present research will reveal
whether there is an empirical basis for this possibility.
Third, we investigate how need threats relate to mood (specifically enjoyment).
Although ostracism is an aversive experience, adult, and adolescent research shows
no consistent connection between need threat and mood (Sebastian et al., 2010;
Williams, 2007). This could be attributable to the nature and timing of measurement
or to participants’ coping strategies. Adults may be able to enjoy themselves even when
cyberostracized because they may treat it as ‘just a game’, and recognize that needs can
be satisfied beyond the game. Thus, there could be a closer relationship between need
threats and mood among younger than older participants.
A fourth question is whether ostracism from in-group and out-group members has
similarly negative impacts on children and adolescents. Previous research on adults
suggests that ostracism is equally painful regardless of whether is perpetrated by ingroup our out-group members. We varied whether the game was played with females
or with males. Younger children are known to show greater gender in-group preference
than older children (Powlishta, Serbin, Doyle, & White, 1994; Verkuyten & Thijs, 2001).
Thus, we might expect larger effects for same-gender ostracism among 8- to 9-year-olds
than among adolescents and adults.
Sixty-eight males and ninety-eight females participated, of whom 41 were 8- to 9-yearolds, 79 were 13- to 14-year-olds, and 46 were introductory psychology students (mean
age = 20 years). Gender was balanced within the younger age groups. There were
more females (34) than males (12) among the students. Participants completed the
Cyberball game in a university or school computer room. Participants were ethnically
homogeneous, White middle class from the south-east of England. For the 8- to 9-yearolds, the procedure was introduced verbally as well as on-screen so that the (female)
experimenters could ensure children understood the instructions.1
Design and materials
The design was 4 (needs) × 3 (trial: first inclusion, second inclusion, ostracism) × 3 (age
group) × 2 (sequence of ostracism: first trial, second trial). Age group and sequence
were between-participants factors with random assignment to condition. Needs and
trial were within-participants factors. Additionally, gender of participant was treated as
a factor and gender of excluder was varied with random assignment.
Williams et al.’s (2000) Cyberball game was redesigned for children as well as adults.
Specifically, the original game is displayed in a small area of the computer screen which
children, four adolescents, and thirteen students were excluded from the study because of non-completion due to time
constraints (lessons starting). Five students were excluded because they were more than 3SD above the age range.
Dominic Abrams et al.
might have been difficult for some children to view. The new version is set in a larger
format that fills the screen. The original presents amorphous figures to represent the
players whereas in the new version the other players are depicted by their names
to reduce children’s possible inference that the other players are not real people (a
question that arose during pilot work). The presentation of names also makes it easier to
incorporate gender information about the other players unobtrusively and continuously
while the game is played.
Participants were asked to sit silently at a computer, viewing a demonstration screen
showing three players passing the ball to each other. Two players were positioned
centre left and right hand, the third was positioned centre bottom of the screen. A
yellow ball was ostensibly thrown between the players in a looping movement to mimic
throwing. After a brief demonstration, participants were asked to enter their age in a
box on screen and to check a box showing their gender. They were informed that they
would be playing the game with two people who were using computers elsewhere. The
left- and right-hand player’s positions were populated with names (both male or both
female), while the centre player was labelled ‘YOU’. Participants were shown how to
left-mouse-click the player to whom they intended to throw the ball. Having checked
that they could operate this effectively, the game began.
Three rounds of the game (trials) were presented, in the sequence ostracism–
inclusion–inclusion, or inclusion–ostracism–inclusion, ensuring that the experiment
concluded with an inclusion experience. In line with the methods used by Williams
and colleagues, trials involved 12 tosses of the ball among the three players. In inclusion
trials, participants (and the other players) each received the ball four times evenly spread
across the trial. In the ostracism trial, participants only received the ball one-sixth of the
time (twice from each player) at the start of the trial, following which the other players
only passed the ball to each other. At the end of each trial separate screens presented
items for the manipulation check, a question about each of the four needs, and about
their enjoyment. Response scales were depicted using icons rather than a numeric scale,
because previous research established that iconic scales were understood easily by 8- to
9-year olds (e.g. Abrams et al., 2009).
The manipulation check simply asked ‘how much did they throw you the ball’,
where the icons ranged to denote a lot (1) to not at all (scored as 5). Following pilot
work to ensure children could comprehend the measures, the remaining items were
adapted from past research (van Beest & Williams, 2006; Zadro, Boland, & Richardson,
2006; Zadro et al., 2004). Participants were asked how they felt during the game. As in
Williams et al. (2000), each need was assessed using a single item, the order of which
was randomized in each trial. Self-esteem was measured by the question ‘I felt good
about myself’, belongingness was measured by the question ‘I felt like the odd one out’,
control was measured by the question ‘I felt in charge during the game’ and meaningful
existence was measured by the question ‘I felt invisible’. Next, mood was measured with
the item, ‘I enjoyed playing the game’. Responses were made by selecting an icon that
ranged from 1 (very much) to 5 (very little).2
2 Enjoyment is not a pure measure of mood. However, comparable to Williams et al., in another data set involving children
aged 8 (N = 81), we measured both sadness and enjoyment. Within inclusion and ostracism trials these were quite highly
correlated (rs = .64 and .57, respectively) giving us good grounds for assuming the enjoyment measure partially taps mood
Ostracism: children, adolescents, adults
Overall effects of ostracism versus inclusion
Table 1 shows that the manipulation of ostracism was effective. Participants experienced
receiving the ball significantly less in the ostracism than the inclusion trials. All four
needs states were higher, and mood was worse, following ostracism than inclusion. As
hypothesized, these differences were significant within all three age groups. Overall,
control need threat was significantly higher than all others (ps < .001), esteem need
threat was higher than threats to belonging and meaning (ps < .01) and belonging
and meaning need threat did not differ significantly. Because there were no differences
between need threats on the two inclusion trials we averaged these for subsequent
analyses. However, as a precaution, we also conducted analyses comparing the ostracism
trials only with the inclusion trial that followed ostracism. These yielded precisely the
same pattern of significant findings and are available on request from the first author.
We conducted a 4 (needs) × 2 (type of trial: inclusion vs. ostracism) within-participants,
by 3 (age) by 2 (sequence: ostracism first, inclusion first) between-participants ANOVA.
There were significant main effects of age, F(2, 160) = 5.15, p < .01, 2 = .06, and
sequence, F(1, 160) = 5.20, p < .05, 2 = .03 but not their interaction. There were
significant main effects of trial, F(1, 480) = 305.03, p < .001, 2 = .66, need, F(3,480) =
37.87, p < .001, 2 = .191, and their interaction, F(3,480) = 14.14, p < .001, 2 =
.08. The trial by need interaction indicates that size of the effects of ...
Purchase answer to see full