1

Social comparison and social identity: Some prospects for

intergroup behaviour

JOHN C. TURNER

University of Bristol

A bstract

Recent studies have reported that the variable of social categorization per se is
sufficient for intergroup discrimination. This paper presents an explanation of
these findings in terms of the operation of social comparison processes between
groups based on the need for a positive ingroup identity. The relationship between
perceived social identity and intergroup comparison is elaborated theoretically,
and it is argued that social comparisons give rise to processes of mutual differentia-
tion between groups which can be analysed as a form of ‘social’ competition. Social
competition is distinguished from realistic competition (conflict of group interests).
New data is reported which strengthens this interpretation of the ‘minimal’
categorization studies. It is found that minimal intergroup discrimination takes
place in the distribution of meaningless ‘points‘ as well as monetary rewards and
that social categorization per se does not lead to intergroup behaviour where the
subjects can act directly in terms of ‘self‘. Other studies on intergroup biases are
reviewed to argue for the generality of social competition in intergroup situations.

1. Introduction

Recent studies (Tajfel, 1970; Tajfel et at., 1971) have explored the role played
by social categorization in intergroup behaviour. Tajfel et al. (1971) have drawn
the conclusion that under certain conditions the mere classification of subjects
into in- and outgroupers is a sufficient as well as necessary condition to induce
forms of ingroup favouritism and discrimination against the outgroup. The results
of these studies which have since been amply confirmed and refined in various
forms by further work (Doise et al., 1972; Doise and Sinclair, 1973; Tajfel and
Billig, 1974; BiKg and Tajfel, 1973; Turner 1972, 1973) point to a wide gap
in the applicability of existing theories of intergroup behaviour. In particular, the

Eur. J . SOC. Pvchol. 5(1), p p . 5-34.

6 John C.Turner

intergroup discrimination demonstrated under the exacting conditions of these

experiments seems to fall outside of the domain of either ‘rational’ or ‘irrational’
models of intergroup conflict in the wide classification of these models employed
by Coser (1956).
It should also be noted that, as the evidence has accumulated, it has become
more and not less difficult to explain the data in terms of relatively familiar
processes operating ‘incognito’ in the experimental situation (see the above studies).
Turner (1972, p. 4), in assessing the results, concluded that ‘there is no compelling
evidence that the controlling variable is similarity/dissimilarity of in- and out-
groupers, common fate or even anticipation of interaction, although doubtless
these variables have their independent effects’.
An experiment performed by Billig and Tajfel (1973) to separate the variables
of interindividual similarity and social categorization should be mentioned in
particular. It seems that even where classification into groups is on an openly
random basis (thus eliminating any perception of ingroup similarity by the
subjects), intergroup discrimination still occurs. Both social categorization and
similarity were manipulated as independent variables in this study. In the
categorization (C) conditions subjects were explicitly divided into groups, and in
the non-categorization conditions (c) there was no mention of ‘groups’. In the
similarity conditions (S) subjects were told that they were assigned code numbers
and/or group membership on the basis of previously expressed picture preferences.
In the non-similarity conditions (3) code numbers and/or group membership were
openly assigned by chance. Thus there were four conditions: CS,CS, C S and a.
The dependent variable was the distribution of monetary rewards between other
subjects referred to only by their code number and group affiliation if they had
one. Significant ingroup favouritism was found in both C conditions but in neither
of the c conditions. S subjects produced greater ingroup favouritism than the 3
subjects on only one of four measures and an overall measure. Thus whilst both
variables have some effect, that of categorization is clearly the stronger and cannot
be explained in terms of similarity.
In a later article Tajfel (1972) has presented a body of ideas linking social
categorization with social identity which seems capable of explaining much of the
data from experiments on ‘minimal groups’ and other related data from the area
of intergroup competition. This paper will discuss some of his and other ex-
periments in the light of these ideas. The aim is to develop Tajfel’s interpretation
in applying it to the more recent data and perhaps to develop some guidelines for
reconstructing what at the moment is an analytic approach more along the lines
of a systematic predictive theory. The paper’s theme is not to test but to clarify,
and both data and theory will be presented selectively to that end.
 Social comparison and social identity 7

2. A theoretical outline

2.1 Social comparison processes and social identity: A definition and some
assumptions
Tajfel’s linkage of social categorization and social comparison processes with the
theme of social identity will be briefly outlined:
A) Social identity will ‘refer to the individual’s knowledge that he beIongs to
certain social groups together with some emotional and value significance to him
of this group membership’ (p. 292), whilst social categorization is a ‘means of
systematizing and ordering the social environment particularly with regard to its
role as a guide for action, and as a reflection of social values’. It also provides
a ‘system of orientation which creates and defines the individual’s own place in
society’ (p. 293). An individual defines himself as well as others in terms of his
location within a system of social categories - specifically social group member-
ships - and social identity may be understood as his definition of his own position
within such a system.
B) ‘It can be assumed that an individual will tend to remain a member of a
group and seek membership of new groups if these groups have some contribution
[or can be made to have some contribution] to make to the positive aspects of his
sQcial identity, i.e., to those aspects of it from which he derives some satisfac-
tion. . .’ (p. 293). Tajfel goes on to say that if a group does not satisfy this re-
quirement the individual will tend to leave it. If leaving a group presents certain
difficulties then there may be reinterpretation of the group’s attributes so that
its unwelcome features (e.g., low status) are either justified or made acceptable
or the situation may be accepted for what it is with engagement in social action
which will lead to desirable changes in the situation.
C) ‘No group lives alone - all groups in society live in the midst of other
groups. In other words, the “positive aspects of social identity” . . . the re-
interpretation of attributes and the engagement in social action. . . only acquire
meaning in relation to, or in comparison with, other groups’ (p. 293-4). Thus
Tajfel notes: ‘The characteristics of one’s group as a whole (such as its status, its
richness or poverty, its skin colour or its ability to reach its aims) achieve most
of their significance in relation to perceived differences from other groups and the
value connotations of these differences. . . A group becomes a group in the sense
of being perceived as having common characteristics or common fate only because
1. Except where otherwise indicated, all
quotations in this section are from H. Tajfel
(1972).
8 fohn C.Turner

other groups are present in the environment’ (p. 295).

To summarize the argument so far, membership of a particular group in regard
to its function of social identity is related to a positive evaluation of its attributes
in comparison with other groups. It can be said that the important dimensions of
intergroup comparison from the standpoint of social identity are those associated
with values, most of which will be culturally derived.
D) ‘A social group will, therefore, be capable of preserving its contribution to
those aspects of an individual’s social identity which are positively valued by
him only if it manages to keep its positively valued distinctiveness from other
groups. . . social comparisons between groups are focussed on the establishment
of distinctiveness between one’s own and other groups’ (p 296).
This can be expressed in another way: An individual’s need for positively valued
identity requires that where an intergroup comparison can be made in terms of
a dimension whose poles have a clear value differential, then his own group must
differentiate itself relative to other groups on that dimension towards the positively
valued pole.
If it is the comparative perspective that links social categorization with positive
identity, it is as well to clarify briefly Tajfel’s extension of Festinger’s (1954)
theory of social comparison processes. The basic hypotheses of Festinger’s theory
relevant to this clarification are I, I1 and IV,respectively, that there exists in the
human organism a drive to evaluate his opinions and abilities; to the extent that
objective, non-social means are not available, people evaluate their opinions and
abilities by comparison with the opinions and abilities of others; and that there
is a unidirectional drive upward in the case of abilities which is largely absent in
opinions. The distinction between abilities and opinions in the last hypothesis
rests on the fact that ‘with respect to abilities, different performances have intrin-
sically different values in Western culture’ whereas ‘no opinion in and of itself
has any greater value than any other opinion’ (Festinger, 1954, pp. 124-125).
Firstly, with regard to hypothesis I, Tajfel (1972) argues that an individual’s
membership group is an important object of evaluation with regard to its function
of social identity. Tajfel is attempting a social psychological analysis of a simple
but unspoken theme in Festinger’s ideas - that the individual has a need to
evaluate himself. To Berger’s (1966) declaration that ‘the individual realises
himself in society’, the corollary is being added that the individual evaluates
himself in society.
Secondly, Tajfel argues that: ‘The criterion of “objectivity” cannot be based
on classifying phenomena as being of a rcsocial” or a “non-social” nature, with
the presumed attendant consequence that opinions about it can be tested respect-
ively by “social” or “non-social” means.’ Whilst it is true that ‘certainty can very
 Social comparison and social identity 9

often be more easily reached about the physical than about the social means of
testing. . . this is not a theoretical distinction between what appears and does not
appear as “objective reality” ’ (p. 294). The criterion of objectivity may be defined
instead in terms of the awareness that there exist alternatives to the judgment one
is making; thus ‘objectivity’ may be equally conferred through physical testing
or a high social consensus. The range of social comparison processes, therefore,
may be extended to include ‘both the social context (or significance) of “non-
social” testing and the cases where the high social consensus about the nature of
a phenomenon is sufficient to confer the mark of “objectivity” on opinions about
it’ (p. 295). Evaluation of the membership group is not ‘social’ by default - where
nonsocial, physical means of appraisal are unavailable. The positive or negative
value of the membership group is intrinsically related to social comparison just
as group membership itself is an element of social and not physical reality.
Likewise the objectivity of a particular evaluation is a matter of the degree of
social consensus that exists regarding it (at least amongst those sharing similar
values). There cannot be nonsocial, physical criteria for the ‘objeotive’ status
of social identity. As Tajfel puts it: ‘The only “reality” tests that matter with regard
to group characteristics are tests of social reality’ (p. 295).
Thirdly the important comparative dimensions for social identity parallel those
of abilities rather than opinions, i.e., they are value-laden. In this sense the
individual‘s need to evaluate himself in society is more correctly expressed as a
need to make a favourable or positive evaluation of himself in society. The im-
portance of this parallel will be brought out in considering the relation of social
identity to competition.

