67 least contributes to, or increases, the intensity of ingroup affiliations. The existence or strength of the ingroup are thus seen as phenomena derived from the relations between the ingroup and its outgroups. In some cases, this consists of presumed intragroup effects of various kinds of direct or projected outgroup hostility in others, it is seen as a direct result of an &dquo objective&dquo conflict of interests between the groups. The emphasis remains the same. There is no doubt that this outgroup-ingroup sequence of attitudes and beha- viour has a great deal of validity, both intuitively and as a result of a great mass of empirical evidence. But the emphasis - understandable as it has been, mainly for social reasons - is too one-sided. An adequate social psy- chological theory of intergroup behaviour must take into account both causal directions: from ingroup processes to outgroup behaviour and attitudes as well as the opposite one which has been until now the principal object of theory and research. Even if it is true that originally many groups are created as a common shelter for their members from outside threats and dangers (human or not), it is equally true that in any complex society an individual confronts from the beginning of his life a complex network of groupings which presents him with a network of relationships into which he must fit himself. One of the most important and durable problems that is posed to an individual by his insertion into society is to find, create and define his place in these networks. It is reasonable to assume that both his ingroup and outgroup attitudes and behaviour must be determined, to some extent at least, by this continuing pro- cess of self-definition. An early version of these ideas led to the first experiments we conducted in Bristol some three or four years ago (Tajfel, 1970a Tajfel et al., 1971). Their aim was to establish minimal conditions in which an individual will, in his behaviour, distinguish between an ingroup and an outgroup. In order to create such minimal conditions we attempted to eliminate from the experimen- tal situations all the variables that normally lead to ingroup favouritism or discrimination against the outgroup: face-to-face interaction conflict of inte- rests any possibility of previous hostility any utilitarian or instrumental link between the subjects��� responses and their self-interest. In addition, we en- abled the subjects to chose amongst a variety of strategies in their responses, some of which were more &dquo rational&dquo or &dquo useful&dquo than responses which would have created a differentiation between the groups. The subjects first perfor- med a relatively trivial task (guessing numbers of dots in rapidly projected clusters, or expressing preference for the paintings of one of two fairly abstract painters, Klee and Kandinsky). They then worked separately in individual cubicles. Their task was to decide (on a number of payment matrices) how points, worth money, should be divided between two other subjects. They knew what was their own group membership (under- or over-estimation of dots or preference for one or the other painter), and the group membership of those between whom they were dividing the money but these individuals were designated by code numbers, and their identity was unknown. The by guest on July 22, 2009 http://ssi.sagepub.com Downloaded from

68 results were very highly significant in the direction of awarding more money to members of the ingroup. In the second set of experiments, the matrices were so constructed that we could assess the separate &dquo pull&dquo of several varia- bles on the decisions. These variables were: maximum joint profit (i.e. the strategy of awarding the maximum joint amount on each matrix, so that all the subjects together - who knew each other well before the experiments - could get the greatest possible amount of money out of the experimenters) maximum profit for members of the ingroup maximum difference in favour of the ingroup at the price of sacrificing both the above advantages and fair- ness of choices. Of these variables, the first - maximum joint profit - exerted hardly any pull on the decisions maximum ingroup profit was important, but sometimes not nearly as important as achieving maximum difference in favour of the ingroup. Fairness was also a significant variable and served to mode- rate the excesses of ingroup favouritism. Two simple and overlapping explanations are available to account for these results: a &dquo normative&dquo one and a &dquo learning&dquo one. The first is that our school- boy subjects, aged 15 to 16 years, saw the situation as one of &dquo team competi- tion&dquo in which one should make one���s own team win at whatever cost. The second, that - in a new situation - they engaged in ingroup behaviour which had been reinforced on countless occasions in the past. While both these explanations are sensible they are also quite uninteresting because not genui- nely heuristic. If our subjects had chosen strategies of choices leading to maximum joint profit, the same explanations could still serve, in one form or another. If they had chosen only the strategy of fairness without that of ingroup favouritism, one could still &dquo explain&dquo their responses starting from norms and previous reinforcements. My argument is not that these explana- tions are invalid. It is rather that, in addition to their capacity to explain indiscriminately all kinds of results, they are at a level of generality which prevents them from