Categorization and intergroup anxiety in contact between British and Japanese nationals

Two studies examined the relationship between categorization, intergroup anxiety and intergroup attitudes (intergroup bias and negative a€ect). Study 1 consisted of a survey of 236 British and Japanese nationals. Study 2 was a longitudinal study of 54 Japanese students studying in the UK. Of the three categorization variables (interpersonal, superordinate and intergroup), only intergroup categorization was shown to have a relationship to generalized intergroup attitudes. In addition, intergroup anxiety and quality of contact were associated with ingroup bias and negative a€ect to the outgroup. Study 2 revealed an interaction between intergroup categorization and quality of contact in predicting negative a€ect. Intergroup anxiety was also associated with increased intergroup categorization. It is concluded that the e€ects of categorization during contact are still poorly understood, and that intergroup anxiety is a far more powerful variable in contact than the current literature acknowledges. Copyright # 1999 John Wiley & Sons, Ltd. Despite more than forty years' intensive research, there is still no clear consensus on how to improve intergroup relations through intergroup contact. In recent years, social identity theory (Tajfel, 1978; Tajfel & Turner, 1979) has led to an emphasis on the role of categorization. Opposing models have each argued that speci®c levels of categorization (interpersonal, intergroup or superordinate) are necessary if contact is to have a positive e€ect on intergroup relations as a whole (Brewer & Miller, 1984; Gaertner, Mann, Murrell & Dovidio, 1989; Hewstone & Brown, 1986). A separate line of research has concentrated on the a€ective experience of contact and, in particular, on the role of intergroup anxiety (Stephan & Stephan, 1985). In this perspective, intergroup anxiety is viewed as both a determinant and a consequence of intergroup contact. With a few notable exceptions (e.g. Islam & Hewstone, 1993), CCC 0046±2772/99/040503±19$17.50 Received 5 August 1997 Copyright # 1999 John Wiley & Sons, Ltd. Accepted 12 June 1998 European Journal of Social Psychology Eur. J. Soc. Psychol. 29, 503±521 (1999) *Correspondence to: Dr. Katy Greenland, School of Psychology, Cardi€ University, PO Box 901, Cardi€ CF1 3YG, UK. e-mail: greenlandKA@cardi€.ac.uk Contract grant sponsor: Economic and Social Research Council. little research has attempted to trace the link between categorization and intergroup anxiety. The research to be presented will examine the e€ects of both categorization and intergroup anxiety on contact between British and Japanese nationals. In contrast to much of the current categorization literature, it will employ real groups and contact outside the laboratory. Further, the relationship between categorization and intergroup anxiety will be explored, and the implications for this relationship on contact will be examined. CATEGORIZATION AND INTERGROUP CONTACT The decategorization model of Brewer and Miller (1984) emphasizes the role of interpersonal perception during contact. It suggests that by discouraging the use of category-based perception, participants are less likely to display the range of ingroup favouring biases that are associated with psychologically salient categories (Brewer, 1979). Further, according to this model, if participants are successful in viewing outgroup members in individuated, personalized terms, the psychological utility of the category should be reduced. Participants should not only develop more positive attitudes to speci®c outgroup members, but may also be less likely to utilize the category in the future or in di€erent contexts. A number of experimental studies have provided empirical support for the decategorization model (Bettencourt, Brewer, Croak &Miller, 1992; Markus-Newhall, Miller, Holtz & Brewer, 1993; Miller, Brewer & Edwards, 1985). The recategorization model of Gaertner and his colleagues also involves attempting to reduce the salience of the group boundary (Gaertner et al., 1989, Gaertner, Mann, Dovidio, Murrell & Pomare, 1990). However, instead of deconstructing respective group categorizations, Gaertner et al. suggest that participants should be encouraged to recategorize both inand out-group members into a larger superordinate category. By subsuming both groups into a single, larger group (e.g. being students), distinctions between erstwhile inand outgroup members should be reduced. In this way, ingroup favouritism is again less likely to occur, and the probability of the original category being utilized in future may be reduced. Some ®eld and experimental evidence supports this model (e.g. Dovidio, Gaertner, Validzic, Matoka, Johnson & Frazier, 1997; Gaertner et al., 1989, Gaertner, Rust, Dovidio, Bachman & Anastasio, 1994; Gaertner, Dovidio & Bachman, 1996). The categorization model of Hewstone and Brown (1986) contrasts sharply with these two models. Hewstone and Brown argue that the e€ects of contact will not generalize unless the categories maintain some minimal psychological salience. In this way the interaction can be regarded as `intergroup' rather than interpersonal* (Brown & Turner, 1981). Hewstone and Brown argue that intergroup categorization does not inevitably lead to ingroup favouritism, as long as opportunities exist for mutually positive intergroup comparisons along reciprocal dimensions (Mummendey & *The use of the term `intergroup' in this formal, technical sense requires that social categories are psychologically salient and that there is some evidence of perceived group homogeneity and behavioural intergroup uniformity (Brown & Turner, 1981). Thus, it goes beyond the looser and more generic use of `intergroup', which simply denotes an interaction involving members of di€erent groups. 504 K. Greenland and R. Brown Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 503±521 (1999) Schreiber, 1984; van Knippenberg & van Oers, 1984). In support of this model there is evidence which suggests that non-typical group members may be subtyped as exceptions to the rule: stereotype discon®rming information associated with them is therefore ignored (Johnston & Hewstone, 1992; Rothbart & John, 1985; Weber & Crocker, 1983). More directly, some evidence suggests that contact with a typical outgroup member is associated with attitude change to the outgroup as a whole (Desforges, Lord, Ramsey, Mason, van Leeuwen, West & Lepper, 1991; Desforges, Lord& Pugh, 1997; vanOudenhoven, Groenewoud&Hewstone, 1996; Vivian, Brown & Hewstone, 1995; Wilder, 1984), while contact that contains only personalized information is not (Scarberry, Ratcli€e, Lord, Lanicek & Desforges, 1997). Leaving to one side the contrasting predictions (and evidence) based on these models, it is worth identifying some limitations shared by all three approaches. First, they have relied primarily on laboratory methodologies and ad hoc groups. In this context, participants may have less of an investment in the group and their categorization processes may be more malleable. In real groups, individuals may actively resist the (de)emphasis on certain levels of categorization and reject interventions that attempt to achieve this. Real category memberships may also remain salient when the di€erences between the groups involved reinforce stereotypes (Lee & Duenas, 1995). The current categorization literature has therefore largely failed to address the reality and in ̄exibility of categorization in the real world. The same literature has also concentrated on examining the one-shot e€ects of di€erent levels of categorization, rather than investigating the use of categorization over time (Pettigrew, 1986; Stephenson, 1981; van Oudenhoven et al., 1996). Most importantly for the research to be presented, however, it can be argued that all three models are strongly cognitive in emphasis and have neglected more a€ective processes. Di€erent kinds of categorization may have di€erent a€ective consequences, and these in turn may in ̄uence categorization processes. INTERGROUP ANXIETY AND INTERGROUP CONTACT Stephen and Stephan (1985) suggested the term `intergroup anxiety' to denote the anxiety that an individual may feel when anticipating or experiencing contact with someone from another group.* Stephan and Stephan suggested a number of antecedents of such anxiety, many of which are embedded in the history of intergroup relations and the social context. For example, an individual may fear that they will be discriminated against, that the group may reject them, or that their identity or selfesteem may be threatened. Individuals may experience more intergroup anxiety where there has been a history of discrimination, where the perceived di€erences between groups are large, or where the individual has had minimal previous contact. There are presumably other, more idiosyncratic, reasons why a person may experience anxiety in such group settings, which may have little to do with speci®cally intergroup variables. There is a sizeable literature examining the consequences of a€ect and mood on cognition and judgement, and a number of competingmodels to account for how some *Note that `intergroup' is used here in its generic sense simply to indicate that a member of the other group may be present. This does not imply that the contact is necessarily construed psychologically at an intergroup level in the Brown and Turner (1981) sense, although this is also possible. Categorization and intergroup anxiety 505 Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 503±521 (1999) of the observed e€ects might come about. It is beyond the scope of the current paper to review these e€ects and competing models in depth. Instead we will focus on one speci®c aspect of this research: the relationship between generic anxiety and stereotyping.Wewill