The Issue of Self and Identity Is One of Crucial Importance Within Social Psychology (For
The self as determinant of ingroup favoritism
Running head: The self as determinant of ingroup favoritism
When “I” turns into “we”:
The self as a determinant of favoritism towards novel ingroups
Sabine Otten
Department of Social Psychology,
Friedrich-Schiller-University Jena, FRG
Key words: ingroup favoritism; positive ingroup default; self-anchoring; social
discrimination; social identity
Abstract
Initially, Social Identity Theory (SIT) seemed to have provided a straightforward concept how the self relates to ingroup favoritism: Based upon the distinction between personal and social identity, it posits a striving for positive ingroup distinctiveness as a means to enhance the self-concept. This reasoning was crucially influenced by findings in the Minimal Group Paradigm, demonstrating that mere categorization into two social groups can suffice to elicit favoritism even towards anonymous ingroup members. Although this effect was replicated in numerous studies, its interpretation in the framework of SIT has been questioned. In this context, my own empirical program proposes a link between self-concept and positive ingroup distinctiveness that relies on cognitive rather than motivational processes. Evidence is presented demonstrating that ingroup favoritism does not necessitate an explicit social comparison between ingroup and outgroup, and that the self can function as a heuristic to give meaning to novel ingroups. Hence, these findings suggest that – under certain conditions of the intergroup context – ingroup favoritism can rely on an intra- rather than an intergroup process.
(172 words)
When “I” turns into “we”: The self as a determinant of favoritism towards novel ingroups
The issue of self and identity is of crucial importance within social psychology (for a recent survey see Baumeister, 1998). As Turner and Oakes (1997, p. 365) put it: “The nature of the self and its relationship to social cognition is the theoretical core of social psychology“.
Hence, it is not surprising that this theme plays a central role in influential theories on intergroup behaviour, especially in Social Identity Theory (SIT; Tajfel & Turner, 1979; 1986) and Self-Categorization Theory (SCT; Turner, Hogg, Oakes, Reicher & Wetherell, 1987). In both theories, the notion of a social identity, defined by group memberships and the evaluations and emotions attached to these membership is pivotal (see e.g. Tajfel, 1978; Tajfel & Turner, 1986; Turner, 1999). However, whereas SIT is basically concentrating on the prediction of intergroup relations and the phenomenon of ingroup favoritism (see Brown, 2000a), SCT’s core issue is the more general question of the conditions and the (cognitive) consequences that arise when people categorize themselves as group-members with shared characteristics rather than as unique individuals. While SIT focuses rather on the motivational factors affecting group members’ perception and behavior, SCT focuses on social identity as “the cognitive mechanism that makes group behaviour possible” (Turner, 1984, p. 527). The basic goal of the empirical program presented in this chapter was to investigate basic determinants of ingroup favoritism. Thus, SIT rather than SCT was the reference point from which the research question was derived.
Although much broader in its scope, SIT is closely associated with experiments conducted in the so-called minimal group paradigm (MGP; Tajfel, Billig, Bundy & Flament; see also Rabbie & Horwitz, 1969). This paradigm was originally planned to provide a baseline in order to test subsequently the necessary and sufficient conditions for ingroup favoritism and outgroup derogation. Hence, intergroup behaviour was analyzed in a situation of ‘mere categorization‘: Participants were anonymously assigned to one of two distinct, novel social categories; there was neither intra- or intergroup interaction, nor an opportunity to directly fulfil self-interests by allocations or evaluations, nor was there a functional relation between the categorization dimension, on the one hand, and the evaluation or allocation dimension on the other hand. However, the meanwhile commonplace and often replicated finding (see e.g. Brewer, 1979; Brewer & Brown, 1998; Brown 2000) was that already under these restricted conditions ingroup favoritism occurred.
