540757
research-article2014
GPI0010.1177/1368430214540757Group Processes & Intergroup RelationsJans et al.
Group Processes &
Intergroup Relations
Article
The development of group
influence on in-group identification:
A multilevel approach
G
P
I
R
Group Processes & Intergroup Relations
2015, Vol. 18(2) 190­–209
© The Author(s) 2014
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1368430214540757
gpir.sagepub.com
Lise Jans,1 Colin Wayne Leach,2 Randi L. Garcia,2 and Tom Postmes1
Abstract
Research on in-group identification typically focuses on differences in individuals’ identification at the
individual level of analysis. We take a multilevel approach, examining the emergence of group influence
on identification in newly formed groups. In three studies, multilevel confirmatory factor analysis
confirmed two dimensions of identification—self-definition and self-investment (Leach et al., 2008)—
at both the individual and the group level. As expected, the group had greater influence on individuals’
identification the more group members interacted with each other. This was shown in experiments with
varying amounts of real interaction (Study 1), in a longitudinal study of student project groups (Study
2), and in a longitudinal study that experimentally mimicked the development of online communities
(Study 3). Together, these studies support a developmental model of identification at the group level
that has implications for the understanding of social identity and small-group dynamics.
Keywords
group dynamics, group formation, in-group identification, interaction, multilevel, small group, social
identity
Paper received 3 February 2014; revised version accepted 13 May 2014.
Despite the widespread assumption that individuals’ identification is at least partly a function of
group-level processes, most research on identification focuses on the individual level of analysis
(see Ashmore, Deaux, & McLaughlin-Volpe,
2004; Ellemers, Spears, & Doosje, 1999; Haslam,
Turner, Oakes, McGarty, & Reynolds, 1997).
Thus individual differences in in-group identification have been shown to affect individuals’ selfconcepts, relations with others, and social
behavior (for reviews, see Ashmore et al., 2004;
Ellemers et al., 1999). However, it is typically
unclear to what extent individuals’ identification
is based on membership of a shared in-group that
also affects other group members’ identification.
Technically speaking, the influence of the ingroup on members’ identification cannot be
examined at the individual level of analysis that is
1University
2University
of Groningen, The Netherlands
of Connecticut, Storrs, USA
Corresponding author:
Lise Jans, Department of Social Psychology, University of
Groningen, Grote Kruisstraat 2/1, 9712 TS, Groningen, The
Netherlands.
Email: [email protected]
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
191
used in most previous research; it must be examined by making appropriate comparisons between
groups. In this paper we examine in-group influence on in-group identification. We use a multilevel modeling statistical approach to study
group-level (between-group) and individual-level
(within-group) emergence of in-group identification in newly formed groups.
The Multilevel Nature of
Identification
Identification is typically assumed to be more
than a chronic individual difference (e.g.,
Ashmore et al., 2004; Ellemers et al., 1999). It has
been operationalized as the extent to which people feel attached to a specific group in a particular
context (Ellemers et al., 1999) because research
suggests that the nature of the group can influence members’ identification. For example,
group-level factors such as group size, status, and
salience (Brewer, 1991; Brewer, Manzi, & Shaw,
1993; Ellemers, van Knippenberg, De Vries, &
Wilke, 1988; van Dick, Wagner, Stellmacher, &
Christ, 2005) affect individuals’ degree of identification. Despite this acknowledgement that the
group can influence group members’ identification, the extent to which the group influences
identification has rarely been addressed.
Group influence on identification would be
reflected in group members being more similar in
their degree of identification than individuals in
general (see Figure 1a). To examine this group
influence on in-group identification, withingroup and between-group differences should be
considered in conjunction with each other. That
is, it requires a multilevel approach in which individuals are nested within groups (see Arrow,
McGrath, & Berdahl, 2000; Bliese, 2000).
However, up until now, little social psychological
research has examined identification with ingroups at both the individual level and the group
level (see van Dick, van Knippenberg, Hägele,
Guillaume, & Brodbeck, 2008, for an exception).
In a multilevel approach, in-group identification at the group level refers to the part of individuals’ identification that is influenced by the
particular group to which the individual belongs.
In contrast, in-group identification at the individual level refers to the part of individuals’
identification that is solely based on the individual and his or her personal representation of the
group (for general discussions, see Bliese, 2000;
Kenny, Kashy, & Bolger, 1998). Thus at the individual level, identification is independent of the
group and of other group members, and reflects
group members’ idiosyncratic representations
of their in-group. For example, research shows
that individuals’ in-group identification can be
partly anchored in personal self-perceptions
(Otten & Epstude, 2006; van Veelen, Otten, &
Hansen, 2011).
Without explicit attention to the group level
of analysis in a multilevel approach, the degree of
individuals’ identification based in the group
itself cannot be estimated. Neither can it be distinguished from the more idiosyncratic influence
of the individual on his or her identification.
The Emergence of Group
Influence
Individuals can identify with a group, even if they
are not yet a member (Amiot, De la Sablonnière,
Terry, & Smith, 2007). When the group is
unknown, individuals are likely to use their personal characteristics as a reference point for their
identification (van Veelen, Otten, & Hansen,
2013). Similarly, upon entering a group, new
group members’ identification is predominantly
based on idiosyncratic representations of the
group (van Veelen, Hansen, & Otten, 2013). In
other words, without group knowledge (e.g.,
about the content of group identity, norms, and
values) the group is unlikely to influence individuals’ identification. Indeed, research suggests that
over time, through a process of socialization, is
new members’ identification increasingly
informed by characteristics of the group (Amiot
et al., 2007; van Veelen, Hansen, et al., 2013).
Thus we predict that in newly formed groups (in
which identity or norms do not yet exist), the
emergence of group influence on individuals’
identification can progress in a similar manner.
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
192
Individual-level variance
Constant
Group-level variance
In-Group Identification
b.
In-group identification
Constant
a.
Decreasing
Amount of group interaction
In-Group Identification
d.
In-group identification
Increasing
c.
Amount of group interaction
Amount of group interaction
Amount of group interaction
Figure 1a–d. The emergence of two group-level properties of in-group identification with greater group
interaction: decreasing individual-level variance and increasing group-level variance. Group 1 is represented
by squares, and Group 2 is represented by circles. Group means are represented by x’s. Left brackets indicate
group-level variance whereas right brackets indicate individual-level variance.
At the start of group formation, group members may not have any (shared) knowledge about
the group, and group identification is most likely
an individual-level property, reflecting group
members’ idiosyncratic representations of their
in-group. Interaction is an important factor
through which shared knowledge of the group
can emerge (see Kerr, Aronoff, & Messé, 2000;
Postmes, Haslam, & Swaab, 2005), as interaction
allows members to elaborate on the shared experience of group membership (Arrow et al., 2000;
van Knippenberg, De Dreu, & Homan, 2004).
Thus over time, with increased interaction, the
group may increasingly become a shared reference point that influences group members’ identification. In other words, we expect that in-group
identification at the group level can emerge with
increased interaction.