2.2 Comparison, competition and conflicts of interests

The relevance of these ideas to the phenomena of intergroup competition can now
be sketched; the drat3 will assume more substance as we look at the empirical
findings. Firstly a more general question will be dealt with. The point in arguing
that the acquisition of positive social identity is linked to social categorization
via intergroup comparison has been to develop a coherent account of what people
will do under theoretically specifiable conditions; but it could be said that the
account’s affective or motivational component, its ‘motor’, is somewhat tautologous.
People will seek a more positive social identity where ‘positive’ itself denotes no
more than certain shared beliefs as to what aspects of social identity are desirable
and sought. The answer must be that insofar as the value connotations associated
with a comparative dimension are shared, indeed generated, socially, they are
specifiable independently of and prior to the behaviour of any one group. (This
10 John C. Turner

is not to deny processes of reinterpretation and justification leading to value

conflict.) In the light of the sociocultural context of social behaviour the assertion
that people desire a positive rather than a negative evaluation of themselves is
a declaration of an empirical regularity, and the theoretical task is as much to
show that apparently dissimilar phenomena are aspects of the same regularity
as to find some ‘deeper’ reason why. It is not the intent of this discussion to
explain the nature of ‘value’; it is taken as a premise, a stepping stone, from
which hopefully sense can be made of other things.
The fact that the value differential associated with a social comparison is shared
by differing groups provides the link between social identity and competition. From
the earlier discussion it should be clear that positive social identity is inextricably
a matter of mutual comparisons between groups - a comparison where, for
instance, any two groups are attempting mutual but asymmetrical differentiation
from each other towards the positively valued pole of the relevant dimension.
It could be said that there is a process of competition for positive identity, for
each group’s actions are attempts not at some absolute degree of value but at
positively valued differentiation and thus are relative to the other group’s actions.
Mutual comparison compels continuous reciprocity in the standards each group
sets itself and ensures consequently a spiralling rivalry until some final ‘inequity’
is reached. This derivation of competition from comparison at the intergroup
level is similar to Festinger’s analysis of the unidirectional drive upwards in the
case of abilities at the interindividual level; in essence both rest on the asymmetry
built into comparisons associated with differential value. The above description
is couched in terms of the ‘pure’ case; its relevance to concrete experiments will
be touched upon later. For the moment it suffices to say that we have been led
to processes of competition that might be expected to operate in many intergroup
situations but do not require conditions defined by realistic conflict of interests
or rivalry for a material reward or goal. The processes of competition which follow
from the notion of social identity - hereafter called ‘social competition’ - arise
from the very nature of an intergroup situation where ‘comparable’ action is
possible, that is, associated with a shared value differential.
Positive social identity can, of course, be described as a goal or reward, and
social competition can be defined as the outcome of rivalry between groups for
this goal; but any causal similarity between social competition and, for instance,
Sherif‘s (1966) ‘conflict of interests’ theory implied by the use of such terminology
would be misleading. Sherif‘s theory has as its basis the postulate of self-interest,
not self-evaluation. His model is of the group or individual with specific goals or
interests which exist in principle quite independently of an intergroup situation
and where intergroup competition is a contingent expression of a particular inter-
 Social comparison and socid identity I1

relationship of these goals, a negative interdependence. The relations between

groups are a function of the relations between group goals and may range from
conflict through cooperation to complete fusion, i.e., the dissolution of the inter-
group situation. Thus Sherif’s theory can as easily explain cooperation as competi-
tion - in terms of the positive as opposed to negative interdependence of group
interests - whereas social competition theory indicates a unilateral tendency to
competition in the relations between groups. Positive interdependence with regard
to positive identity is intrinsically impossible; therefore positive identity is not
to be classed, theoretically speaking, as just another one of many group interests
which may or may not lead to intergroup rivalry. The importance of social com-
petition is that it requires no goal which is desirable to both groups of itself and
could therefore act as a reward in some non-intergroup situation.
A hypothetical example of a situation in which the two theories would make
differing predictions should help to clarify the distinct causal processes they assume.
Consider a phase I of this situation: Two face-to-face groups working unbeknown
to each other on identical tasks. The tasks are highly involving for the subjects
in that they accept task performance as reflecting certain highly valued qualities;
for example, subjects could be art students and the task some problem of aesthetics
or design. Suppose also that both groups will receive a reward, such as credit
points towards an art diploma, for satisfying certain low, easily attainable standards
of task performance - all subjects are confident that they can reach these
standards. The standards are defined ‘absolutely’, without reference to the per-
formance of others. Now as phase I1 suppose that in the middle of the task activity
both groups are brought together within sight and sound of each other and
informed that they are working on identical tasks for the same reward. In this
latter situation social competition theory would predict the evocation of social
comparison processes and the devolopment of competitive feelings and behaviours
between the groups. In terms of the rewards for task activity specified in this
example Sherif‘s theory should not predict the development of competitive
relations between the groups; before Sherif could legitimately make this predicition,
phase I1 would have to incorporate new instructions that the credit points would
be given only to the better performing group. Sherif could posit the spontaneous
introduction by the groups of other goals, which were negatively interdependent,
into the second phase to explain any competition which might develop; but this
would only contribute to turning his previously fruitful account of one kind of
competition into a generally applicable but theoretically futile tautology. The
few experiments (cf. Ferguson and Kelley, 1964; Rabbie and Wlkens, 1971)
whose conditions resemble this example, i.e., an intergroup situation involving
separate task activity but with no reward contingent on intergroup rivalry, have
12 John C. Turner

reported a tendency for competition between groups to be triggered.

To conclude this introductory discussion it is worth making an explicit distinc-
tion between four kinds of intergroup competition. This classification will be
helpful from the standpoint of gauging how much of an explanatory burden social
competition may be required to carry. Firstly, there is competition which is
characterized primarily by the independent desires of various groups for a material
reward which can be gained by only one group. ‘Material‘ is not meant narrowly;
it could, for example, encompass such things as control of a political or social
institution, The theme of this competition is expressed by the notion of a ‘conflict
of interests’. At the other pole, there is what we have referred to as ‘social com-
petition’, arising from the social comparative aspects of social identity as they
interact with shared values. Its generally necessary conditions (their sufficiency
is a larger problem) are the salience of the intergroup situation and the possibility
of differentially valued actions relevant to the particular social categorization into
‘groups’.
The third type of competition is defined by the overlap between the first two:
Where a material reward to some extent valued of itself serves as a token or
symbol of a value differential associated with a possible social comparison
between groups. It is an open question whether competition in this case has its
own distinct behavioural repercussions or whether this kind of situation tends to
collapse into one of the other two types, depending on, for instance, just how
much independent value is possessed by the token or the degree of arbitrariness
in the relation between symbol and that which is symbolized. This is also an
important question because intuitively many workers tend to assume that the
use of, for example, small monetary rewards produces a situation of conflict of
interest, whereas in fact the results often seem more intelligible if one assumes
that the reward had an especial effect only insofar as it helped to make salient
the possibility of what we referred to as social competition.
A fourth type of competition or competitive situation is worth suggesting
although it will not be discussed in any detail; it is the counterpart of the third
type in that it is defined to some extent by an overlap between the first two and
is presumably a form of transition between them. It differs in the direction of
transition; here it is a social-competitive situation which gives rise to a conflict
of interest. This might happen where comparison results in a stable and explicit
inequity between two groups and thus the desire for positive self-evalution leads
to directly conflicting group interests with regard to the maintenance of the
comparative situation as a whole. Articles by Thibaut (1950), Kelley (1951) and
Manheim (1960) provide some support for the idea that such a stable inequity
is an important factor determining the development of social competition into
 Social comparison and social idenfity 13

intergroup hostility.
No hard and fast answers are implied by this brief analysis of intergroup com-
petition; the aim is to set up, if only temporarily, certain guidelines in order to
approach some current problems from a new and perhaps more promising angle.