SIT offers an explanation for the so-called ‘mere categorization effect‘ by postulating a need for positive social identity (see Brown, 2000; Tajfel & Turner, 1986; Turner, 1999). It is assumed that the self-concept comprises two components, personal and social identity. By treating or evaluating ingroup members more favorably than outgroup members, social identity can be ensured or enhanced. Thus, establishing positive ingroup distinctiveness serves the general motive of self-enhancement (see Sedikides, 1993; Sedikides & Strube, 1995). Accordingly, it is “presumed to be a causal connection between intergroup differentiation ... and self-esteem” (Brown, 2000, p. 755; see also Abrams & Hogg, 1988; Hogg & Abrams, 1990). Turner (1999) summarizes this assumption as follows:
“The basic hypothesis, which is at the psychological heart of the theory, is the notion that social comparisons between groups relevant to an evaluation of social identity produce pressures for intergroup differentiation to achieve a positive self evaluation in terms of that identity” (p. 18).
In fact, there is some evidence indicating that exerting relative ingroup bias increases self-esteem (Lemyre & Smith, 1985; Oakes & Turner, 1980); however, the findings are not unequivocal, and especially for the complementary derivation, threatened self-esteem as predictor of ingroup favoritism, the findings are not at all convincing (see Crocker & Schwartz, 1985; Crocker, Thompson, McGraw & Ingerman, 1987; Hogg & Abrams, 1990; Rubin & Hewstone, 1998). Hence, it is not surprising that notwithstanding the many successful replications of ingroup favoritism in the minimal group paradigm, its interpretation in terms of a striving for positive social identity is still – or rather increasingly -- controversial (e.g. Cadinu & Rothbart, 1996; Diehl, 1989; Gaertner & Insko, 2000; Hogg & Abrams, 1990; Messick & Mackie, 1989; Mummendey, 1995). According to Brown (2000) this discussion raises “the possibility of recognising a wider range of cognitive motives associated with social identification than those specified by SIT”(p. 756).
My own research program can be understood as a contribution to such discussion; however, rather than focusing on alternative motivational accounts, the present approach concentrates on basic cognitive processes that might account for ingroup favoritism even towards anonymous fellow group members in a minimal intergroup situation. Starting from the observation that a favorable attitude towards novel ingroups does not necessitate explicit social comparisons with the outgroup, a model is proposed, in which the self is not seen as immediately motivating and profiting from ingroup favoritism, but rather as a source of information from the definition of novel own groups can be derived.
More specifically, a research program will be presented that provided evidence in basically four domains:
a)Positive ingroup default: A set of study demonstrates that there is evaluative favoritism towards novel ingroups immediately after social categorization, and without any explicit opportunities to compare with an outgroup.
b)Implicit associations between ingroup and self: Response-time evidence implies that there is an overlap between the mental representations of self and ingroup; in case of ambiguous group judgments, ingroup evaluations are facilitated when matching self ratings.
c)Self-anchoring in explicit measures: There is evidence that – at least under certain contextual conditions – group mebers tend to assimilate the ingroup definition to the self definition in explicit evaluations. There is evidence that this process has a stronger impact than the tendency to differentiate the ingroup positively from the outgroup.
d)Self as heuristic: In a further study it is demonstrated that the process of self-anchoring is determined by the judges’ motivation and ability to use heuristics as a means of impression formation on their novel ingroup.
Positive ingroup default
According to SIT, social comparison with the outgroup is a decisive element in the process by which social categorization can turn into the creation of positive ingroup distinctiveness. Thus, the question is whether individuals that have been assigned to a novel social category will express a biased ingroup evaluation already before having engaged in explicit social comparisons (as typically requested in experiments on intergroup allocations and evaluations). In fact, Maass and Schaller (1991) argue that there is an “initial categorization-based ingroup bias” such that "group members seem to approach their task with the rudimentary hypothesis that their own group is better than the opposing group" (p. 204).
In the present context, crucial issues were: Can such positive ingroup default be demonstrated even for minimal groups? And: How can such initial bias be disentangled from bias that is not the starting point but rather the result of a comparison with the corresponding outgroup? Here, paradigms that study judgmental processes on an implicit level seemed most appropriate. A series of experiments on implicit intergroup bias was conducted by Perdue, Dovidio, Gurtman and Tyler (1990), who demonstrated in learning tasks and lexical decision tasks that (subliminal) global reference to either own groups or other groups tasks primed positive affect in the former, but rather neutral affect in the latter condition. In sum, their findings imply that words relating to ingroup enhance the accessibility of positive trait information.