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
193
The emergence of group influence on ingroup identification through interaction can take
two different, though related, forms. On the one
hand, group influence can be the result of “consensualization” (Haslam et al., 1997). Research
suggests that through interaction, group members come to consensualize on aspects of their
identity, such as self-stereotypes, norms, and values (Haslam et al., 1997; Meeussen, Delvaux, &
Phalet, 2013; Sherif, 1935; Smith & Postmes,
2009; Thomas & McGarty, 2009). Through interaction, group members thus come to increasingly
agree on what the group and group membership
is like. Therefore, group members’ identification
may become increasingly alike (see Figure 1b).
On the other hand, group influence can be the
result of “polarization.” Research suggests that
through interaction, groups increasingly polarize
on aspects such as opinions and attitudes (see
Moscovici & Zavalloni, 1969; Turner, 1991, for a
review). Thus with increased interaction, groups
can become increasingly different from other
groups. For example, some groups will be
friendly, efficient, and enjoyable whereas other
groups will be unfriendly, inefficient, and joyless
(Postmes, Spears, & Lea, 2000). These betweengroup differences can influence the extent to
which a group invites members to identify. Over
time, groups may become increasingly different
in their degree of identification (see Figure 1c), as
some groups may encourage higher identification
than other groups.
In sum, we expect group influence on ingroup identification to emerge as a function of
interaction. This might occur, because group members become more similar to each other in their
degree of identification, because groups become
more different from each other in their degree of
identification, or because these processes cooccur (see Figure 1d).
Overview
To examine the multilevel nature of identification
and the role of interaction in its emergence, we
use multilevel modeling in a secondary analysis
of several small-group studies.1 First we compare
two very similar experiments (Studies 1a and 1b)
that varied mainly in the amount of interaction
within the group. Second, in Study 2, we explore
the multilevel nature of identification longitudinally by examining students’ identification with a
work group at three different points in time.
Finally, in Study 3 we explore the multilevel nature
of identification in experimentally created online
groups longitudinally, after 1 and 2 weeks of
online posting.
In all these studies, identification was measured with Leach et al.’s (2008) well-validated
two-dimensional model of in-group identification that distinguishes self-definition from
self-investment. In this model, self-definition
refers to the extent to which individuals perceive themselves and other in-group members
as a category (Turner, Hogg, Oakes, Reicher, &
Wetherell, 1987), and is reflected in individuals’
self-stereotyping as a typical group member and
perceiving the in-group as homogenous. Selfinvestment refers to individuals’ psychological
investment in the in-group (Tajfel, 1978), and is
reflected in a sense of solidarity with group
members and feeling satisfied about one’s group
membership.2 The present studies are the first
to examine this model in small groups and at
both the individual and the group level, simultaneously. We will explore the emergence of
group influence on both dimensions as a function of interaction.
Study 1
In Study 1, we compare two very similar experiments that varied in the amount of interaction. In
both studies, groups were formed solely for the
purpose of the experiment. In Study 1a, participants interacted face-to-face for 10 minutes. In
Study 1b, the amount of face-to-face interaction
was twice that in Study 1a.3
Method
Study 1a. In Study 1a, university students (134
women, 37 men) were randomly assigned to 57
different three-person groups. Participants were
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
194
seated around a table and interacted in a group
task in which they had to make a team shirt (see
Jans, Postmes, & van der Zee, 2012). They had
approximately 10 minutes to finish the task.
Afterwards, participants completed questionnaires, measuring their identification with the
group among others. Eleven items were adapted
from Leach et al. (2008) to assess the four components of in-group identification relevant to newly
formed groups: individual self-stereotyping, ingroup homogeneity, solidarity, and satisfaction.
For example, an individual self-stereotyping item
stated “I am similar to the average member of this
group,” and an in-group homogeneity item stated
“Members of this group have a lot in common
with each other.” A satisfaction item stated “I am
glad to be in this group,” and a solidarity item
stated: “I feel solidarity with this group.”
Responses were given on 7-point scales ranging
from 1 (fully disagree) to 7 (fully agree).
Study 1b. Study 1b was very similar in design to
Study 1a, but allowed for twice as much interaction. University students (115 women, 41 men)
were randomly assigned to 39 different four-person groups. They were seated around a table, and
interacted in two group tasks. The first task was
the same 10-minute task used in Study 1a. In a
second task, participants interacted an additional
10 minutes to build a house of Lego blocks.
Afterwards, participants completed questionnaires measuring identification with their group
among others. Responses were given on 7-point
scales (1 = fully disagree; 7 = fully agree).
Analyses. MPlus (Muthén & Muthén, 1998–
2010) was used to perform a multigroup
(between-studies) multilevel confirmatory factor
analysis (CFA) that specified the two-dimensional model of in-group identification at both
the individual (Level 1) and the group level (Level
2). In the model, the individual self-stereotyping
and in-group homogeneity components loaded
on a self-definition factor, and the solidarity and
satisfaction components loaded on a self-investment factor. We used scale scores for each of the
components, to minimize the number of
estimated parameters at the group level, relative
to N groups. The items at the group level are
represented by ovals in Figure 2a, because they
are the latent variables (i.e., group intercepts) and
not the measured variables. Additionally, the
model was specified in two ways that ensured
measurement equivalence across studies and
across the individual and the group levels (see
Figure 2a) and reduced the model’s complexity.
First, all item loadings for all factors were fixed
to 1 at both levels. In this way we forced every
item to have an equivalent loading on its expected
factor. Second, the item error variances were also
fixed to zero at the group level. We tested these
two sets of constraints in all our models and the
fit did not worsen with these constraints.
Multilevel CFA allows us to assess the extent
to which groups influence the two dimensions of
in-group identification, within each study. In multilevel CFA, group influence is shown in the intraclass correlation (ICC)—the proportion of the
variance in individuals’ responses attributable to
group membership. The ICC ranges from 0 to 1.
Thus a higher ICC for self-definition (or selfinvestment) shows that individuals within groups
are more similar in their degree of self-definition
(or self-investment) than individuals in general
(Bliese, 2000; Kenny & LaVoie, 1984). If the ICC
were 0, individuals within groups would be no
more similar in their self-definition (or selfinvestment) than individuals in general. If the
ICC were 1, individuals within groups would be
complete replicates of one another in their degree
of self-definition (or self-investment). To see
whether the ICC differed between studies, we calculated 95% confidence intervals as suggested by
McGraw and Wong (1996). Group influence on
self-definition and self-investment can be examined even more precisely with the two statistics
 τ 00 
used to calculate the ICC 
2 
 : individual τ 00 +σ 
level variance (σ2) and group-level variance (τ00)
(see Bliese, 2000).
We conducted our analyses in two steps. As a
baseline, Step 1 estimated a model that did not
constrain individual- and group-level variances of
self-definition and self-investment to be equal
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
195
1
e1
In-group
homogeneity
1
1
1
e2
1
e3
Individual
self-stereotyping
Satisfaction
1
1
1
e4
Self-definition
Self-investment
Solidarity
Individual level
Group level
0
1
e5
In-group
homogeneity
1
0
1
e6
1
Self-definition
Individual
self-stereotyping
0
1
e7
Satisfaction
1
0
e8
1
1
Self-investment
Solidarity
Figure 2a. Two-dimensional model of in-group identification.
across studies. This model provides the same
results as when we would have run the specified
model in each study separately. In Step 2, we provide a formal test of the differences in variance
across studies. We statistically assess whether individual-level variance of self-investment and selfdefinition decreased, and group-level variance of
self-investment and self-definition increased over
studies. If a factor variance can be constrained to
be equal across studies with no significant decrease
in model fit, it shows that this constrained model is
superior to the baseline model. Because a model
that constrains a factor variance to be equal across
studies is more parsimonious, it should be preferred to the less parsimonious baseline model. In
other words, an equally good fitting constrained
model indicates that there is little difference in the
factor variance across studies.