3. The experiments I: Social categorization in intergroup behaviour

3.1 Why discriminate?

Billig (1972,p. 1) provides this description of Tajfel’s experiment in ingroup-
outgroup discrimination: ‘His subjects are divided into two groups [categorized]
on the basis of performance on some trivial task. These groups are as “minimal”
as possible; there is no interaction between the group members. Each subject is
told only which group he is in. He does not know who are the other members of
his group or who are in the other group. He has to do another task, this time on
his own. This task consists of awarding real money to the other subjects, but he
does not know the identity of the person to whom he is awarding the money.
All he knows is that he can award amounts of money to “a member of
his own group” (other than himself) or to ‘% member of the other group”. It has
been consistently found that subjects in this situation show ingroup favouritism,
i.e., they award more money to members of their own groups than to members
of the other group. These choices seem to be irrational at face value: they do not
benefit the chooser nor are they designed to extract as much money as possible
from the experimenter.’
Although the subjects’ choices were not designed to extract as much money
as possible from the experimenter (the Maximum Joint Profit strategy), it could
be argued that they were meant to benefit the chooser: Each subject was told that
he would receive the total amount of money allotted to him by the other subjects,
so if he assumed others would tend to choose as he did, there is perhaps a sub-
jectively rational basis for giving money on a ‘group’ basis where an intergroup
choice is required. However, the above description does not make clear the
crucial findings: ‘When the Subjects have a choice between acting in terms of
maximum utilitarian advantages to all (the Maximum Joint Profit, MJP, strategy)
combined with maximum utilitarian advantage to members of their own group
(the Maximum Ingroup Profit, MIP, strategy) as against having their group win
on points at the sacrifice of both these advantages (the Maximum Difference, MD,
in favour of the ingroup strategy), it is the winning that seems more important
to them. It is clear from the analysis of the findings that this is a deliberate
14 John C. Turner

strategy adopted for their choices, although they are aware of the existence of
the alternative strategies’ (Tajfel et al., 1971, p. 173).
Thus there is evidence for a deliberate strategy of intergroup discrimination
followed even when the subjects’ group getting more than the outgroup directly
conflicted with simple ‘material’ gain for the ingroup and when this discrimination
was evoked even by a ‘flimsy’ ad hoc intergroup categorization. Tajfel interpreted
these results ‘in terms of a “generic” social norm of ingroup-outgroup behaviour
which guided the Subjects’ choices since the Subjects classified the social situation
in which they found themselves as one to which this norm was pertinent. . .’
(Tajfel et al., 1971, p. 174). The questions that must be answered are what is
the psychological nature of this ‘generic’ norm (what kind of discrimination is
this), and why is the social situation classified as one to which the norm is
pertinent? The submission of this paper is that the experimental situation was of
a kind to evoke processes of social competition.
The subjects, who were schoolboys and acquainted with each other, were made
fully aware in the first part of the experiment and by way of the instructions in
the choice booklets, that they were divided into two groups, i.e., a dichotomous
social categorization was accomplished.
The major independent variable was not simply ‘group classification’. There
was superimposed on mere classification into groups the individuals’ own locations
within that classification. This was done in such a manner that there was no overlap
in membership between the dichotomized social classes; for this reason alone
it is possible to collapse ‘membership’ and ‘groupness’ into an ‘ingroup-outgroup’
dimension.
Thus, given that the experimenter provided, if only temporarily, a social identity
for each subject in the sense that each subject was specifically located within a
social category, two findings support strongly the case for social competition.
Firstly, whenever choices differed from a fairness strategy they consistently
tended to benefit the ingroup (altruism was possible but not displayed); and
secondly, experimental or rather ‘matrical’ variations designed to uncover the
relative pull of certain variables within this strategy of ingroup benefit showed
that what seems important is to achieve maximum difference between groups
rather than absolute monetary benefit for one’s ingroup. The first finding argues
for the existence of a definite common value differential predictably related to
the achievemenVpossession or reception of morelless money. The second, on the
other hand, makes it clear that the value associated with money does not derive
simply from the independent desire of each group for money, i.e., this is not
primarily a ‘conflict of interest’ type of discrimination. The ingroup chooses to
obtain more money but ‘more’ understood precisely as a means to intercategory
 Social comparison and social identity 15

differentiation, a process of achieving distinctiveness from the other group, both

because ‘more’ money symbolizes the positive sign of the value attached to the
direction of differentiation as well as the socially relative nature of valued dis-
tinctiveness.
The results of the experiment are consistent with the general proposition deriving
from the earlier theoretical discussion - that in any intergroup situation where
categorization fulfils the functions of social identification and is relevant to
comparison on a differentially valued dimension, members will attempt intergroup
differentiation to provide for themselves a positive self-evaluation.
Turner (1973) performed a study to corroborate this interpretation of Tajfel’s
experiment. He argued that the process underlying intergroup discrimination was
primarily social comparison and not the desire for monetary reward and that
therefore money functioned merely to represent a shared dimension of comparison
associated with value. If this were the case, then the same effects should be found
where choices were in terms of a purely comparative dimension whose differential
value was completely abstract, i.e., a continuum on which no point had per se
either a concrete meaning or a specific value except in terms of an overall attribu-
tion of comparative value between persons or groups. The experiment as a whole
had four conditions with 22 subjects in each; half the subjects were in the money
(M) condition (the usual Tajfellian situation) and half were in the value (V)
condition. The V subjects were told that the task was to distribute ‘points’; the
points stood for nothing in particular but were to be thought of as in a game
where ‘the more points you get the better’ - ‘more’ could be taken in either a
relative or an absolute sense, and ‘you’ could refer to yourself, your group or all
you boys together. Each subject was told that he would discover at the end of
the experiment how many points he had gained. Every subject filled out two
choice booklets: The usual one where the subject could only choose for two other
subjects (00),and one where every choice was for the subject himself and another,
who was either an in- or an outgrouper (SO). Half of both M and V subjects
filled out the booklets in the order of 00 - SO and half in the reverse order.
The rest of the data will be described later in the paper; for the moment we are
concerned with the difference between M and V results in the 00 booklets. The
points continuum is a pure ‘value’ dimension since the points have no meaning
except in terms of value; likewise it is purely comparative since the value of the
points is given relatively not absolutely. (Of course, the subjects do not have to
use the dimension comparatively.) Since it was hypothesized that the intergroup
discrimination in Tajfel’s experiments was the result of a comparison process, the
first prediction was that similar discrimination would occur in the V condition
as in the M condition. To the extent that the use of points had a more purely
16 lohn C. Turner

comparative significance than the use of points representing money, it was pre-
dicted secondarily that discrimination would be greater amongst V than M subjects
as measured by the strength of the relative winning strategy (MD). This latter
prediction is important if the pull of MD is to be taken as evidence for comparison
processes: Subjects using a dimension whose only meaning is comparative should
reflect a greater preoccupation with that variabie. There was intergroup discrimina-
tion in both M and V conditions as measured by the pull of MIP and MD on MJP
(see Table 1; cf. Tajfel et. al. for a description of methods of assessment of
these ‘pulls’) and MD on MJP and MIP, and the pull of MD was significantly
stronger in V than in M conditions (see Table 2).

Table 1. Mean scores for the pull of muximum ingroup profit plus maximum
difference in favour of ingroup on muximum joint profit in the other:
Other booklets

Conditions Money Value

Before self Other choices 82.36 (2.57) b2.52 (2.98)
After self Other choices c1.8 (2.8) 4.29 (4.74)

1. There are no significant main effects nor 2. Scores may range from - 12 (maximum
interaction by analysis of variance; the discrimination against the ingroup) to +12
figures in parentheses are standard devia- (maximum discrimination in favour of the
tions. ingroup). All treatments show significant
ingroup favouritism (all tests are one-tailed):
a :z = 4.22, p < .0001 b : z = 3.91, p <.OW1
c : z = 2.95, p < ,002 d : z = 4 . 1 6 , ~< .OOO1

Table 2. Mean scores for the pull of maximum difference in favour of ingroup
on maximum ingroup profit plus maximum joint profit in the other:
Other booklets

Conditions Money Value

Before self Other choices 81.19 (1.62) ’2.5 (2.96)
After self Other choices cO.73 (2.6) d3.97 (4.46)

1. Money versus value: F = 11.59. df = 2. Scores may range from -12 (maximum
1/84. p <.01; there are no other significant difference in favour of outgroup) to +12
effects by analysis of variance. (Figures in (maximum difference in favour of ingroup).
parentheses are standard deviations.) Three treatments show significant ingroup
favouritism (all tests are one-tailed):
a : z = 3.38, p <
.0001 b : z = < 3 . 8 8 , ~< .OOO1
c : z = 1.3 p < .1 n.s. d : z = 4 . 1 , p<.OOOl
 Social comparison and social identity 17