Perdue and collaborators (1990) did not refer to specific groups in their experiments; as ingroup designators they used terms like "we", "us", as outgroup designators terms like “they” and “them”. The authors themselves acknowledge that their experiments “do not demonstrate whether it is the in-group and outgroup terms themselves or whether it is the cognitively represented social entities that they signify that are source of these attitudinal biases” (p. 483). Whereas their findings can be read as evidence for intergroup bias on the implicit level, they do not parallel the minimal group paradigm, where the ingroup-outgroup distinction was conceptually unrelated to existing social schemata. Hence, their data can not demonstrate whether there is such thing as a ‘positive ingroup default’. However, such demonstration was provided in a series of studies by Otten and collaborators (Otten & Moskowitz, 2000; Otten & Wentura, 1999). Instead of using unspecific primes for ingroup and outgroup, Otten and Wentura (1999) used the labels of categories which they had introduced in a minimal categorization procedure immediately before the lexical decision task. Again, there was an affective priming effect for ingroup labels such that their preceding subliminal presentation facilitated the classification of positive as compared to negative traits, whereas no affective congruency effects emerged for outgroup primes. Moreover, there was a significant correlation between implicit positive ingroup attitudes (as indicated by responses in the lexical decision task) and explicit measures of ingroup preference.
Otten and Moskowitz (2000) provided evidence for an implicit bias towards minimal ingroups in a different paradigm. They combined a minimal categorization procedure with a probe task demonstrating spontaneous trait inferences (STIs; e.g. Uleman, Hon, Roman & Moskowitz, 1996). The idea was that reference to the minimal ingroup should facilitate spontaneous inferences with regards to positive but not negative traits, whereas valence of traits should not affect STIs about outgroup members. At the beginning of the experiments, participants were individually categorized into a novel social category, allegedly based on their perceptual style when structuring visual information. The second part of the study was introduced as dealing with the structuring of verbal information. The ingroup versus outgroup condition was realized by claiming that the sentences that were presented stemmed from either ingroup or outgroup members who had described activities from their daily life. Following each sentence (which were either trait-implying or not) a target word was presented and participants had to decide quickly, whether this word was in the sentence or not. It was hypothesized that trait inference would be facilitated (thus interfering with the correct rejection of the word as not in the sentence) by references to the ingroup, but only for traits that were positive and that had been implied by the preceding sentence (for more details, see Otten & Moskowitz, 2000). The 2x2x2 (trait implication of sentence x valence of trait x group) ANOVA showed the predicted three-way interaction. The longest response latencies for correctly rejecting trait words were measured when positive traits followed sentences describing ingroup members performing behaviors that implied the respective trait. No effects of valence and group was found for trait words following non-trait implying sentences (see figure 1 for the results on trait-implying sentences).
Taken together, these experiments provide convincing evidence that there is ingroup favoritism on an implicit level even towards novel laboratory groups. The experiments by Otten and collaborators (Otten & Moskowitz; Otten & Wentura, 1999) did not involve any explicit social comparison and avoided reference to unspecific ingroup and outgroup designators as used by Perdue and colleagues (1990); hence, the findings support the assumed positive ingroup default that is the starting point or baseline rather than the result of intergroup evaluations. However, these results do not provide evidence about a process that can account for the immediate, automatic emergence of a positive ingroup stereotype. At this point, reference to the self seems worthwhile. As will be argued in more detail below, it is assumed that the positive value of novel ingroups is based on their association with the typically positively evaluated self (see e.g. Baumeister, 1998; Diener & Diener, 1996; Taylor & Brown, 1988).