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
196
1
e1
In-group
homogeneity
1
1
e2
Individual
self-stereotyping
1
1
1
e3
1
e4
Satisfaction
In-group
identification
1
Solidarity
Individual level
Group level
0
1
e5
In-group
homogeneity
1
0
1
e6
Individual
self-stereotyping
1
0
1
e7
1
Satisfaction
In-group
identification
1
0
e8
1
Solidarity
Figure 2b. One-dimensional model of in-group identification.
Results
The baseline model specified the two dimensions
of in-group identification at both the individual
and the group level, in both studies. This baseline
model fits the data quite well, χ2(20) = 42.31, p =
.003, CFI = .967, RMSEA = .083, AIC = 3239.23
(Hu & Bentler, 1999), and fits the data better than
a comparable model that specified only a single
factor of general identification (see Figure 2b),
Δχ2(8) = 140.91, p < .001, AIC = 3364.14. We
therefore look at the results for self-definition
and self-investment, separately.
Group influence. In Study 1a, after 10 minutes of
interaction, the group explained 12% of the total
variance in self-definition, and 26% of the total
variance in self-investment. Both confidence
intervals did not include zero (see Figure 3a–b).
Thus 10 minutes of face-to-face interaction
appeared to be enough to enable the group to
influence individuals’ degree of self-investment
in the group and self-definition by the group.
Study 1b allowed for twice as much interaction as
Study 1a, and this seemed to be reflected in the
higher ICC (see Table 1). In Study 1b, the group
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
197
1.0
1.0
0.9
0.9
0.8
0.8
ICC for self-investment
ICC for self-definition
Jans et al.
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Study 1a
0.0
Study 1b
Study 1a
Study 1b
Figure 3a–b. ICC and 95% confidence intervals for self-definition and self-investment in Study 1.
Table 1. Group influence statistics and grand means for in-group identification in Study 1.
Dimensions
Study
ICC
1a
1b
1a
1b
.115a
Self-definition
Self-investment
.172a
.257a
.404b
Variance
Individual-level
Group-level
.757 (.120)***
.954 (.151)***
.471 (.082)***
.552 (.087)***
.098 (.084)
.198 (.115)*
.163 (.073)*
.374 (.124)**
M
SD
3.81a
4.18b
4.37a
4.92b
0.99
1.16
0.88
1.01
Note. Responses were given on a 1 (totally disagree) to 7 (totally agree) scale. The means and ICCs with different superscripts for
each dimension are significantly different, p (two-sided) < .05. The standard errors for the variances are reported between
brackets. A z-test was used to test whether variances were significantly higher than 0 (one-tailed, as suggested by Snijders &
Bosker, 1999, p. 90).
†p < .100; *p < .050; **p < .010; ***p < .001.
explained 17% of the variance in self-definition,
and a notable 40% of the variance in self-investment. This is quite high for groups of such small
size (Bliese, 2000).
Inspection of the 95% confidence intervals
around the ICC in the two studies (Figure 3a–b),
suggests that only the ICC for self-investment
differed significantly between studies (Cumming
& Finch, 2005). With the increased interaction
from Study 1a to Study 1b, the influence of the
group on self-investment increased.
Examination of the variance of each dimension
of in-group identification at the individual and the
group level, offers a more fine-grained analysis than
the ICC. It allows us to examine whether increased
group influence on self-investment is the result of
group members becoming more similar in their
self-investment, and/or groups becoming more
different in their self-investment.
Within-group similarity. In both studies, the individual-level variance of self-investment and selfdefinition were significant (see Table 1).
Although the group influenced group-members’
identification, individual differences in identification remained. Constraining the individuallevel variance of self-investment across studies
did not significantly reduce model fit, Δχ2(1) =
0.46, p = .248, AIC = 3237.70. Neither did constraining the individual-level variance of
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
198
self-definition, Δχ2(1) = 1.05, p = .152, AIC =
3238.29.4 Thus, group members were not more
similar in their identification with the group in
the study with more interaction, than in the
study with less interaction. This suggests that
the stronger group influence on self-investment
in Study 1b compared to Study 1a was not the
result of consensualization.
Between-group difference. Group influence on individuals’ identification can also be reflected in
groups being different from each other in identification. Although in Study 1a the ICC for selfdefinition and self-investment were significantly
different from zero, groups did not significantly
differ in self-definition (see Table 1). It might be
that the small size of the three-person groups
hampered statistical power at the group level
(see Kenny et al., 1998). In Study 1b, group variance was higher, and significant for both selfdefinition and self-investment (see Table 1).
Thus after 20 minutes of interaction, groups
differed from each other on both dimensions of
identification.
A formal test of the differences in group-level
variance between studies showed that groups did
not differ more from each other in their degree
of self-definition with more interaction: The variance of self-definition could be constrained at
the group level across both studies without
worsening fit, Δχ2(1) = 0.52, p = .236, AIC =
3237.75. However, in line with the differences in
ICC for self-investment, groups did differ more
on self-investment as a function of interaction.
Model fit declined when the group-level variance
of self-investment was constrained to be equal
across studies, Δχ2(1) = 2.51, p = .057, AIC =
3239.74. Thus group influence on self-investment increased from Study 1a to Study 1b,
because groups polarized in their degree of selfinvestment with more interaction.
Discussion
The results of Studies 1a and 1b confirm that
identification is more than an individual difference. The group to which individuals belonged
influenced their self-definition to some extent
and quite strongly influenced their self-investment. Furthermore, results supported our prediction that group influence on group members’
identification can increase with group interaction. However, this was only the case for selfinvestment. Self-investment was more strongly
influenced by the group in the study with 20 minutes of interaction than in the study with 10 minutes of interaction. In particular, it seemed that
with increased interaction, groups differed more in
the extent to which they promoted investment of
group members (although the overall degree of
identification also went up; see Table 1). This suggests that group polarization might be the underlying process for the increased group influence. Thus
Study 1 provides initial support for the idea that
interaction plays a key role in the emergence of
group-level identification across two similar studies
that varied in the amount of interaction. But a limitation of Study 1 is that these studies differed on
more factors besides interaction. In order to
remove this confound, Study 2 examines the role
of interaction on the multilevel nature of identification in a longitudinal study.
Study 2
In Study 2 we examine group influence on ingroup identification longitudinally in real-life
groups. Student project groups were followed
over the course of a term, and their identification
was measured at three different time points.