serving as a point of departure for new and more searching insights about intergroup processes. It is the choice by the subjects of these particular norms based on these particular reinforcements which defines the problem and provides a point of departure for some research questions about the psychology of intergroup relations. This is so particularly in view of the fact that the results have since been replicated in several experiments both in Britain (Billig, 1972 Billig and Tajfel, 1973 Tajfel and Billig, 1974 Turner, 1973) and elsewhere (Deutsch et al., 1971 Doise et al., 1972 Sole et al., 1973). The problems of an individual���s self-definition in a social context, briefly mentioned above, can be restated in terms of the notion of social identity. We need to postulate that, at least in our kinds of societies, an individual strives to achieve a satisfactory concept or image of himself. This was one of the bases of Festinger���s (1954) theory of social comparison. Festinger, however, was almost exclusively concerned with social comparisons made between indi- viduals and with evaluations of oneself and others made by means of these by guest on July 22, 2009 http://ssi.sagepub.com Downloaded from

69 inter-individual comparisons. This inter-individual emphasis neglects an important contributing aspect of an individual���s self-definition, namely that he is a member of numerous social groups and that this membership contributes, positively or negatively, to the image that he has of himself. Four linked concepts will be employed in order to proceed with this dis- cussion. They are: social categorization, social identity, social comparison and psychological distinctiveness. The process of categorization, as it is used by the human individual in order to systematise and simplify his environment, presents certain theoretical con- tinuities between the role played by categorizing in perceptual activities and its role in the ordering of one���s social environment. For our purpose, social categorization can be understood as the ordering of social environment in terms of social categories, that is of groupings of persons in a manner which is meaningful to the subject. Therefore, in our discussion the term &dquo group&dquo denotes a cognitive entity that is meaningful to the subject at a particular point in time and must be distinguished from the way in which the term &dquo group&dquo is used in much of the social psychological literature where it denotes an &dquo ob- jective&dquo (most often face-to-face) relationship between a number of people. In other words, social categorization is a process of bringing together social objects or events in groups which are equivalent with regard to an individual���s actions, intentions, attitudes and systems of beliefs. The second concept we must introduce here is that of social identity. For our purposes we shall understand social identity as that part of an individual���s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the emotional significance attached to that membership. Social categorization must therefore be considered as a system of orientation which creates and defines the individual���s own place in society. As Berger (1966) has written: &dquo Every society contains a repertoire of identities that is part of the objective knowledge of its members&dquo (p. 106). &dquo Society not only defines but creates psychological reality. The individual realises himself in society - that is, he recognizes his identity in socially defined terms and these definitions become reality as he lives in society&dquo (p. 107). Several consequences regarding group membership follow upon this &dquo recog- nition of identity in socially defined terms&dquo . They can be described as follows: 1) 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 contri- bution to make to the positive aspects of his social identity i.e. to those aspects of it from which he derives some satisfaction. 2) If a group does not satisfy this requirement, the individual will tend to leave it unless a) leaving the group is impossible for some &dquo objective&dquo reasons or, b) it conflicts with important values which are themselves a part of his accep- table social identity. _ by guest on July 22, 2009 http://ssi.sagepub.com Downloaded from

70 3) If leaving the group presents the difficulties just mentioned, then at least two solutions are possible: a) to change one���s interpretation of the attributes of the group so that its unwelcome features (e.g. low status) are either justified or made acceptable through a reinterpretation b) to accept the situation for what it is and engage in social action which would lead to desirable changes in the situation (of course, there may be various combinations of a) and b), such as, for example, when the negative attributes are justified and social action to remove them is undertaken at the same time). 