3.2 Why identify?

If we define social identity as ‘the individual’s knowledge that he belongs to

certain social groups together with some emotional and value significance to
him of this group membership’, then a problem arises - what emotional and
value significance does membership of these ‘minimal’ groups have? In what
respect does a ‘minimal’ group contribute to social identity such that a member
is led to social competition? Our hypothesis is that where an intergroup situation
allows a positive self-evaluation on some dimension, then the individual con-
fronting this situation will define himself in social terms relevant to that dimension.
This proposition generalizes the assumption made about group membership in
point (B) of our original theoretical outline: That membership and aspiration to
membership of a group is partly a function of its perceived ‘competitiveness’ in
relation to other groups. Here ‘competitiveness’ is meant in a ‘comparative’ and
not in a ‘conflictual’ sense. Thus each schoolboy subject could have acted in
terms of the minimal ad hoc grouping, or in terms of the real-life grouping inherent
in the fact that the boys already knew each other via the classroom. However,
it was only the ad hoc dichotomy that was relevant to the criteria provided for
possible intergroup differentiation and thus relevant to the expression of the
desire for positive self-evaluation in this limited social situation. To act wholly
in terms of the real-life grouping in this situation would be to forgo a favourable
comparison. If a purely conceptual identification is made in terms of an ad hoc
categorization and is then ‘used’ in order to discriminate, to compete, for certain
reasons, one must find other reasons to account for the original conceptual
identification. The gist of the idea now being proposed is that the conceptual
identification itself is dependent on its relevance to the possibilities of social
competition within a situation. (That the ad hoc grouping has an effect in this
experiment may derive from the fact that at its own Zevel [as a ‘group’], the class-
room grouping was not explicitly involved in intergroup comparison. This is not
an argument against the social competition notion but a prediction one might
make from it.)
Thus whilst it is true empirically that under certain conditions categorization
per se is sufficient for intergroup discrimination, the proposition may be theo-
retically misleading to the extent that it suggests that ‘acceptance’ by subjects of a
categorization is automatic and its use inevitable rather than indicating that ‘ac-
ceptance’ itself is to be explained in terms of how the category is used in a given
situation. The danger of the former view is that it tends to take as a datum what
has yet to be established - that certain presciptions for behaviour somehow inhere
in a categorization such that it must produce its specific normative effect. A fuller
18 John C. Turner

discussion of this question can be found in Turner (1972). For the moment it
suffices to ask how far the available evidence lends support to the position adopted
here, namely, that subjects will identify with that category (of those provided) in
terms of which they can be socially competitive; or how far the evidence supports
the alternative view that categorizations possess inherent norms of ingroup bias
or at least automatically trigger off in some way intercategory differentiation.
Several studies have employed the technique of ‘coinciding’ a real-life categorial
division with a intergroup division in a game-playing situation (Wilson and
Kayatani, 1968; Doise, 1969; Uejio and Wrightsman, 1967; Swingle, 1969; Har-
ford and Cutter, 1966; Heller, 1966; Wrightsman et al., 1967; Hatton, 1967).
This ’coincidence’ should presumably accentuate intergroup differentiation if
categorization has its specific additional effect - in the main the results are
negative.
The weakness of these studies - that control was not exercised over the
various confounding and diluting influences inevitably associated with real-life
categories such as particular states of ihtergroup relations and attitudes - was
overcome in a small experiment by Turner (1972). He used an identical set-up
to that of Tajfel et a2. (1971) with one important modification: Self-interest, in
this case monetary, was explicitly and directly involved in every choice. Whereas
Tajfel’s subjects could only distribute money to others, Turner’s subjects always
chose for themselves and another; half the others were ingroupers and half were
outgroupers (only one experimental condition is being described). Thus in this
design the possibility of competition was not confounded with the use of one
available categorization - as it is in the normal paradigm; the subjects could
compete just as well in terms of self-other as of in- and outgroup. If categorization
per se has its specific effect, then subjects should be more competitive to out-
groupers than to ingroupers; however, if as argued above a category will be used
only if it is relevant to a possible self-evaluation in a situation, there is no reason
to expect any intergroup discrimination since there is no favourable evaluation
to be gained through the group dichotomy not already, indeed more, available
through the self-other dichotomy. In fact self-other but not intergroup competition
was found; subjects treated outgroupers and ingroupers in the same fashion.
Since the number of subjects in the experiment was small and also since
instrumental rather than social competition could be used to explain the lack
of intergroup effect, a further experiment was run in which self-other non-
monetary choices were required. This experiment (Turner, 1973) has been
described above: 44 subjects made self-other (SO) choices before other-other
choices (00);44 subjects chose in the reverse order (00-SO); half of both
00-SO and SO-00 subjects distributed points representing money (M) and half
 Social comparison and social identity 19

just distributed points (V). We are only describing the SO choices of the SO-00
subjects, both for the M and V subjects. The V subjects represent a similar
situation of category-conflict as the M subjects but without any complication of
monetary self-interest, i.e., of instrumental competition. The prediction was con-
firmed that for both M and V subjects there would be self-other but not inter-
group discrimination.* Thus, it has been found possible to delineate the conditions
under which categorization into groups is not sufficient for intergroup discrimina-
tion and moreover to demonstrate this in such a way as to support the view that
these phenomena cannot be derived from processes of instrumental self- or group
interest (see Table 3).

Table 3. Mean scores for the pull of maximum profit plus maximum difference
in favour of self on maximum joint profit in the self: Other booklets.

Conditions Money Value

(Type of other) = Ingrouper Outgrouper Ingrouper Outgrouper
Before other: Other choices 82.75 (5.31) b2.68 (5.33) f5.05 (5.37) d5.98 (5.4)
After other: Other choices e2.29 (5.17) 5.27 (5.78) 82.43 (5.59) 6 . 8 6 (4.73)

(Figures in parenthese are standard devia- (F = 13.7; df 1/84; p <.Ol), but not in the
tions.) before other: Other choices (F = 0.25; df
1. Ingrouper versus outgrouper. F = 8.84, 1/84 as.).
df 1/84, p < .01. This effect is explicable
in terms of the significant interaction be- 2. Scores may range from -12 (maximum
tween ingrouper versus outgrouper and be- discrimination against the self in favour of
fore versus after other: Other choices. F = the other) to +12 (maximum discrimination
5.14, df = 1/84, p < .05.There was a sig- in favour of self against the other). All
nificant simple effect for ingrouper versus means show significant discrimination in
outgrouper in the after other: Other choices favour of self (all tests are one-tailed):
a : z = 2.38, p < .01; b : z = 2.31, p < .01; c : z = 4.32, p < .OO01;
d : z = 5.09, p < .0001; e :z = 2.03, p < .025; f : z = 4.19, p < .Owl;
g : z = 1.995, p < .025; h : z = 5.68, p < .0001.
It seems that the answer to our initial question ‘what emotional or value signifi-
cance does membership of these minimal groups have?’ is that subjects will
identify with a social category to the extent that such identification enables them
to achieve value significance, to the extent that it is the category most relevant to

2. Table 3 shows that there is in fact a discrimination in the before-money choices

significant main effect for the ingrouper whilst the slight tendency in the before-
versus outgrouper type of other. However value condition does not approach signifi-
this is almost wholly due to the ‘after’ con- cance.
ditions: There is no tendency for intergroup
20 John C . Turner

the desire for positive self-evaluation in the experimental situation. Thus, it can
be said that in the experiment by Tajfel et al. it was not the division into groups
which caused discrimination but rather that the group dichotomy was the only
existing categorization through which a more basic motivation might be expressed
- the subjects therefore had to use the categories provided.
The conception that social competition can function as a basis for identification
with a social category has relevance to fundamental problems of social identity
which cannot be discussed here. However, to conclude this section some further
evidence from Turner’s (1973) experiment can be cited in order to demonstrate
the legitimacy of the notion of ‘identifying with’ rather than of just ‘using’ a
category. In the above experimental data, identification and use of a category
were operationally synonymous: Where discrimination has occurred, identification
was assumed as some intermediate process between categorization by the ex-
perimenter and the observed behaviour of the subjects. The research problem,
therefore, is to demonstrate the effect of this intermediate process in contra-
distinction to what would be expected on the basis of the hypothesis that subjects
merely used the categories provided. As mentioned above 44 subjects filled out
two booklets in the order SO-00, and another 44 subjects filled them out in the
reverse order. It was predicted and confirmed that there would be no significant
intergroup effect in the SO choices of the SO-00 subjects: They had no reason to
use the group dichotomy. The 00-SO subjects also had no reason to use the
group dichotomy in their SO choices, but they were confronted with the SO
choices after the 00 situation in which they had no choice but to use the group
categorization in order to compete. If some process of identification takes place
in using the group dichotomy, i.e., the subjects tend to define themsehes in terms
of that category, then an intergroup effect should be found on the SO booklets
of those subjects who first filled out the 00 booklets. This result was predicted
and found: These subjects were more competitive to outgroupers and less com-
petitive to ingroupers unlike the subjects who filled out the SO booklets first
(see Table 3). This result was true not only for V but also for M subjects - in
other words outgroup discrimination and ingroup altruism conflicting directly with
monetary self-interest were found. M subjects in a situation where they could give
money directly to themselves sacrificed money in order to favour other ingroupers.
Of course, self-other competition still predominated; but it is hard not to be
impressed by the fact that such an artificial and flimsy categorization could have
an influence in a situation characterized by obvious and explicit self-interest.
We need not discuss the rest of the detailed findings to sustain this interpretation;
however, the data for the pull of MD on MS plus MJP is worth presenting
(Table 4) to show that M subjects do clearly perceive and are influenced by
 Social comparison and social identity 21

monetary self-interest as compared with V subjects.