Method
In total 280 psychology students (Nwomen = 203,
Nmen = 63, Nunknown = 14) of a second-year
research course participated. They were divided
into 62 project groups, ranging from three to nine
members on the basis of their personal preferences (43 groups with four members; Mgroup size =
4.52, SDgroup size = 1.24). Over a period of 3
months, these groups had at least eleven group
meetings with their supervisor but probably met
up more often than that because students worked
together on group assignments. The project ended
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
199
with a large plenary session. We measured identification at three time points. T1 took place at the
beginning of the third group meeting at which
point students had not yet worked on group
assignments. T2 took place at the beginning of
the fifth meeting, when students had worked on
one group assignment. Finally, T3 took place at
the end of the plenary session. At this point,
group members had worked on six group assignments together. Five participants quit the course
before T1, two after T1, and again five after T2. In
total, there were 268 students left at T3.
To examine the effect of interaction on the
multilevel nature of in-group identification, we
tested the models described in Study 1. However,
instead of studies, we used time points as groups
in the multigroup multilevel CFAs.5
Results
The two-dimensional model fits the data very
well, χ2(30) = 46.10, p = .030, CFI = .990,
RMSEA = .046, AIC = 6878.17 (Hu & Bentler,
1999), and better than a single factor model of
general identification (see Figure 2b), Δχ2(12) =
518.29, p < .001, AIC = 7372.46.
Group influence. At the beginning of the third group
meeting, the percentage of variance explained by
the group was 21% for self-definition, and 29%
for self-investment. Thus the group already influenced identification at this early stage. Two weeks
later, the percentage of variance in self-definition
and self-investment changed a little (see Table 2).
At the end, after approximately 11 group meetings, the group explained 27% of the variance in
self-definition and 39% of the variance in
self-investment.
To examine whether self-definition and selfinvestment differed across the three time
points, we examined the 95% confidence intervals around the ICC (see Figure 4a–b; Cumming
& Finch, 2005). The confidence intervals suggest that the ICC for self-definition only
increased significantly from T2 to T3, while
there was no difference between T1 and T2
or T1 and T3. The increase of ICC for self-
investment does clearly seem to reflect a linear
effect of interaction. Group influence on selfinvestment at T3 was significantly higher than
group influence at T2 and T1. These patterns
may also be reflected in changes in individualand group-level variance.
Within-group similarity. There were significant
individual differences on both dimensions of ingroup identification at all time points (see Table
2). Group members did not become more similar
in their self-investment in the group and selfdefinition by the group with their greater interaction over time. Constraining the individual-level
variance of self-definition across time points did
not significantly reduce model fit, Δχ2(2) = 3.15,
p = .103, AIC = 6877.32. The same was true
for constraining the individual-level variance of
self-investment across time points, Δχ2(2) = 1.52,
p = .234, AIC = 6975.69.
Between-group difference. The higher ICC from T1
to T3 seems to be reflected in groups becoming
increasingly different from each other in both
self-definition and self-investment (see Table 2).
However, the group-level variance of selfdefinition, Δχ2(2) = 0.78, p = .338, AIC = 6874.95,
and self-investment, Δχ2(2) = 1.34, p = .256, AIC
= 6875.51, could be constrained across time
points without worsening model fit. Thus there is
no evidence that group influence on identification increased over time because groups became
more polarized in their degree of identification.
Discussion
The results of Study 2 are largely in line with our
predictions and the results of Study 1.
The results show that identification is a twodimensional, two-level construct. Despite
individual differences in self-definition and selfinvestment, the group influenced both dimensions. The results suggest that with interaction,
group influence on both dimensions increases.
But similar to Study 1, the effect of interaction on
group influence appears to be stronger on the
self-investment dimension. At the final
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
200
Table 2. Group influence statistics and grand means for in-group identification in Study 2.
Dimensions
Time
ICC
Self-definition
Self-investment
1
2
3
1
2
3
.213ab
.159a
.268b
.292a
.320a
.391b
Variance
Individual-level
Group-level
.711 (.083)***
.935 (.109)***
.749 (.090)***
.375 (.050)***
.462 (.058)***
.389 (.050)***
.193 (.073)**
.177 (.083)*
.274 (.090)***
.155 (.051)***
.217 (.066)***
.250 (.066)***
M
SD
3.98a
3.99ab
4.17b
4.76a
4.83a
5.23b
1.01
1.11
1.08
0.80
0.88
0.88
1.0
1.0
0.9
0.9
0.8
0.8
ICC for self-investment
ICC for self-definition
Note. Responses were given on a 1 (totally disagree) to 7 (totally agree) scale. The means and ICCs with different superscripts for
each dimension are significantly different, p (two-sided) < .05. The standard errors for the variances are reported between
brackets. A z-test was used to test whether variances were significantly higher than 0 (one-tailed, as suggested by Snijders &
Bosker, 1999, p. 90).
† p < .100; *p < .050; **p < .010; ***p < .001.
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Time 1
Time 2
Time 3
0.0
Time 1
Time 2
Time 3
Figure 4a–b. ICC and 95% confidence intervals for self-definition and self-investment in Study 2.
measurement, groups explained approximately
25% of variance in individuals’ self-definition,
and a notable 40% in self-investment. This corresponds with the variances explained after 20 minutes of interaction in Study 1b. Furthermore, it
seems that the overall degree of identification
also increased (see Table 2). However, we did not
find evidence for the process through which the
group influenced identification. This may be
because the groups already had met two times
before the first measurement point. Furthermore,
because Study 2 focused on students’ identification with groups they were not randomly
assigned to, the initial effects at T1 may have
been affected by self-selection. Therefore, in the
final study we focus on online groups to which
members are randomly assigned. This allows us
to better explore the extent to which interaction
is necessary for group influence on identification
to emerge.
Study 3
Study 3 was designed to allow members of newly
formed experimentally created groups to engage in
indirect online interaction. Participants were randomly assigned to small online groups. The online
group operated solely through asynchronous
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
201
online communication (i.e., participants logged on
at different times). Over the course of 2 weeks,
individual members posted approximately three
messages to each other. This sort of computermediated communication involves only minimal
interaction, as individuals respond to each other’s
messages over varying lengths of time. However,
past research shows that this sort of online group
can be sufficient to produce a sense of groupness—an online community as it were (Postmes,
Spears, & Lea, 1998). Identification was measured
after 1 and 2 weeks of online interacting.
Method
First-year university students (N = 140) were randomly assigned to 28 different five-person online
discussion groups.6 Group members were
assigned a number and a color that represented
them online. Participants were asked to introduce
themselves to each other by posting at least three
messages in the first 4 days of the study. In these
posts, participants typically wrote about their
characteristics, such as where they lived, what
their hobbies were, and what their strengths were.
Typical posts were: “I am 18 years old, and I am
sharing a house with another student, I like to go
out and play hockey, I am studying psychology
and I would like to work with children”; “I am
really spontaneous and I like to please people”; “I
have a Facebook account, just as the other group
members. But in contrast to the others I am not
religious.” Five participants did not post any messages and thus their data were excluded from
analyses. The remaining participants (N = 135, 17
men, 117 women, 1 unknown) posted an average
of 3.13 messages (SD = 0.88, range = 1–5). The
average message consisted of 64 words.
After 4 days of acquaintance, participants
were sent an online questionnaire. This included
a first measure of identification with their online
group (T1). Responses were given on a 6-point
scale that ranged from 1 (fully disagree) to 6 (fully
agree). Students had 3 days to complete the questionnaire that was sent to them.