4) No group lives alone - all groups in society live in the midst of other groups. In other words, the &dquo positive aspects of social identity&dquo in 1) above, and the reinterpretation of attributes and engagement in social action in 3) above, only acquire meaning in relation to, or in comparisons with, other groups. It is this comparative perspective that links social categorizing with social identity. In his theory of social comparison processes, Festinger (1954) hypo- thesized that &dquo there exists, in the human organism, a drive to evaluate his opinions and his abilities&dquo . His second major hypothesis in the same paper was that &dquo to the extent objective, non-social means are not available, people evaluate their opinions and abilities by comparison respectively with the opi- nions and abilities of others&dquo . But there are some difficulties with respect to the conception that social comparisons only take place &dquo to the extent that objec- tive, non-social means are not available&dquo . Festinger���s example is that &dquo one could, of course, test the opinion that an object was fragile by hitting it with a hammer&dquo . I can confirm the opinion that a bed is for lying-down-on by lying down on it until I discover that this particular bed in this particular room of the castle belonged to the Duke of Urbino and is most definitely not for lying- down-on. Very often, the &dquo objective non-social means&dquo that may appear to an observer to be available for the testing of opinions do not have much validity unless they are used in conjunction with the significance that they acquire in their social setting. The cases which lie outside this range are usually trivial in the analysis of social behaviour. In addition, social reality can be as &dquo objective&dquo as is non-social reality, and conversely &dquo objectivity&dquo can be as &dquo social&dquo as it is &dquo physical&dquo . In some cultures, thunder and lightning are as indisputably signs of anger of supernatural powers as they are bursts of sound and light. The criterion of &dquo objectivity&dquo cannot be based on classifying phenomena as being of a &dquo social &dquo or a &dquo non-social&dquo nature, with the presumed attendant consequence that opinions about them can be tested respectively by &dquo social&dquo or by &dquo non-social&dquo means. It can instead be defined in terms of the awareness (or the degree of subjective probability) that there exist alternatives to the judgement one is making. A low (or nil) probability that alternatives to one���s opinions exist may be due to the consistency over time in the checking of these opinions through non-social means, as in Festinger���s example of fra- gility and hammer but it may also be due to the very high social consensus by guest on July 22, 2009 http://ssi.sagepub.com Downloaded from

71 about the nature of a phenomenon, independently of whether the phenome- non is thought of as being &dquo physical&dquo , &dquo natural&dquo or &dquo social&dquo . It is undoubt- edly true that certainty can very often be more easily reached about the physi- cal than about the social means of testing, but this is not a theoretical distinc- tion between what appears and does not appear as &dquo objective reality&dquo . It cannot be said that a human organism turns towards social means of validating opinions only when non-social means for doing so are not available. There are many examples, both in the history of science in our own culture and in the systems of knowledge of other cultures, of procedures which follow the opposite course i.e. they do not use the &dquo physical&dquo means of testing which are, in principle, available because of the very high (or complete) social consen- sus about the nature of a phenomenon. Therefore, &dquo social comparison processes&dquo have an even wider range of appli- cation than Festinger was willing to assign to them. The range of application includes both the social context (or significance) of &dquo non-social&dquo testing, and the cases where the high social consensus about the nature of a phenomenon is sufficient to confer the mark of &dquo objectivity&dquo on opinions about it. In his theory, Festinger (1954) was mainly concerned with the social testing of opi- nions about characteristics of individuals, and with the resulting &dquo relative simi- larity in opinions and abilities among persons who associate with one another (at least on those opinions and abilities which are relevant to that association)&dquo . The theory was primarily addressed at the within-group effects of the process of social comparison (such as pressures towards uniformity in a group) while &dquo comparisons with members of a different status group, either higher or lower may sometimes be made on a phantasy level, but very rarely in reality&dquo . Though Festinger qualifies this statement by adding that comparisons between groups that differ are not completely eliminated, the focus of his discussion remains on individuals comparing themselves with other individuals. On the basis of our discussion so far, we are now able to make some general statements about social categorization into groups in relation to its function &dquo as a system of orientation which creates and defines the individual���s own place in society&dquo . The first concerns the &dquo objective reality&dquo of comparisons focusing on an individual as an individual and comparisons based on an indi- vidual���s membership of a particular social group. With regard to the first issue, it can be said that the only &dquo reality&dquo tests that matter with regard to group characteristics are tests of social reality. 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 connotation of these differences. For example, economic deprivation acquires its importance in social attitudes, intentions and actions mainly when it becomes &dquo relative depri- vation&dquo easy or difficulty access to means of production and consumption of goods, to benefits and opportunities, becomes psychologically salient mainly in relation to comparisons with other groups the definition of a group (natio- by guest on July 22, 2009 http://ssi.sagepub.com Downloaded from