Self-categorisation, commitment to the group and group self-esteem as related but distinct aspects of social identity

The aim of this study is to show that, when examining social identi®cation, it is both possible and important to distinguish between self-categorisation, commitment to the group, and group self-esteem, as related but separate aspects of group members' social identity. This was demonstrated in an experiment (Nˆ 119), in which Ingroup Status (high/low), Ingroup Size (majority/minority), and Group Formation (self-selected/ assigned group membership) were manipulated orthogonally. The results of this study con®rm that these three aspects of social identity can be distinguished as separate factors in a principal components analysis. Furthermore, as predicted, the three aspects are di€erentially related to manipulated group features, as well as displays of ingroup favouritism. Group members' self-categorisations were only a€ected by the relative size of the group, while group self-esteem was only in ̄uenced by group status. A€ective commitment to the group depended both on group status and on the group assignment criterion. Importantly, only the group commitment aspect of social identity mediated displays of ingroup favouritism. Copyright # 1999 John Wiley & Sons, Ltd. Although social identi®cation plays a key role in social identity theory, relatively little attention has been devoted to the question of how exactly this concept should be de®ned theoretically, or how it can be measured empirically. Consequently, investigations of social identity-related processes have often not systematically included social identi®cation as a dependent measure, have only measured it indirectly, or merely inferred social identi®cation from other responses such as intergroup di€erentiation. The main goal of this investigation is to take a closer look at the conceptualisation and measurement of social identi®cation, by trying to distinguish between di€erent aspects of social identi®cation, relating these to speci®c group features, and investigating them as mediators of social behaviour. CCC 0046±2772/99/020371±19$17.50 Received 13 October 1997 Copyright # 1999 John Wiley & Sons, Ltd. Accepted 3 March 1998 European Journal of Social Psychology Eur. J. Soc. Psychol. 29, 371±389 (1999) *Correspondence to: Dr Naomi Ellemers, Social Psychology, Free University, Van der Boechorststraat 1, 1081 BT Amsterdam, The Netherlands. e-mail: [email protected] At a theoretical level, the de®nition proposed by Tajfel (1978), which is most commonly cited, maintains that social identity is `. . . that part of an individual's selfconcept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional signi®cance attached to that membership.' (p. 63). On the basis of this de®nition, it is assumed that three components may contribute to one's social identity: a cognitive component (a cognitive awareness of one's membership in a social groupÐself-categorisation), an evaluative component (a positive or negative value connotation attached to this group membershipÐgroup self-esteem), and an emotional component (a sense of emotional involvement with the groupÐa€ective commitment). The measurement scales that were developed to tap identi®cation with social group re ̄ect this common conception, as they all seem to incorporate the three components proposed in Tajfel's de®nition (see Ellemers & Mlicki, unpublished manuscript, for an overview). However, social identi®cation is usually treated as a unidimensional construct (cf. Ellemers, 1991). A notable exception is a study by Hinkle, Taylor, Fox-Cardamone and Crook (1989), who distinguish three factors in the group identi®cation scale developed by Brown, Condor, Mathew, Wade and Williams (1986). Although Hinkle et al. (1989) argue in favour of a multi-component conceptualisation of group identi®cation, the components they distinguish show substantial intercorrelations (between 0.43 and 0.58), which seems to have been taken as an indication that a common treatment as one factor would be acceptable for practical purposes. More importantly, this imprecision at the operational level is often re ̄ected in conceptual treatments of social identity, and has resulted in a considerable amount of theoretical confusion. Empirical investigations of social identity have often studied how people respond to arti®cially constructed or so-called minimal groups in experimental research paradigms (see Brewer, 1979). Although it was demonstrated with this method that mere categorisation can be sucient to induce people to behave in terms of their group membership (cf. Tajfel, Billig, Bundy, & Flament, 1971), this is not necessarily the case for members of natural groups. Indeed, it may be argued that in the minimal group paradigm, the categorisation into di€erent groups provides the only way for people to make sense of the experimental situation. Hence, group aliation is the only social cue that may be used to guide one's behaviour towards other participants (see also Jetten, Spears, &Manstead, 1996). However, in more natural social contexts, people who acknowledge that they belong to a particular social category do not necessarily feel committed to that group, or behave in terms of that group membership. Instead, they might prefer to belong to another group, or simply be indi€erent to this particular categorisation. Accordingly, it has been pointed out that seemingly robust experimental phenomena such as the display of ingroup favouritism were not consistently found as a consequence of mere categorisation into natural groups (cf. Mullen, Brown, & Smith, 1992). The key proposal of social identity theory, however, is that it is the extent to which people identify with a particular social group that determines their inclination to behave in terms of their group membership. In this sense, social identi®cation is primarily used to refer to a feeling of a€ective commitment to the group (i.e. the emotional component), rather than the possibility to distinguish between members of di€erent social categories (the cognitive component). Therefore, as a ®rst step we think it is important to distinguish cognitive awareness of one's group membership per se (self-categorisation) from the extent to which one feels emotionally involved 372 N. Ellemers et al. Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 371±389 (1999) with the group in question (a€ective commitment). Indeed, there is recent empirical evidence that people who belong to the same social group may show di€erential responses, depending on the extent to which they feel a€ectively committed to that group (cf. Branscombe & Wann, 1994; Doosje, Ellemers, & Spears, 1995; Ellemers & Van Rijswijk, 1997). Accordingly, it has also been demonstrated that self-stereotyping (denoting a cognitive awareness of one's group membership) can be distinguished from a€ective commitment to the group (see Spears, Doosje, & Ellemers, 1997) at the measurement level, in the sense that they emerged as two separate clusters of items in a principal components analysis. A second respect in which we would like to specify our conceptualisation of ingroup identi®cation is by distinguishing the extent to which people feel emotionally involved with their group (a€ective commitment) from the value connotation of that particular group membership (group self-esteem). It has repeatedly been argued (cf. Tajfel & Turner, 1979) and demonstrated (cf, Ellemers, 1993) that the two often covary, in the sense that a€ective commitment tends to be stronger in more positively evaluated groups (because these groups may contribute more to a positive social identity), while people are inclined to distance themselves from less attractive groups. However, and again this can be seen most clearly in the case of natural groups (when leaving the group does not constitute an easy or attractive option), this does not imply that the two necessarily go together, or that these concepts can be used interchangeably. Indeed, recent empirical evidence clearly reveals that, provided their identity as members of a distinct social group is suciently important, people may show signs of strong emotional involvement while simultaneously acknowledging or even emphasizing the negative characteristics of their group (see Mlicki & Ellemers, 1996). In terms of our conceptual analysis, this implies that self-categorisation (the cognitive component) as well as a€ective commitment to a speci®c group (the emotional component) can be distinguished from group self-esteem derived from the value connotation of that particular group membership (the evaluative component). More importantly, we want to argue that this distinction should be made, to be able to understand how they are a€ected di€erentially by speci®c characteristics of the group or the social context. Indeed, on the basis of previous theory and research it is possible to hypothesize which group characteristics are most likely to a€ect the three di€erent components of identi®cation. Furthermore, we predict that these components are di€erentially related to displays of ingroup favouritism in evaluative responses or outcome allocations. Relative status can be considered a central group characteristic in both theory and research on social identity and intergroup relations. The general argument is that a low group status position results in unfavourable comparisons between the ingroup and relevant other groups, which may frustrate attempts to derive a positive social identity from one's group membership. As a result, members of lower status groups are expected to show less social identi®cation than members of groups with higher status. Indeed, empirical investigations have con®rmed that ingroup identi®cation is generally less in lower status groups than in groups with high status (Ellemers, Van Knippenberg, De Vries, & Wilke, 1988; Ellemers, Van Knippenberg, & Wilke, We have opted for the term `a€ective' commitment because we aim to refer to a sense of emotional involvement in the group, rather than the commitment that stems from interdependence or normative considerations (see Allen & Meyer, 1990). Self-categorisation 373 Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 371±389 (1999) 1990; Sachdev & Bourhis, 1987). However, in line with our previous analysis, we want to argue that it is mainly the evaluative component of social identity (group selfesteem), that is a€ected by relative group status. If this were indeed the case, this would also help us understand inconsistent empirical ®ndings with respect to the socalled self-esteem hypothesis in social identity theory (see Hinkle & Brown, 1990; Hogg & Abrams, 1990; Long & Spears, 1997). Essentially, the contradictory issue seems to be that, on the one hand, members of lower-status groups are expected to show ingroup favouritism as a means to boost their social identity. On the other hand, to the extent that the current low status of their group results in low ingroup identi®cation, this should preclude them from undertaking such group-level behaviour. However, if we assume that low group status negatively a€ects the evaluative component of identi®cation only, while the level of a€ective commitment (the emotional component) can remain unchanged, it becomes clear that it is the combination of a threat to group self-esteem and strong a€ective commitment which should elicit attempts to depict the ingroup in a positive way. A second important issue which has generated a substantial amount of empirical research concerns the e€ects of relative group size on ingroup identi®cation and ingroup favouritism (e.g. Gerard & Hoyt, 1974; Mummendey & Simon, 1989; Sachdev & Bourhis, 1984; Simon & Brown, 1987). Again, to date, the precise nature of these e€ects has not been established unambiguously. We have argued elsewhere that (seemingly) inconsistent results may have been obtained in previous research because minority versus majority group membership has often been used to refer to di€erential status as well as di€erential group size (see Ellemers & Van Rijswijk, 1997). However, when the two are disentangled, it turns out that minority group size results in stronger ingroup identi®cation than majority group size. Other than group status e€ects, it seems that this is not due to the fact that membership in a minority group is more attractive but because it is more salient or distinctive than majority group membership (see McGuire, McGuire, Child, & Fujioka, 1978; McGuire & Padawer-Singer, 1976). Simon and Hamilton (1994) indicate that membership in a small or distinctive group implies a relatively large overlap between the collective