In the second week, participants were asked to
debate two topics of interest to students (being
completely drunk and obtaining one’s “propaedeutic”—first-year qualification). After 4 days of
discussing, students received another questionnaire which measured identification (T2).
Results
The two-dimensional model fits the data very
well, χ2(20) = 18.87, p = .530, CFI > .999,
RMSEA < .001, AIC = 2435.49, and better than
the comparable single-factor model, Δχ2(8) =
167.02, p < .001, AIC = 2586.51.
Group influence. After 4 days of exchanging messages online, already 18% of the total variance
in individuals’ self-definition was explained by
group membership, while only 2% of the variance in self-investment was explained by group
membership (see Table 3). The 95% confidence
interval showed that the ICC for self-investment
was statistically equivalent to zero (see Figure
5a–b). In other words, the group did not explain
any variance in self-investment yet. At T2, participants had been interacting with each other
online for 2 weeks. The ICC values in Table 3
show that with this increased interaction, the
ICC for self-definition, and self-investment
were higher. Twenty-nine percent of the total
variance in individuals’ self-definition is now
explained by their group membership, and 11%
of the variance in self-investment. Thus with
more interaction, the group seemed to have a
stronger influence on group members’ identification. Inspection of the 95% confidence
around the ICC values across both time points
(Figure 5a–b), suggests that ICC increased significantly from T1 to T2, for both self-definition
and self-investment. Thus, with increased online
interaction, group influence on individuals’
investment in the group and definition by the
group increased.
Within-group similarity. At the individual level,
both self-definition and self-investment had
statistically significant variance at both time
points (see Table 3). Thus individuals differed a
good deal in their degree of self-definition and
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
202
Table 3. Group influence statistics and grand means for in-group identification in Study 3.
Dimensions
ICC
Time
Self-definition
Self-investment
1
2
1
2
.182a
.286b
.022a
.105b
Variance
Individual-level
Group-level
.504 (.086)***
.322 (.067)***
.622 (.098)***
.546 (.092)***
.112 (.067)*
.129 (.062)*
.014 (.048)
.064 (.057)
M
SD
4.13a
4.13a
3.86a
3.91a
0.85
0.77
0.85
0.85
1.0
1.0
0.9
0.9
0.8
0.8
ICC for self-investment
ICC for self-definition
Note. Responses were given on a 1 (totally disagree) to 6 (totally agree) scale. The means and ICCs with different superscripts for
each dimension are significantly different, p (two-sided) < .05. The standard errors for the variances are reported between
brackets. A z-test was used to test whether variances were significantly higher than 0 (one-tailed, as suggested by Snijders &
Bosker, 1999, p. 90).
† p < .100; *p < .050; **p < .010; ***p < .001.
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Time 1
Time 2
0.0
Time 1
Time 2
Figure 5a–b. ICC and 95% confidence intervals for self-definition and self-investment in Study 3.
self-investment. Still, individual-level variance
at T2 seemed to decrease for both dimensions
of identification. A formal test of these differences showed that model fit decreased when
individual-level variance was constrained across
time points for self-definition, Δχ2(1) = 2.86,
p = .046, AIC = 2436.35, but not for selfinvestment, Δχ2(1) = 0.32, p = .286, AIC =
2433.81. Thus group members’ self-definition,
but not group members’ self-investment,
became more “consensualized” with more
online interaction.
Between-group difference. After 4 days of exchanging
messages online, there was only significant variance
between groups in their self-definition (see Table
3). Some groups had somewhat lower self-definition and some had somewhat higher self-definition
than others. But with only minimal interaction
within the groups, groups did not significantly differ in their degree of self-investment (recall that the
ICC was very low on this dimension). At T2,
group-level variance increased for both dimensions
of identification. However, group-level variance for
self-investment still did not reach significance (see
Table 3). Thus even after 2 weeks of online interaction, groups did not differ in self-investment.
Model fit was not reduced when the group-level
variance of self-definition, Δχ2(1) = 0.03, p = .429,
AIC = 2433.52, or self-investment, Δχ2(1) = 0.44,
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
203
p = .254, AIC = 2433.92, were constrained across
time points. Thus there is no evidence that group
influence on identification increased over time
because groups became more polarized in their
degree of identification.
Discussion
Study 3 provides more insight in the necessity
of interaction for group influence to emerge on
identification. After some minimal online interaction at T1, only self-definition appeared to be
affected by the group. In contrast, self-investment seemed to be no more than an individual
difference. More interaction seems necessary for
self-investment to be influenced by the group.
After more online interaction, the amount of
group influence on both dimensions increased.
Thus in online communities as in face-to-face
groups, the group influence on identification
strengthens as a function of interaction. This
supports the idea that psychological processes
of self-categorization and social identification
may operate in computer-mediated communication as in face-to-face interaction (Postmes et al.,
1998). At T2 in the study with on-line groups,
degree of variance in self-definition as in the
studies with face-to-face groups. However, the
explained variance by the group in self-investment was much lower here than in the face-toface groups examined in the preceding studies.
This suggests that group influence on selfinvestment may be more dependent on interaction (and maybe also more direct interaction)
than group influence on self-definition.
Interestingly, the underlying process for
increased group influence in the online communities of Study 3 appeared to be different from
the real-life groups in Study 1 and 2. Rather
than groups becoming more different in selfinvestment, group members became more similar
in their self-definition over time. It might be that
the underlying processes for emerging group
influence on group identification are different for
online groups than for face-to-face groups. This
finding resonates with research that shows that
social categorization may have a strong influence
on the formation of online groups and communities (e.g., Spears & Postmes, in press).
General Discussion
We examined the multilevel nature of identification by exploring the development of group
influence on in-group identification in newly
formed groups. The results of three studies
supported the suggestion that identification is
more than an individual difference (Ashmore et
al., 2004; Ellemers et al., 1999), as it is at least
partly based in the group. Thus in-group identification is a multilevel construct. At the same
time, the results of Study 3 suggest that in some
cases identification may be purely based on an
idiosyncratic representation of the group; with
only limited online interaction, the group had
no influence on individuals’ self-investment.
This supports research suggesting that when a
group is (still) unknown, group members use
themselves as an anchor to infer group identification (van Veelen, Hansen, et al., 2013; van
Veelen, Otten, et al., 2013).
The results of all three studies also support
our prediction that group influence on group
identification increases as a function of interaction. Particularly for self-investment, identification became increasingly a group-level property
as group members interacted with each other.
These results corroborate findings showing that
new members in preexisting groups first need to
get to know the group, before they can use the
group as a basis for their identification (van
Veelen, Hansen, et al., 2013). Our research suggests that also in newly formed (face-to-face and
online) groups, the group becomes an increasingly important source for group members’
degree of identification. Over time, with increased
interaction, group members can develop shared
knowledge about the group (see Kerr et al., 2000;
Postmes et al., 2005), and the experience of being
a group member (Arrow et al., 2000; van
Knippenberg et al., 2004). The interactive experiences within the group thus seem to create common reference points that inform group
members’ identification.