Implicit associations between ingroup and self
Ample evidence for the close link between self-definitions and the definitions of own groups on an implicit level was provided in a series of experiments by Smith and collaborators (Smith & Henry, 1996; Smith, Coats & Walling, 1999; Coats, Smith, Claypool & Banner, 2000). Their research was based upon a paradigm developed by Aron and collaborators in order to demonstrate what they call “self-expansion” in interpersonal relationships (e.g. Aron, Aron, Tudor & Nelson, 1991). The cognitive connection between self and ingroup is reflected in response time evidence: Dichotomous judgments of self and ingroup, respectively, were significantly facilitated on those dimensions on which self definition and ingroup definition (as measured in previously administered questionnaires) matched. Smith and collaborators (Smith et al., 1999) conclude that self and ingroup are linked in connectionist networks of memory. When mental representations of self and ingroup overlap, the two stimuli elicit similar patterns of activation. Thereby, activating one concept facilitates congruent responses with regards to the other.
Interestingly, on this implicit level, there is little evidence for any systematic links between outgroup and self or ingroup and outgroup. In terms of the need for positive ingroup distinctiveness as assumed by SIT (see above) or the optimal distinctiveness theory by Brewer (1993), one might have expected that responses were facilitated on those trait dimensions characterized by mismatches with the outgroup. However, typically no such evidence was found (Coats, Smith, Claypool & Banner, 2000; Smith & Henry, 1996; Smith et al., 1999; see Brewer & Pickett, 1999, for one exception).
Although the empirical evidence revealing links between self and ingroup on the implicit level is broadly unequivocal, these findings are not yet sufficient to account for ingroup favoritism towards completely novel, arbitrary groups. Smith and collaborators tried to use social categories that were not heavily stereotyped (e.g. students with different majors; members or non-members of sororities/fraternities), but these were not completely novel groups. In fact, the task of evaluating ingroup and outgroup on a large set of trait dimensions can hardly be realized for arbitrary social categories. Therefore, when replicating the study by Smith and collaborators (1999) Epstude and Otten (2000) also refrained from analyzing minimal groups. Instead, they focused on how realistic social categories were judged in a forced-choice response format in the case of judgmental ambiguity (as indicated by the previous paper-pencil-ratings). As Hogg and Mullin (1999; see also Grieve & Hogg, 1999) have demonstrated, uncertainty is a crucial feature in minimal intergroup settings. Typically, this uncertainty is assumed to stem from the novel, ill-defined social categorization; however, one might argue that uncertainty could also be elicited by focusing on trait dimensions that are not clearly defined for the (realistic) groups in question. In the original studies, traits that were not judged as either applicable or as not applicable to the respective targets (answer “4” on the 7-point bipolar scale) were excluded from further analysis. However, in our study (where we used gender as categorization criterion) exactly these dimensions were of central interest. Data indicated that judgments about ambiguous trait were facilitated when adapted to self ratings, thus replicating the pattern of the match-mismatch effect as already obtained for clearly defined traits; the interaction between type of dichotomous ingroup judgment (response: yes, no) and previous self rating (trait applies; trait not applies) was significant for both ambiguous ingroup-traits and well-defined ingroup-traits. Again, matches or mismatches with the outgroup had no effects on response latencies.
The original findings by Smith and collaborators reveal that there is a firm link between the concepts of self and ingroup. Within the connectionist model, it does not really matter whether the link between the two targets stems from a definition of self in terms of the ingroup (as discussed by SCT; Turner et al., 1987) or whether it stems from a definition of the group in terms of the self, that is, from self-anchoring (see Cadinu & Rothbart, 1996; see also Krueger, 1998, for the more general model of egocentric social projection). In this context, the study by Epstude and Otten (2000) suggests a possibility to disentangle and analyze the direction of activation patterns as manifested in response latencies.
Self-anchoring in explicit measures of ingroup favoritism
The findings summarized in the previous section provide convincing evidence for a firm association between self evaluations, on the one hand, and ingroup evaluations, on the other hand. Implicitly, such link can already account for positive ingroup judgments, as there is much evidence (at least in western cultures) for the overall tendency to see the self as positive and ‘above average’ (e.g. Baumeister, 1998; Matlin & Stang, 1978, Taylor & Brown, 1988; Triandis, 1989). Thus, in the following more direct tests of whether assimilating self- and ingroup-definition can account for positive and positively distinct ingroup judgments will be presented.