self and the individual self. In a similar vein, from an optimal distinctiveness perspective (see Brewer, 1991) it would seem that, compared to inclusion in a majority group, minority group membership o€ers a better opportunity to balance the need to retain some sense of individuality with the need to belong to a group, which should result in a greater readiness to perceive or de®ne oneself as a group member. Accordingly, we would argue that it is mainly the cognitive component or self-categorisation aspect of ingroup identi®cation that is a€ected by relative ingroup size. Finally, we aim to identify group characteristics which primarily in ̄uence the emotional component of ingroup identi®cation, that is, the extent to which people feel a€ectively committed to a particular group. In our view, this component is essential as this is supposed to constitute the main determinant of individual-level (such as distancing oneself from the group) versus group-level responses (e.g. displays of ingroup favouritism) to a common identity threat (see also Ellemers, Spears, & Doosje, 1997). In relation to this issue, we argue that a fundamental distinction can be made between assigned versus achieved (or self-selected) group memberships. Although this distinction has been noted in the literature (e.g. Luhtanen & Crocker, 1991), neither in theoretical accounts nor in empirical work has systematic attention been devoted to possible di€erential responses to membership in these two kinds of 374 N. Ellemers et al. Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 371±389 (1999) groups. Nevertheless, from the original formulations of social identity theory (Tajfel, 1978; Tajfel & Turner, 1979) as well as from previous empirical work we may infer that the nature of one's group membership is likely to make a di€erence. In laboratory research (Ellemers, Wilke, & Van Knippenberg, 1993) it has been established that individuals who feel that their inclusion in a lower-status group is unjust (but are nevertheless assigned to this group by the experimenter) are likely to compete with their fellow ingroup members in order to leave this group, indicating relatively little group commitment. In contrast, to the extent that people have voluntarily committed themselves to membership in a particular group, they are more inclined to show group solidarity, even when the group turns out to be unsuccessful (see Turner, Hogg, Turner, & Smith, 1984). In a similar vein, Cio and Garner (1996) have demonstrated that people who have actively applied for membership in a particular group aremost likely to behave in accordance with their groupmembership. Indeed, the observation that systematically di€erent results are obtained with research among natural versus arti®cial groups (see Mullen et al., 1992), may have to be attributed to the fact that arti®cial groups are usually created by assigning people to a particular experimental group, while a classi®cation of research participants into natural groups (e.g. according to their study major, university town, or political aliation) is generally more likely to involve self-selected group memberships. Accordingly, recent empirical evidence has demonstrated that di€erent results obtained with these two kinds of research paradigms can at least to some extent be ascribed to the fact that the level of ingroup identi®cation tends to be higher as a result of natural compared to arti®cially created group memberships (Jetten et al., 1996). We want to argue that this di€erential group commitment essentially occurs because of the basis on which people are included in a particular group (i.e. assigned versus self-selected group memberships). Although in practice this distinction may covary with the distinction between arti®cial and natural groups as we have argued above, in our view this points to an empirical confound, rather than to a conceptually necessary or inherent combination of features of these two kinds of groups. Indeed, membership in some natural groups (such as gender groups or ethnic groups) is assigned, rather than self-selected, while arti®cial groups for laboratory investigations can also be created by letting participants choose or earn membership in a particular group (cf. Wright, Taylor, & Moghaddam, 1990). In this study we will therefore systematically vary the way in which participants are included in one of two arti®cially created groups by either having the experimenter decide (assigned group membership) or letting people indicate themselves to which of two groups they think they belong (self-selected group membership). To summarise the above argument, it seems both necessary and informative to distinguish between di€erent components of social identity as possible responses to group membership and group features. The ®rst aim of this study therefore is to demonstrate that this is not only a possible conceptual distinction but that it can also be made empirically. For this purpose, we will investigate how the three aspects of social identity are di€erentially a€ected by important group features, notably the relative status and size of the group, and the basis of group formation (i.e. assigned or self-selected group membership). Furthermore, we aim to assess whether the three components of social identity play a di€erent role as mediators of grouplevel behaviour. Speci®cally, we predict that it is essentially a sense of emotional involvement with the group (a€ective commitment), rather than the cognitive Self-categorisation 375 Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 371±389 (1999) (self-categorisation) or evaluative (group self-esteem) component of ingroup identi®cation which predisposes people to show ingroup favouritism. In order to investigate this, we manipulated the group assignment criterion (self-selected versus assigned group membership), the relative size (minority versus majority) and status (high versus low) of arti®cially created groups. In addition to measuring the three components of social identity, we also included separate measures of personal selfesteem and personal identity (a personal-level equivalent of commitment to the group seems irrelevant), to investigate under what circumstances these measures at the personal level show opposite or parallel results to the group-level measures (Turner, 1985).