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
204
We did not have clear expectations about the
specific process underlying increased group influence on group identification. Group influence on
group identification can either be the result of a
process of “consensualisation” in which group
members come to agree on the characteristics
of the group and the experience of group membership, or it can be the result of a process of
“polarization” in which groups become increasingly different from each other in characteristics
and experiences. The results of the three studies
seem to suggest that both processes occur. In
Study 1 (and somewhat in Study 2), groups
became increasingly different in their degree of
self-investment as a function of face-to-face interaction. In contrast, in Study 3, with online communities, group members became increasingly
similar in their degree of self-definition. Although,
this latter pattern may be particular to online
interaction, it could also be that group influence
on the self-definition dimension emerges more
through the process of consensualization, while
group influence on self-investment emerges more
through the process of polarization.
In addition to finding support for our main
hypotheses, our results also corroborate and
extend the two-dimensional model of in-group
identification by Leach et al. (2008) by showing
that the two-dimensional model is applicable to
newly formed small groups, at both the individual
and the group level of analysis. Thus at the individual level, identification with small groups
seems to be two-dimensional, like identification
with large social categories such as sex, ethnicity,
nationality, or university. This suggests that the
psychological side of groups may be similar
despite notable differences in size and other features (see Jans, Postmes, & van der Zee, 2011;
Postmes, Baray, Haslam, Morton, & Swaab, 2006;
Turner et al., 1987). Identification with various
types of groups can be conceptualized by a
dimension that captures the extent to which individuals perceive themselves and other in-group
members as a category (i.e., self-definition), and a
dimension that captures the extent to which individuals’ are psychologically invested in the ingroup (i.e., self-investment). Moreover, the
differential influence of the group on individuals’
self-definition and self-investment provides further evidence of the validity of the distinction
between these two dimensions of identification
(see also Leach, Rodriguez Mosquera, Vliek, &
Hirt, 2010; Leach et al., 2008; Shepherd, Spears,
& Manstead, 2013).
Our results suggest that interaction leads to
stronger group influence on self-investment than
on self-definition. Self-investment seems to require
more interaction than self-definition to be influenced by the group. It seems that for the group to
influence self-definition there only needs to be a
group (group-being), while group influence on
self-investment requires actual interaction (groupdoing). This distinction between group-being and
group-doing echoes a familiar distinction between
categorical and dynamic groups (Deaux & Martin,
2003; Postmes et al., 2005; Prentice, Miller, &
Lightdale, 1994; Wilder & Simon, 1998). However,
rather than characterizing groups as either categorical or dynamic, our findings suggest that most
groups tend to be and do at the same time.
Individuals’ self-definition is shaped by their
perceptions of the group and themselves as
being a category (Leach et al., 2008). This selfdefinition reflects the extent to which group
members have a “depersonalized” perception of
group members (in-group homogeneity) and
themselves (self-stereotyping; see Turner et al.,
1987). From the moment a group comes into
being, it can serve as a salient category (because it
is especially homogeneous or distinctive; Arrow
et al., 2000; Lickel et al., 2000), and thus serve as
a common reference from which group members
can deduce their self-definition (Leach & Vliek,
2008; Postmes et al., 2005). Indeed, there are circumstances where group-being alone suffices to
produce a sense of self-definition, as evidenced
in research showing that categorization of a zerohistory “minimal group” against an out-group
can lead group members to stereotype themselves
in group terms (Cadinu & Rothbart, 1996).
In contrast, self-investment is anchored in a
different set of socialpsychological processes
than self-definition. On the basis of the present
research it appears that common experiences as a
result of extended interaction are necessary for
the group to influence group members’
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
205
self-investment in the group. Future research can
begin to explore these differential effects of
group-doing and group-being on the emergence
of group-level identification.
The importance of multilevel approaches. Individuals’
in-group identification seems to be influenced by
the particular group as well as by the particular
individual. Inattention to group-level effects is
likely to lead to an overestimation of individuallevel effects (for discussions, see Bliese, 2000;
Kenny et al., 1998). This may inadvertently reinforce the view that in-group identification is
solely an individual difference. On the other
hand, focusing only on measurement at the group
level ignores the potentially important variance
across individuals that may explain their experience and behavior. Thus whenever social psychological phenomena may operate at multiple levels
of analysis it is important to examine these phenomena with a multilevel approach.
In addition, when in-group identification is
considered as a predictor of group- and individual-level outcomes, identification at the group
level could also be examined. For instance, the
question of which groups are most likely to
engage in collective action is one that may be best
answered with attention to between-group differences in in-group identification. Although strong
in-group identification at the individual level
increases individuals’ willingness for collective
action (for a review, see van Zomeren, Postmes,
& Spears, 2008), strong in-group identification at
the group level might help to turn this willingness
into specific action. Being in a group that is highly
identified may provide the sort of social support
required for complicated or dangerous collective
action (see van Zomeren, Spears, Fischer, &
Leach, 2004). Distinguishing individual-level
from group-level variance as we have done here
allows for a more precise examination of the
group-level processes at work.
Limitations and Future Directions
Multilevel analyses of individuals in groups are
difficult for a number of practical reasons (e.g.,
finding enough participants). In general, multilevel analyses of the sort we performed here tend
to be low in statistical power (Kenny et al., 1998).
It seems important to also explore the grouplevel properties of in-group identification with a
large number of large-scale groups typically
examined in social psychological studies of ingroup identification (e.g., ethnicity, nationality,
university). This may help us examine whether
previously found effects of for example group
status on identification, are also reflected in
equivalent degrees of identification with a wide
variety of groups with similar status.
Our examination of small groups may not
parallel exactly the emergence of group-level
properties of identification with preexisting
social categories, such as nationality. In these
large groups, it is impossible to interact directly
with all members of the group. Still, new members’ identification may become increasingly a
function of group characteristics as a result of
socialization (see van Veelen, Hansen, et al.,
2013). Thus direct interaction with the entire
group cannot be the only means through which
members achieve shared group experiences. It is
likely that modes of indirect or symbolic interaction have a stronger influence on the emergence of group-level self-investment in large
social categories. For example, entrepreneurs of
identity, the media, history, and intergroup
events (such as important sporting events or
war) can provide shared knowledge of the group
and shared experiences of group membership
(see also Postmes et al., 2006). Furthermore,
since larger groups can often be subdivided into
smaller groups, interaction within these subgroups might be high, while interaction between
subgroups (such as social classes or age groups)
is low. Therefore, degrees of self-investment
may be similar within subgroups but different
between subgroups.
Conclusion
Although group identification is often treated as a
characteristic that varies across individuals, conceptually the notion of “social identity” implies
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
206
that identification is not merely an individual-level
phenomenon. Groups can influence members’
degree of self-definition by the group and selfinvestment in the group. Where in-group identification is investigated as having both individual- and
group-level properties, differences between
groups and similarities within groups can be
examined simultaneously. Both of these grouplevel effects have seldom been studied by social
psychologists. A multilevel approach such as that
used in the present research enables a better
examination of the uniquely social psychological
interplay of the individual and the group as distinct and yet interlinked phenomena. Using this
approach, we can conclude that group processes
have predictable and strong effects on the emergence of group-level patterns of identification.
However, researchers interested in the relation
between individuals and their group should realize
that at times identification might have little to do
with shared group membership.
2.
3.