Table 4. Mean scores for the pull of maximum difference in favour of self on
maximum profit in favour of self plus maximum joint profit in the self:
Other booklets

Conditions Money Value

(Type of other) = Ingrouper Outgrouper Ingrouper Outgrouper
Before other: Other choices 80.86 (3.73) b0.38 (4.24) 4.52 (5.39) d5.64 (5.97)
After other: Other choices e2.56 (4.1) f2.09 (5.67) g3.02 (5.09) 64.56 (4.43)

(Figures in parentheses are standard devia- F = 4.82, df = 1/84, p . < .05. No other
tions.) effects are significant.
1. Money versus value: F = 14.95, df =
1/84, p < .001. 2. Scores may range from -12 (maximum
Interaction between money versus value and difference in favour of other) to +12
before versus after other: Other choices: (maximum difference in favour of self):
a :z = -1.06, a s . c :z = 3.85, p < .OW1
One-tailed d :z = 4.34, p < .0001
tests e :z = 2.87, p < .005 tests g : z = 2.72, p < .005
f : z = 1.69, n.s. h :z = 4.73, p < .OOO1

3.3. Convergence and divergence

Our discussion has widened from an analysis of competition to include, albeit in
a limited way, the problem of the acquisition of social identity. The need for
positive self-evaluation has certain definite consequences for intergroup relations,
yet, at the same time, actual and/or potential comparisons between social groups
can function as a source of pressures to redefine the ‘self’. Social identity is not
only expressed but is also in some sense tested in the active comparison of ‘in’
- and ‘out’ - categories, and the outcome of the process may be a redrawing of
the boundaries between the social categories which define what the individual is
and what he is not. The connecting thread between competition and the development
of social identity per se is found in understanding self-evaluation as a social
process: It is against social reality that the individual measures himself, and it is
through social reality that he, and not just others, can come to see himself in a
new light. What is being proposed is a framework for the analysis of self-evaluation
as a social process; this will perhaps be clearer if we consider briefly the require-
ments for social competition.
22 John C. Turner

a) Both the categorization and location of individuals within it must be explicit,

non-overlapping, not in direct conflict with more strongly felt group member-
ship and relevant to the criteria for differentiation.
b) The dimension of comparison must be recognized and shared by both groups.
c) Both groups must share a similar attribution of value to a range of possible
actions.
d) There must be room for differentiation: Where a stable value inequity exists
between the two groups that action cannot remedy, it is unlikely that com-
petition will persist as a means to positive self-evaluation. Conversely, the
strongest competition might be expected to occur where in reality there is
the least reason to prefer one group to the other. This requirement is really a
general one that there be no handicap or hindrance perceived as rendering
action futile.
To make these requirements determinants of competition is simply to focus on
action as a dependent variable. Insofar as categorization, comparison and values
are necessary and interdependent aspects of the evaluation of self in society,
‘action’ is merely one form of psychological activity that may lead to a positive
social identity. In any concrete situation other alternative avenues may be used:
There may be changes in (1) self-categorization, i.e., changes in group identity,
(2) the dimensions of comparison and (3) the values associated with a particular
comparison. The discussion in the previous section of the conditions under which
an identification is made is precisely an example of focussing on (l), treating the
location of the self within a category as a dependent variable.
These changes include the cases where new groups are formed, new dimensions
of comparison sought out and new values invested in a comparison. It is often
difficult to distinguish between use of new comparative dimensions and changes
in associated values since the phenomena are closely related empirically: One
makes comparisons which are worth making. The conceptual distinction between
(2) and (3) is worth retaining for those not uncommon situations where two
persons or groups relate themselves in terms of the same characteristics but attach
positive and negative value to different poles of the comparative continuum. An
‘inferior’ group in a situation in which requirement (d) is not met has several
options - to dissolve itself, to leave the situations so as to escape pressures for
mutual comparison between the groups, to adopt values so different from the other
group’s as to make itself non-comparable or to fight to legitimize a new valued
comparison on which it is or can be superior to the other group. In this context
to seek legitimacy for a cornparison means to gain acceptance of it by the other
group, either directly through persuasion or coercion or indirectly through ap-
pealing to some external or higher social authority. This last intergroup situation
 Social comparison and social identity 23

is not social-competitive; it is subsumed by the fourth type of competitive situation

in our initial analysis and is perhaps best described as a ‘conflict of values’.
It is worth citing some evidence that these alternative avenues are used. In
regard to (1) Mann (1961) using a procedure which fixed the actual group member-
ship of individuals found support for the hypothesis that ‘in a hierarchy of groups
the majority of persons in any group will prefer to direct their positive relations
to the hierarchical region comprising the membership group and higher groups and
their negative relations to the hierarchical region comprising the membership group
and lower groups’. Positive and negative relations here refer primarily to member-
ship aspiration.
In regard to (2) Lemaine (1966) put children’s groups in a situation of inter-
group comparison involving the construction of a hut where one group was
handicapped with respect to that particular dimension and found that the handi-
capped group responded by seeking out and attempting to legitimize a new com-
parative dimension (involving the hut’s environs) on which it was clearly superior.
Experimental corroboration of (3) is lacking, but a study by Peabody (1968)
is suggestive. He had judgments about Chinese living in the Philippines Filip-
pinoes, Americans and Japanese made by the first two groups using 14 pairs of
scales designed to separate evaluative and descriptive aspects. The results showed
that the groups tended to disagree about their evaluation of each other but agreed
about the descriptive characteristics of each group and that all groups tended to
evaluate themselves positively. Also relevant to the changing of values as a means
to a positive intergroup comparison is Fishman’s (1968) notion of the ‘ideologizing’
of intergroup differences, the creation of values associated with intergroup dif-
ferences. Kelman (1969) consequently suggests ‘that an ethnic group becomes a
nation when it begins to ideologize its customs and way of life. That is it goes
beyond the conception of “this is the way we do things’’ to a conception of “there
is something unique, special and valuable about our way of doing things”. It is
ideologizing of this sort that makes it possible to develop allegiance to and invest
one’s identity in a collectivity that goes beyond one’s primary group’ (Kelman,
1969, p. 284).
Our list of requirements for social competition is still tentative, somewhat vague
and probably incomplete; however, it is suggestive for designing future research
and for considering the facts about intergroup competition that have already
accumulated. Social competition seems to possess distinctive determining condi-
tions; possibly it will have distinctive psychological consequences for intergroup
relations. Thus it may be asked whether ‘discrimination’, understood as arising
from social competition, necessarily implies outgroup depreciation as is assumed
in the conflict of interest situation? Other experimental work, closely related to
24 John C. Turner

Tajfel’s, is useful in considering the form taken by socially competitive discrimi-

nation.

4. The experiments II:Some data on intergroup differentiation

We shall begin by looking at two investigations of intergroup behaviour undertaken

by Rabbie and co-workers (Rabbie and Horwitz, 1969; Rabbie and Wilkins,
1971). Factors extraneous to social comparison are present in both these ex-
periments, so there can be no unequivocal claim that social competition underlies
their results. Rabbie and Horwitz impose common fate on their experimental
groups and vary the nature of its external origin whilst Rabbie and Wilkins allow
within-group interaction in their experimental situation. Nevertheless there are
good reasons for examining these studies. The former creates an intergroup
situation where one group’s gain is contingent upon the deprivation of another
group, i.e., a conflict of interest situation. And yet the resulting ingroup biases
in evaluation cannot be explained within the framework of Sherif‘s theory. In
addition, other data appear to contradict the idea that categorization into groups
per se is sufficient for intergroup discrimination.
The latter study adresses itself directly to intergroup competition. The authors
take conflicts of interest (negatively interdependent group goals) as their definition
of the competitive situation and in so doing again reveal its inadequacy. They
create conflict and social comparison situations and find competition in both.
Further, the intergroup differentiation they find in both situations is amenable
to interpretation in terms of comparison but not conflict - this is so despite the
fact that originally the comparative situation was not expected to yield competi-
tion at all. This second experiment is a direct validation of an earlier discussion
of the conditions under which social competition would be aroused.
Rabbie and Horwitz (1969) investigated the minimal conditions that produce
a more favourable evaluation of the ingroup than the outgroup. The experiment
comprised four conditions - in each of which subjects were classified into two
groups. In the control condition (group classification per se ) the groups were
neither rewarded nor deprived. In the three experimental (common fate) con-
ditions the groups were either rewarded or deprived by chance alone, by the
arbitrary choice of the experimenter or by the choice of one of the groups. Thus
in the experimental conditions there were two overlapping variations: Firstly
half the subjects received a reward and half a deprivation, and secondly subjects
experienced three kinds of responsibility for this reward or deprivation. The
dependent measures were ratings of personal attributes of in- and outgroup
 Socid comparison and social identity 25