Acknowledgements
We thank David A. Kenny for his guidance and his
comments on a previous draft.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit
sectors.
Notes
1. The between-condition differences in Study
1b of the present paper are reported in detail
in another paper (see Jans et al., 2012). Studies
1a, 2, and 3 are currently unpublished data. The
primary purpose of Studies 1a, 1b, and 3 was to
examine the effects of the nature of the interaction on group outcomes and on mean levels
of self-investment. Important for the present
paper is that although the nature of the interaction may have varied across studies, the amount
of interaction was identical for each condition.
Important for the present analyses, too, is that we
carried out multigroup multilevel CFAs to check
if differences between condition would qualify
the results that we focus on in the present paper.
These analyses showed that differences between
4.
5.
6.
conditions were independent of the effects
reported here (see endnotes for each study for
more details).
The hierarchical model of in-group identification by Leach et al. (2008) specifies a third component of self-investment: The extent to which
the group is a central aspect of someone’s selfconcept. But the items measuring this component were excluded from the studies reported in
this paper. The reason was that we expected a
floor effect on measures of centrality in newly
formed groups. To illustrate why, consider
that members of a newly formed group that
had interacted for 10 or 20 minutes would be
asked questions such as: “This group is central
to who I am.” When designing the research we
thought that such feelings would be so unlikely
that responses to the question would not be
meaningful and interpretable. We decided not to
include this dimension.
In both Studies 1a and 1b, groups were randomly assigned to a 2 (diversity: homogenous
vs. heterogeneous personalities) x 2 (social identity formation: inductive vs. deductive) design.
Participants were told that group members had
either very similar or very different personalities. In deductive group formation, members
were given a team logo whereas in the inductive
group formation each member contributed to
the design of a logo. A comparison between the
baseline model and the fully constrained model
did not provide any evidence for differences in
variance across conditions in Study 1a, Δχ2(18)
= 19.43, p = .366, and in Study 1b, Δχ2(18) =
13.21, p = .779.
When the chi-square difference test was used to
test a variance component, we divided the p-value
for the chi-square difference test by 2, as recommended by Berkhof and Snijders (2001).
Instead of treating timepoint as a group, we could
have set it up as a latent multilevel growth curve
model. That would allow for estimating the slope
in group-level identification over time. But what
we are interested in—changes in group variance
overtime—would still need to be assessed by fixing the variances to be equal.
Groups were randomly assigned to a common
characteristics versus individual characteristics
condition, in which they were asked to discuss
either the commonalties or differences of group
members. Comparisons between multigroup
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
207
multilevel CFAs did not provide any evidence
for differences in variances across conditions. A
model in which we constrained all variances and
covariances to be equal across conditions did not
significantly decrease model fit in comparison to
a baseline model, Δχ2(6) = 7.17, p = .306.
References
Amiot, C. E., De la Sablonnière, R., Terry, D. J., &
Smith, J. R. (2007). Integration of social identities in the self: Toward a cognitive-developmental
model. Personality and Social Psychology Review, 1,
364–388. doi:10.1177/1088868307304091
Arrow, H., McGrath, J. E., & Berdahl, J. L. (2000).
Small groups as complex systems: Formation, coordination, development, and adaptation. Thousand Oaks,
CA: Sage.
Ashmore, R. D., Deaux, K., & McLaughlin-Volpe, T.
(2004). An organizing framework for collective
identity: Articulation and significance of multidimensionality. Psychological Bulletin, 130, 80–114.
doi:10.1037/0033–2909.130.1.80.
Berkhof, J., & Snijders, T. A. (2001). Variance component testing in multilevel models. Journal of
Educational and Behavioral Statistics, 26, 133–152.
doi:10.3102/10769986026002133
Bliese, P. D. (2000). Within-group agreement, nonindependence, and reliability: Implications for
data aggregation and analysis. In. K. J. Klein &
S. W. J. Kozlowski (Eds.), Multilevel theory, research,
and methods in organizations: Foundations, extensions,
and new directions (pp. 349–381). San Francisco,
CA: Jossey-Bass.
Brewer, M. B. (1991). The social self: On being the
same and different at the same time. Personality and Social Psychology Bulletin, 17, 475–482.
doi:10.1177/0146167291175001.
Brewer, M. B., Manzi, J. M., & Shaw, J. S. (1993). Ingroup identification as a function of depersonalization, distinctiveness, and status. Psychological
Science, 4, 88–92. doi:10.1111/j.1467–9280.1993.
tb00466.x
Cadinu, M. R., & Rothbart, M. (1996). Self-anchoring and differentiation processes in the minimal group setting. Journal of Personality and Social
Psychology, 70(4), 661–677. doi:10.1037/0022–
3514.70.4.661
Cumming, G., & Finch, S. (2005). Inference by eye:
Confidence intervals and how to read pictures of data. American Psychologist, 60, 170–180.
doi:10.1037/0003–066X.60.2.170
Deaux, K., & Martin, D. (2003). Interpersonal networks and social categories: Specifying levels
of context in identity processes. Social Psychology
Quarterly, 66, 101–117. doi:10.2307/1519842
Ellemers, N., Spears, R., & Doosje, B. (1999). Social
identity: Context, commitment, content. Oxford, UK:
Wiley-Blackwell.
Ellemers, N., van Knippenberg, A., De Vries, N., &
Wilke, H. (1988). Social identification and permeability of group boundaries. European Journal
of Social Psychology, 18, 497–513. doi:10.1002/
ejsp.2420180604
Haslam, S. A., Turner, J. C., Oakes, P. J., McGarty,
C., & Reynolds, K. J. (1997). The group as
a basis for emergent stereotype consensus.
European Review of Social Psychology, 8, 203–239.
doi:10.1080/14792779643000128
Hu, L., & Bentler, P. M. (1999). Cutoff criteria
for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
doi:10.1080/10705519909540118
Jans, L., Postmes, T., & van der Zee, K. I. (2011). The
induction of shared identity: The positive role
of individual distinctiveness for groups. Personality and Social Psychology Bulletin, 37, 1130–1141.
doi:10.1177/0146167211407342
Jans, L., Postmes, T., & van der Zee, K. I. (2012).
Sharing differences: The inductive route to
social identity formation. Journal of Experimental
Social Psychology, 48, 1145–1149. doi:10.1016/j.
jesp.2012.04.013
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data
analysis in social psychology. In D. Gilbert, S. T.
Fiske & G. Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 233–265). Boston, MA:
McGraw Hill.
Kenny, D. A., & LaVoie, L. (1984). The social relations
model. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 18, pp. 141–182). New
York, NY: Academic Press.
Kerr, N. L., Aronoff, J., & Messé, L. A. (2000). Methods of small group research. In H. T. Reiss & C.
M. Judd (Eds.), Handbook of research methods in social
and personality psychology (pp. 160–189). Cambridge,
UK: Cambridge University.
Leach, C. W., Rodriguez Mosquera, P. M. R., Vliek, M.
L. W., & Hirt, E. (2010). Group devaluation and
group identification. Journal of Social Issues, 66(3),
535–552. doi:0.1111/j.1540–4560.2010.01661.x
Leach, C. W., van Zomeren, M., Zebel, S., Vliek, M.