members, the attributes of each group as a whole and sociometric choices. Thus
the dependent variable in this study is evaluation rather than behaviour. (It should
be remembered that Tajfel’s paradigm prevents interaction. Filling in rating
scales may be construed as ‘action’ no less than choosing monetary rewards, and
to this extent there should be no marked disjuncture between the two ways of
making a discrimination). The difference between Rabbie and Horwitz’s control
condition and the experimental conditions as a whole would seem to be the
salience or importance of classification into groups for the subjects since ‘common
fate’ is in a sense a criterion for group classification rather than something neces-
sarily distinct from it. We are dealing here with a complicated set of independent
variables because the variations within the experimental conditions are not varia-
tions in this classificatory salience: The subjects experience different kinds of
common group fate. If anytbhg, the ‘group responsibility’ condition represents
a superimposition of a ‘conflict of interest’ variable that should make the results
less susceptible to interpretation in terms of social competition.
The experiment yielded two main results: The control (group classification
per se) condition showed no discriminatory evaluations, but the combined ratings
of the experimental conditions showed a significant ingroup bias (mainly due to
the ‘chance’ condition, i.e., where winning or losing was most ‘fair’ - see require-
ment (dj above).
The interesting result (from our point of view) was that where common experi-
ence of reward or deprivation enhanced ‘groupness’, ingroup bias was found
which cannot be explained in terms of conflict of interest. If the bias was a
response.to winning or losing the rewards per se then, as Rabbic and Horwitz point
out, winners should rate the ingroup positively because it mediated a reward,
and losers should rate the outgroup negatively because it mediated a deprivation
(frustratiodaggression theory). Thus the winners should have rated their ingroups
more positively than losers did theirs, and outgroup ratings by losers should have
been lower than outgroup ratings by winners.
Neither result is found. If intergroup differentiation is at all related to the
independent value of the reward it is hard to see why, for instance, there was no
difference in outgroup ratings between subjects who were deprived by chance and
those who believed themselves deprived by a deliberate act of the outgroup. A
cognitive balance explanation (p similar to o induces p likes u, and p dissimilar
to o induces p dislikes 0) is also ruled out since contrasting experiences with out-
group members in the experimental conditions did not generate more negative
ratings of the outgroups in these conditions than in the control condition. Ingroup
bias, it appears, was a matter solely of more positive ingroup evaluation; there
was no outgroup depreciation at all.
26 3ohn C. Turner

This fact alone argues against interpretation of intergroup differentiation in

this experiment as akin to the processes involved in ‘realistic’ intergroup com-
petition. From the standpoint of social identity, groups aim to move to more
positive value in comparison with other groups. It is legitimate to ask whether
social identity is more positive when the outgroup is good but the ingroup better
or when the ingroup is better merely because the outgroup is bad? It would seem
reasonable to argue that it is more positive in the former case, even if we are not
yet in a position to support this with precise theoretical arguments. Requirement
(d) in the previous section of this paper already makes it clear that, if the outgroup
is perceived as ‘bad’ prior to social comparison, then social competition is not
likely to be evoked. An ‘easy’ victory in sport is often valued less than a victory
in a close, hard-fought contest - presumably because under certain conditions
positively valued distinctiveness requires that the outgroup not be depreciated.
The precise nature and generality of these conditions are topics for future research.
If ingroup bias in the experimental conditions is to be interpreted as support
for processes of social competition, its absence in the Rabbie and Horwitz control
condition must be explained. The subjects were divided at random into two groups
of four and seated at either side of a screen to prevent their seeing each other.
Then one group was publicly designated as the ‘blue’ group and the other as the
‘green’; members were repeatedly addressed as the ‘greens’ or ‘blues’ (not green
group or blue group). However, dthough it might seem that classification was
thus accomplished, other aspects of the situation and the rating procedure seem
designed to weaken both the salience of this classification and of the subject’s
location within it. Thus, they were told that division into groups was for ‘adminis-
trative reasons only’, that the ‘subjects would not work together in any way’ and
that members must not talk with one another ‘since this would interfere with your
task later on - to give unbimed impressions of the personality characteristics of
the other subjects in this room’ (not of members of own and other ‘group’?)
[italics mine]. In the ratings of first impressions each subject stood up and read
aloud some personal background information he had prepared and was rated on
scales such as responsibility, consideration, fearfulness, cordiality and soundness
of judgment by the other subjects. Insofar as the basis for using these scales was
‘personal’ background information, the question arises as to what comparison could
or did take place between the groups in terms of which these scales would be
meaningful? (This is not a criticism of design; it suggests rather that Rabbie and
Horwitz were successful in controlling the real source of bias, intergroup com-
parison; it is impossible without a detailed knowledge of the control situation to
choose either the ‘weakness in categorization’ or ‘lack of relevant intercategory
comparison’ interpretation of these results.) While it is true that these scales
 Social comparison and social identity 27

revealed ingroup bias for those subjects who had experienced a common group
experience between classification and impression rating, who were to some extent
forced into a between-group comparison, including as Dann (1972) puts it ‘the
possibility of differential outcome’, it can be said that for the control subjects their
effect would be as much to weaken the relevance of group classification as to
measure its discriminatory consequences.
The instructions prior to the sociometric measure are also slightly ambiguous:
‘We said earlier that you would not work together, but we have changed our
minds about that.. . That is why we would like you to work on a group task
which I will described in a moment.. . We need this information to form the
new group in which you will be working. In this way we can take account of
your preferences’ (Rabbie and Horwitz, 1969, p. 271). If the subjects take these
instructions to mean that they will work on a task according to prior group
classification, then there is evidence (Darley and Berscheid, 1967) that they should
show increased liking for their own group. The fact that there is no sociometric
bias towards own group suggests that they do not anticipate interaction on the
pre-existing group basis - indeed the words ‘to form the new group in which you
will be working [italics mine] indicate exactly the opposite. If existing social identifi-
cation is salient (and a corresponding between-group comparison possible as in the
‘common fate’ conditions), then these instructions allow measurement of inter-
group preferences; if, as with the control subjects, this is not especially so, then
their effect could be to diminish its relevance even more. Thus there are grounds
for thinking that Rabbie and Horwitz’s control condition is not a good test of
whether explicit social identification necessarily produces ingroup bias in evalua-
tion: Explicit classification into groups was probably inadequate, the salience of
group membership may have been weakened by features of the experimental
situation, the rating scales may not have been perceived as relevant to the group
classification owing to prior instructions. Each or all of these factors may have
been responsible for the lack of ingroup bias. A methodological question raised
by this experiment is: When is rating of own and other group on certain scales
merely registering a discrimination and when does it imply a process of com-
parison? Post-hoc analysis of an experimental set-up, as with the above, tends to
be speculative and yields answers of limited utility. What is required are more
experiments designed directly to separate out the possibility of discriminating with
and without a relevant intergroup comparison.
Whilst the response to ‘winning’ or ‘losing’ in Rabbie and Horwitz’s experiment
is not to be explained in terms of conflict of interest, the data do not, of course,
furnish positive evidence for social competition processes. A later study by Rabbie
and Wdkens (1971) provides more compelliig evidence as well as corroboration
28 John C. Turner

for the idea that social competition implies intergroup differentiation primarily
through ingroup bias and not outgroup devaluation. They created three kinds of
intergroup situations: No anticipation of future interaction within or between groups
(NA), no intergroup competition (i.e., no instrumental competition) but task inter-
action in the presence of a task-interacting outgroup (NC) and intergroup com-
petition where the groups worked on similar tasks and were negatively inter-
dependent with regard to a reward (C). Subjects rated the attributes of individual
group members and of the in- and outgroups as a whole - after the preliminary
instructions and then again after the NC and C groups had worked on their tasks.
The NA subjects also made ‘individual’ and ‘group’ ratings twice, and so there is
‘before’ and ‘after’ data for all three conditions although the NA subjects did not
actually work on a task.
NC and C subjects rated the ingroup significantly more favourably than the NA
subjects on the ‘before’ individual ratings, but there was no significant difference
between the conditions on the ‘before’ group ratings of the ingroup. In fact NC
and C subjects gave the other group a more favourable rating than the NA subjects
on two ‘group’ items. Rabbie and Wilkens explain this by the idea that ‘ex-
pectation of interaction with the other group made it more attractive to the subject’
(p. 223). These differences cannot be attributed solely to the anticipation of future
interaction since actual interaction within and between the groups before task
activity was not controlled. Given our discussion of possible reasons for the lack
of ingroup bias in Rabbie and Horwitz’s (1969) control condition, it is worth
noting that in this study there was intergroup differentiation in the NA condition
on the ‘before’ group but not the ‘before’ individual ratings. Since actual interaction
was not controlled in the NA condition, this is not direct evidence for the effect
of minimal group classification, but it does indicate that different results can
accrue according to the individual versus group attribute nature of the ratings.
The NC groups were told that a monetary reward would be given for the at-
tainment of certain ‘absolute’ standards of quality of task-solution; the C groups
believed that they had to do better than the other group to win a prize. A realistic
competition standpoint should predict between-condition differences; a social
competition theory sees no basic difference in the causal psychological processes
evoked by such intergroup situations (with only a smalI monetary reward). A check
on the ‘competitive’ manipulation revealed no significant differences in how com-
petitive the subjects felt with regard to the other group between the C and NC
conditions.
‘Even in the NC condition, the feeling of competitiveness was fairly high.
Apparently, when two groups are working side by side and expect to be compared
with each other, it is very difficult to minimize these feelings under the conditions
 Social comparison and social identity 29