L. W., Pennekamp, S. F., Doosje, B., & …
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Group Processes & Intergroup Relations 18(2)
208
Spears, R. (2008). Group-level self-definition
and self-investment: A hierarchical (multicomponent) model of in-group identification.
Journal of Personality and Social Psychology, 95(1),
144–165. doi:10.1037/0022–3514.95.1.144
Leach, C. W., & Vliek, M. L. W. (2008). Group membership as a “frame of reference” for interpersonal
comparison. Social Psychology and Personality Compass,
2, 539–554. doi:10.1111/j.1751–9004.2007.00058.x
Lickel, B., Hamilton, D. L., Wieczorkowska, G., Lewis,
A., Sherman, S. J., & Uhles, A. N. (2000). Varieties of groups and the perception of group entitativity. Journal of Personality and Social Psychology,
78(2), 223–246. doi:10.1037/0022–3514.78.2.223
McGraw, K. O., & Wong, S. P. (1996). Forming
inferences about some intraclass correlation
coefficients. Psychological Methods, 1(1), 30–46.
doi:10.1037/1082–989X.1.1.30
Meeussen, L., Delvaux, E., & Phalet, K. (2013). Becoming a group: Value convergence and emergent
work group identities. British Journal of Social Psychology, 53(2), 235–248. doi:10.1111/bjso.12021
Moscovici, S., & Zavalloni, M. (1969). The group as
a polarizer of attitudes. Journal of Personality and
Social Psychology, 12(2), 125–135. doi:10.1037/
h0027568
Muthén, L. K., & Muthén, B. O. (1998–2010). Mplus
user’s guide (6th ed.). Los Angeles, CA: Muthén &
Muthén.
Otten, S., & Epstude, K. (2006). Overlapping mental
representations of self, ingroup and outgroup:
Unraveling self-stereotyping and self-anchoring.
Personality and Social Psychology Bulletin, 32, 957–969.
doi:10.1177/0146167206287254
Postmes, T., Baray, G., Haslam, S. A., Morton, T. A.,
& Swaab, R. I. (2006). The dynamics of personal
and social identity formation. In T. Postmes & J.
Jetten (Eds.), Individuality and the group: Advances in
social identity (pp. 215–236). Thousand Oaks, CA:
Sage.
Postmes, T., Haslam, S. A., & Swaab, R. I. (2005).
Social influence in small groups: An interactive model of social identity formation.
European Review of Social Psychology, 16, 1–42.
doi:10.1080/10463280440000062
Postmes, T., Spears, R., & Lea, M. (1998).
Breaching or building social boundaries?
SIDE-effects of computer-mediated communication. Communication Research, 25(6), 689–715.
doi:10.1177/009365098025006006
Postmes, T., Spears, R., & Lea, M. (2000). The formation of group norms in computer-mediated
communication. Human Communication Research,
26(3), 341–371. doi:10.1111/j.1468–2958.2000.
tb00761.x
Prentice, D. A., Miller, D. T., & Lightdale, J. R. (1994).
Asymmetries in attachments to groups and to
their members: Distinguishing between commonidentity and common-bond groups. Personality
and Social Psychology Bulletin, 20(5), 484–493.
doi:10.1177/0146167294205005
Shepherd, L., Spears, R., & Manstead, A. S. (2013).
“This will bring shame on our nation”: The role
of anticipated group-based emotions on collective action. Journal of Experimental Social Psychology,
49(1), 42–57, doi:10.1016/j.jesp.2012.07.011
Sherif, M. (1935). A study of some social factors in perception. Archives of Psychology, 187, 1–60.
Smith, L. G., & Postmes, T. (2009). Intra-group interaction and the development of norms which
promote inter-group hostility. European Journal
of Social Psychology, 39(1), 130–144. doi:10.1002/
ejsp.464
Snijders, T. A. B., & Bosker, R. (1999). An introduction
to basic and advanced modeling. London, UK: Sage.
Spears, R., & Postmes, T. (in press). Group identity,
social influence and collective action online:
Extensions and applications of the SIDE model.
In S. Sundar (Ed.), The handbook of psychology of communication technology. Oxford, UK: Blackwell.
Tajfel, H. (1978). Differentiation between social groups. London, UK: Academic Press.
Thomas, E. F., & McGarty, C. A. (2009). The role of
efficacy and moral outrage norms in creating the
potential for international development activism
through group-based interaction. British Journal of
Social Psychology, 48(1), 115–134. doi:10.1348/014
466608X313774
Turner, J. C. (1991). Social influence. Belmont, CA:
Thomson Brooks/Cole.
Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S.,
& Wetherell, M. S. (1987). Rediscovering the social
group: A self-categorization theory. Oxford, UK: Basil
Blackwell.
Van Dick, R., van Knippenberg, D., Hägele, S.,
Guillaume, Y. R., & Brodbeck, F. C. (2008). Group
diversity and group identification: The moderating
role of diversity beliefs. Human Relations, 61(10),
1463–1492. doi:10.1177/0018726708095711
Van Dick, R., Wagner, U., Stellmacher, J., & Christ,
O. (2005). Category salience and organizational identification. Journal of Occupational
and Organizational Psychology, 78(2), 273–285.
doi:10.1348/096317905X25779
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Jans et al.
209
Van Knippenberg, D., De Dreu, C. W., & Homan, A.
C. (2004). Work group diversity and group performance: An integrative model and research
agenda. Journal of Applied Psychology, 89(6), 1008–
1022. doi:10.1037/0021–9010.89.6.1008
Van Veelen, R., Hansen, N., & Otten, S. (2013).
Newcomers’ cognitive development of social
identification: A cross-sectional and longitudinal
analysis of self-anchoring and self-stereotyping.
British Journal of Social Psychology, 53(2), 281–298.
doi:10.1111/bjso.12038
Van Veelen, R., Otten, S., & Hansen, N. (2011).
Linking self and ingroup: Self-anchoring as distinctive cognitive route to social identification.
European Journal of Social Psychology, 41, 628–637.
doi:10.1002/ejsp.792
Van Veelen, R., Otten, S., & Hansen, N. (2013).
Social identification when an in-group identity is unclear: The role of self-anchoring
and self-stereotyping. British Journal of Social
Psychology, 52, 543–562. doi:10.1111/j.2044–
8309.2012.02110.x
Van Zomeren, M., Postmes, T., & Spears, R. (2008).
Toward an integrative social identity model of
collective action: A quantitative research synthesis of three socio-psychological perspectives. Psychological Bulletin, 134(4), 504–535.
doi:10.1037/0033–2909.134.4.504
Van Zomeren, M., Spears, R., Fischer, A. H., & Leach,
C. W. (2004). Put your money where your mouth
is! Explaining collective action tendencies through
group-based anger and group efficacy. Journal
of Personality and Social Psychology, 87(5), 649–664.
doi:10.1037/0022–3514.87.5.649
Wilder, D., & Simon, A. F. (1998). Categorical and
dynamic groups: Implications for social perception and intergroup behavior. In C. Sedikides, J.
Schopler & C. A. Insko (Eds.), Intergroup cognition
and intergroup behavior (pp. 27–44). Mahwah, NJ:
Lawrence Erlbaum Associates.
Downloaded from gpi.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016