of the present and other experiments (Ferguson and Kelley 1964). Thus psycho-
logically the C and NC conditions are rather similar to the subjects’ (Rabbie and
Wilkens 1971, p. 224; italics mine).
Differences between C and NC conditions could be expected to result from the
fact that explicit negative interdependence helps to define the situation more
clearly as competitive, but if both conditions are basically socialcompetitive,
these differences should be minimal and only of degree. In fact there are no
significant differences between C and NC conditions in the further results to be
presented, i.e., there is no interaction effect for NC/C and ingroup/outgroup
ratings.
Intergroup differentiation was found on the group ratings in all conditions on
both ‘before’ and ‘after’ ratings, although there was also a significant increase in
positive evaluation of both in- and outgroup from ‘before’ to ‘after’ for all con-
ditions. A positive correlation of 0.75 between in- and outgroup ratings was found.
On the individual ratings intergroup differentiation was apparent on the ‘after’
but not the ‘before’ ratings. This increase in differentiation was significant; it seems
that some ‘time’ (social interaction?) was necessary for ingroup bias to manifest
itself in the individual ratings. Increase in positive ingroup evaluation was more
important than outgroup depreciation in the growth of this differentiation. Since
NA subjects did not work on a task between ‘before’ and ‘after’ and since never-
theless the NA data parallels the C and NC data from the point of view of inter-
group differentiation, it is clear that neither the growth of ingroup bias on the
individual nor the stability of differentiation in the group ratings can be explained
in terms of anticipation of interaction or factors involved in task activity itself.
Taking into account all the various facts - the delay in intergroup differentiation
on the individual ratings, its immediacy on the group ratings, the lower ingroup
evaluation of NA subjects on the individual ratings, the higher ingroup evaluation
of the C subjects on all measures, the lower rating values of the NA subjects in
general - it would seem that the overall pattern of the data is most probably
determined by the salience and explicitness of the ‘intergroup’ nature of the
situation for each group, whether conveyed through preliminary instructions or
the type of dependent measure. The effect of lack of anticipation of future inter-
action would be presumably to weaken the subject’s perception of own location
within a given group classification just as the effect of explicit negative inter-
dependence with regard to a reward should be to strengthen it.
The interpretation is tentative owing to the impossibility of assessing the con-
tribution of actual within-group interaction to increased ingroup liking. Although
the obtained positive correlation between in- and outgroup ratings is not evidence
for social competition (in that the circumstances under which this should occur
30 Juhn C. Turner

have not been specified theoretically), it is not inconsistent with it. The correlation
is inconsistent with Sherif‘s realistic competition theory as well as being hard to
explain by between-group interaction (screens were erected between the groups
during activity on the tasks).
Ferguson and Kelley (1964) also provide data that the fact of ‘intergroupness’
rather than realistic competition is sufficient to evoke psychological processes
of a competitive nature. Twenty-two ad hoc groups worked on three tasks. Each
group worked separately but within sight of another group working on the same
tasks. Each group was instructed to do its best but great care was taken to elimin-
ate any suggestion that the groups were in competition with each other. In this
situation the authors found that evaluation of own and other group’s product still
showed significant ingroup bias (even where the outgroup’s product was clearly
superior, the ingroupers tried to minimize its superiority). The crucial factor was
attraction to the group and not ‘cognitive context acquired through taking part in
the production process’. Attitudes towards the group before the work began were
related to the subsequent tendency to overevaluate the group’s products and
‘measures of attitudes toward the group showed an effect similar to the over-
evaluation of the group’s products: one’s own group looks better than the other
group’. Thus division into minimal groups produces the same intergroup differ-
entiation in favour of ingroup in evaluating own and other group’s products that
is apparent in in- and outgroup ratings and monetary discrimination: It is hard
not to suspect that all three facets of differentiation reflect the same process.
To support an interpretation of this study in terms of social competition one
can do no better than cite the authors’ own conclusions: ‘In the post-experimental
discussions we were repeatedly confronted with comments from our subjects that
they felt a strong tendency to compete - that they wanted to excel the other group
in their productions. In view of the minimal nature of these groups we were
impressed by these indications that competition was triggered in this situation.
Apparently when working under conditions of high motivation on an ambiguous
task, one on which performance is difficult to evaluate, the sheer presence of
another group working on the same task leads to a tendency to compare results.
With this comparison, the seeds of competitive feelings and behaviour are sewn’
(Ferguson and Kelley, 1964, p. 227).
Blake and Mouton (1961, 1962) have found the same ingroup bias in actual
knowledge of own and other group’s task solutions; interestingly enough, the type
of item most frequently missed by group members are those that are common to
both group’s solutions, ie., those which, we would argue, blur the distinctiveness
the groups were trying to establish. They also found the same bias when solutions
were assigned but not when solutions were assigned but not revealed; membership
 Social comparison and social identity 31

in a group which ‘owns’ a solution and competition were the critical variables in
this misperception of common items.
Two other studies should be mentioned (Bass and Duntemann, 1963: Wilson
and Miller, 1961) to illustrate the fact that winning or losing serves the function
of providing both groups with shared evaluations of own and other groups. Wilson
and Miller found that when the ingroup won and the other team lost, there were
large positive shifts in ratings of teammates but small positive shiits in ratings of
opponents. When the own team lost and the other team won, rating of team-
mates showed small positive shifts while ratings of opponents showed somewhat
large positive shifts in ratings of teammates but small positive shifts in ratings of
is judged impartially as better than another group’s, that group’s value rises
somewhat for both groups, and the losing group is devalued somewhat by both
groups. Since both groups start with higher evaluation of themselves than the
other these changes are most marked for the losing group. The most severe self-
devaluation of the losing group is only temporary, but the losing group does not
re-attain its initial high level of value.
The outgroup devaluation by the winners which would seem to dispute previous
findings is very small compared to the changes in in- and outgroup evaluations
of the losers. The authors note that ‘we undervalue other groups in comparison
with our own. When another group becomes our competitor, we undervalue it
even more (before announcement of victory or defeat)’. It would seem fairer to say
that we overvalue our own group in comparison with others; initial evaluation of
collaborating and competing groups was roughly similar. The most marked
devaluation of the competing group in comparison with the collaborating group
before the announcement of victory or defeat occurred for the losing group, and
this is perhaps because for some reason that group actually rated the competitor
more highly than the collaborator on the first rating. The general point to be made
from these studies, however, is that if changes in evaluation are to be explained in
terms of the satisfaction or frustration arising from realistic competition processes,
if the situation is one of conflict and not comparison, then outgroup hostility or
depreciation should be greatest for the losing group at the announcement of
outgroup victory. The data show an opposite picture; outgroup victory increases
the value of the outgroup in the eyes of the losing competitor.

5 . Conclusion

It would seem that there is evidence for the utility of a conception of social
competition as distinct from instrumental competition (conflict of interest). Evi-
32 John C. Turner

dence exitsts for processes of intergroup differentiation inconsistent with theories

of reward-mediated attraction to the ingroup or deprivatiom‘frustration-mediated
depreciation of the outgroup, for competition whose motive seems to be comparison
and not rewards in the normal sense. It is possible that social competition is the
parallel process linking over-evaluation of own product and ingroup bias in ratings
and monetary discrimination.
The parameters of what has been loosely called social competition theory
need a fuller elaboration before the term ‘theory’ can be properly used. The
variables that determine competitiveness, whether singly or in combination, need
to be given both conceptual and operational significance - the comparability of
groups, the relevance of a categorization to a comparison, the sharpness or gradu-
alness of a value differential, all these kinds of concepts require a more systematic
exposition.
It has been seen that intergroup discrimination is produced by processes general
to a variety of intergroup situations.
Ingroup bias is produced in situations with or without negative interdependence
as regards a reward, with or without a reward, with or without task activity, inter-
action, behaviour in the narrow sense at all, whether the measure is allocation
of money, evaluation of the group or evaluation of the group’s product. There are
differences of degree in the data, but as far as can be ascertained they seem to
only be of degree. The need for a theory relating directly to the intergroup
situation is striking. This paper has presented some ideas which constitute an
‘analytic framework’ rather than a theory, but it is hoped that this framework will
prove a useful tool in the research that must be done before anything like a
‘theory’ will be feasible.

REFERENCES

Bass, B. M., and Duntemann, G . (1963), tion and similarity in intergroup b e

Biases in the overevaluation of one’s haviour. Eur. J . SOC. Psychol., 3, 27-51.
own group, its allies and opponents. Blake, R. R., and Mouton, 3. S. (1961),
J . confl. Resol., 7, 16-20. Competition, communication and con-
Berger, P. L. (1966), Identity as a problem formity. In I. A. Berg and B. M. Bass
in the sociology of knowledge. Eur. J . (Eds.), Conformity and deviation. New
SOC., 7, 105-115. York, Harper. Ch. 6.
Billig, M. G. (1972), Similarity-attraction - and Mouton, J. S. (1962), Comprehen-
and ingroup favouritism in minimal sion of points of communality in
group situations. University of Bristol. competing solutions. Sociom., 25, 56-
Ms. 63.
- and Tajfel, H. (1973), Social categorha- Coser, L. (1956), The functions of social
 Social comparison and social identity 33

conflict. New York, The Free Press. legitimacy. In J. N. Rosenau (Ed.),

DaM, H.D. (1972), The search for group- International politics and foreign policy.
ness clues. Paper presented at the New York, Free Press. Ch. 26, pp.
EAESP Small Group Meeting on Ex- 276288.
perimental Studies of Intergroup Rela- Lemaine, G. (1966), Inegalit6, comparaison,
tions at University of Bristol. Feb. 4-7. et incomparabilit6: Esquisse d‘une
University of Konstanz. th6orie de l’originalit6 sociale. Bull.
Darley, J. M., and Berscheid, E. (1967), Psychol., 252, 20, 1-2,1-9.
Increased liking as a result of the Manheim, H.C. (1960), Intergroup interac-
anticipation of personal contact. Hum. tion as related to status and leadership
Rela., 20, 29-40. differences between groups. Sociom.,
Doise, W. (1969), Les strategies de jeu 23 (4), 415-427.
l’int6rieur et entre des groupes de natio- Mann, J. W. (1961), Group relations in
nalitCs differentes. Bull. C.E.R.P., 18, hierarchies. J. SOC. Psychol., 54, 283-
13-26. 314.
- Csepely, G., Dann, H. D., Gouge, C., Peabody, D. (1968), Group judgments in
Larsen, K., and Ostell, A. (1972), An the Philippines: Evaluative and descrip-
experimental investigation into the tive aspects. 1. Pers. SOC.Psychol., 10,
formation of intergroup representations. 290-300.
Eur. 1. SOC.Psychol., 2, 202-204. Rabbie, J. M., and Horwitz, M. (1969),
- and Sinclair, A. (1973), The categoriza- Arousal of ingroup-outgroup bias by a
tion process in intergroup relations. chance win or loss. 1. Pers. SOC.
Eur. J. SOC.Psychol., 3, 145-157. Psychol., 13, 269-277.
Ferguson, C. K., and Kelley, H.H.(1964), - and Wilkens, 0. (1971), Intergroup
Significant factors in overevaluations of competition and its effects on intra-
own group’s product. J. abn. SOC. group and intergroup relations. Eur. J.
PSyChol., 69, 223-228. SOC. PSyChOl., 1 (2), 215-234.
Festinger, L. (1954), A theory of social Sherif, M. (1966), Group conflict and co-
comparison processes. Hum. Rela., 7, operation: Their social psychology.
117-140. London, Routledge & Kegan Paul.
Fishman, J. A. (1968), Nationality-national- Swingle, P. G. (1969), Ethnic factors in
ism and nation-nationism. In J. A. interpersonal bargaining. Can. J.
Fishman, C. A. Ferguson, and J. Gupta Psychol., 23, 136146.
(Eds.), Language problems of develop- Tajfel, H.(1970), Experiments in intergroup
ing nations. New York,Wiley. discrimination. Sc. Arner., 223 (3, 96
Harford, T., and Cutter, H. S. G. (1966). 102.
Co-operation among Negro and white - (1972), La CatCgorisation sociale. In
boys and girls. Psychol. Rep., 18, 818. S. Moscovici (Ed.), Zntroduction d la
Hatton, J. M. (1967,Reactions of Negroes psychologie sociale, vol. 1. Paris,
in a biracial bargaining situation. J. Larousse. Ch. 8, pp. 272-302.
Pers. SOC. Psychol., 7, 301-6. - Billig, M. G., Bundy, R. P.,and Fla-
Heller, I. R. (1966), The effects of racial ment, C. (1971), social categorization
prejudice, feedback strategy and race and intergroup behaviour. Eur. J. SOC.
on cooperativecompetitive behavior. Psychol., 1, 149-177.
Unpublished doctoral dissertation. - and Billig, M. G. (1974),Familiarity and
Univ. of California, Berkeley. categorization in intergroup behaviour.
Kelley, H. H. (1951), Communication in J. exp. SOC.Psychol., 10, 159-170.
experimentally created hierarchies. Thibaut, J. (1950), An experimental study
Hum. Rela., 4, 39-56. of the cohesiveness of underpriviIeged
Kelman, H. C. (1969), Patterns of personal groups. Hum. Rela., 3, 251-178.
involvement in the national system: A Turner, J. C. (1972), Competition between
social-psychological analysis of politid self and others, with and without
34 John C. Turner

categorization into groups. University games between opponents of the same

of Bristol. Ms. or of a different race. J. Pers. SOC.
- (1973), Competition and category-con- Psychol., 9, 24-30.
- and Miller, N. (1960), Shifts in evalua-
flict: Self versus group for social value
versus economic gain. University of tions of participants following inter-
Bristol. Ms. group competition. J . a h . SOC. Psychol.,
Uejio, C. K.,and Wrightsman, C. S. (1967), 63, 428-431.
Ethnic group differences in the re- Wrightsman, L. S. et al. (1967), Effects of
lationship of trusting attitudes to co- other person’s race and strategy upon
operative behaviour. Psychol. Rep., 20, co-operative behaviour in a P.D. game.
563-571. Paper presented at meeting of Mid-
Wilson, W., and Kayatani, M. (1968), western Psychological Association,
Intergroup attitudes and strategies in Chicago.

RLsumk Zusammenf assung

De rCcentes Ctudes ont indiquC que la Neuere Untersuchungen haben ergeben, daB
variable de catkgorisation sociale par elle- die Variable der sozialen Kategorisierung
meme suffit ?i l’apparition de discrimination an sich schon ausreicht fur die Erklarung
entre groupes. De cette relation nouvelle- der Diskrimination zwischen Gruppen. In
ment mise-A-jour, cet article prCsente une diesem Aufsatz werden diese Befunde er-
explication B partir de l’action de processus kvart als Wirksamwerden von sozialen Ver-
de comparaison sociale entre groupes fon- gleichsprozessen zwischen Gruppen, die auf
dCe sur le besoin des groupes d’une identiti dem Bediirfnis einer positiven Eigengrup-
positive interne. La relation entretenue entre penidentitat aufbauen. Die Beziehung
identit6 sociale percpe et comparaison inter- zwischen wahrgenommener sozialer Identi-
groupe fait l’objet d’un traitement thCorique tat und dem Vergleich zwischen Gruppen
et il est soutenu que les comparaisons socia- wird theoretisch ausgearbeitet und es wird
les donnent naissance aux processus d’une behauptet, da8 soziale Vergleiche Prozesse
diffkrentiation mutuelle entre groupes qui der gegenseitigen Differenzierung zwischen
peut Stre analysCe comme une forme de Gruppen auslosen, die d a m als eine Form
compCtition ‘sociale’. La comp6tition des sozialen Wettbewerbs analysiert werden
sociale se distingue de la compCtition ‘uti- konnen. Sozialer Wettbewerb wird vom
litaire’ (conflits d’intCr8ts de .groupes). De realistischen Wettbewerb unterschieden
nouvelles donnCes sont introduites qui ren- (Konflikt von Gruppeninteressen). Es wer-
forcent cette interprktation des Ctudes de den neue Daten angefiihrt, die fiir diese
catCgorisation ‘minimale’. On a dCcouvert Interpretation der Untersuchungen iiber
que la discrimination intergroupe minimale ‘minimale’ Kategorisierung sprechen. Es
prend place dans la distribution de ‘points’ l i i t sich nachweisen, daB minimale Diskri-
sans signification aussi bien que de rCcom- mination zwischen den Gruppen sowohl bei
penses monCtaires et que la cat6gorisation der Vergabe von Punkten ohne praktischen
par elle-meme ne conduit pas ?i un com- Wert wie bei finanziellen Belohnungen
portement intergroupe ob les sujets puissent stattfiidet, und daB die soziale Kategori-
agir directement comme tels. D’autres sierung ‘an sich’ kein Zwischengruppenver-
etudes sur les tendances du comportement halten auftreten I i t , bei dem die Individuen
intergroupe sont passhes en revue pour unmittelbar fiir sich selbst agieren konnen.
montrer la gCnCralit6 de la comp6tition Andere Untersuchungen iiber Eistellungs-
sociale dans les situations faisant intervenir verzerrungen zwischen Gruppen werden
plusieurs groupes. dargestellt, um das allgemeine Auftreten des
sozialen Wettbewerbs in Situationen mit
mehreren Gruppen zu belegen.