then go on to review the (very limited) evidence concerning the relationship between intergroup anxiety (in Stephan and Stephan's sense) and stereotyping. There is strong evidence to suggest that generic anxiety is associated with increased stereotyping. Baron, Inman, Kao & Logan (1992) examined illusory correlation (which Hamilton argues can be regarded as an index of stereotyping) among dental patients and observed that those patients who reported high anxiety were more likely to exhibit the illusory correlation e€ect. Wilder and Shapiro (1989b) conducted a series of studies in which participants were made anxious before watching a videotape of a target group. They reported that participants in the high anxiety condition assimilated individual group members into the group. Wilder and Shapiro's account for this e€ect was that participants in the high anxiety condition experienced elevated physiological arousal: this distracted participants such that they were unable to concentrate on individuating information. The suggestion that the e€ects of anxiety is mediated by arousal is further supported by evidence of the e€ects of arousal on stereotyping (Kim & Baron, 1988; Bodenhausen, 1993). Although most of the research in the ®eld appears to have concentrated on the arousal±distraction hypothesis, there are a number of other theoretical accounts that may be of interest. For example, it has been suggested that a negative mood renders negative cognitions more accessible, such that individuals who are in a negative mood are likely to interpret ambiguous information in a negative way (e.g. Esses & Zanna, 1995). Alternatively, Mackie, Hamilton, Schroth, Carlisle, Gersho, Meneses, Nedler and Reichel (1989) have suggested that incongruency between mood and target (e.g. a negative mood but a positive target) requires greater processing, and may therefore lead to increasing reliance on heuristics. However, the theory that contrasts most sharply with the distraction model is the `feelings as information' model (Schwarz, 1990; Schwarz & Clore, 1988). Schwarz and his colleagues argue that a€ect is used to signal well-being (or the lack of it). A negative a€ective state, according to Schwarz, signals a lack of well-being. The adaptive response to this is to engage in problem solving, which ultimately involves processing a€ect-related information more systematically. This model contrasts with the distraction model in two important ways. First, it predicts that the consequences of a€ect on information processing will interact with the perceived relevance of the information to the a€ective state (see the distinction between integral and incidental a€ect below). Second, and most importantly, the `feelings as information' model predicts that anxiety should be associated with reduced stereotyping. Clearly, this prediction is inconsistent with the empirical evidence already described. Although this may be partly attributable to empirical problems with the current evidence (and explored in more depth shortly), it has also been suggested that the `feelings as information' model fails to incorporate arousal suciently (Bodenhausen, 1993; Bodenhausen, Sheppard & Kramer, 1994).* *More recently, Wilder and Simon (1996) have suggested that there is a tension between individuating problem-solving processes, and heuristic processes generated by physiological arousal. This model has yet to be tested empirically. 506 K. Greenland and R. Brown Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 503±521 (1999) Although there are a number of di€erent accounts for the relationship between (negative) stereotyping and anxiety per se, there is little argument about the e€ect, at least within the laboratory. The aims of the current research is to establish how intergroup anxiety may in ̄uence intergroup contact and categorization. There are a number of issues which must be addressed here. First, much of the research on anxiety has been conducted with ad hoc, experimentally generated groups (e.g. Wilder & Shapiro, 1989a, b). In such contexts, it has not always been demonstrated that reduced information processing is associated with increasing access and utilization of outgroup stereotypes: stereotypes may not be activated, or may not be functionally relevant to the situation (Gilbert & Hixon, 1991; Oakes, Haslam & Turner, 1994). Second, the majority of the work has used incidental rather than integral manipulations of anxiety (Bodenhausen, 1993). This means that the source of the anxiety has been unrelated to the task on which participants' performance was being measured. This is an important distinction, since an integral manipulation of anxiety could theoretically facilitate information processing (see Schwarz, 1990, outlined above). This is a hypothesis that has not yet been successfully examined in the literature and is unfortunately beyond the scope of the current paper (see Greenland & Brown, in press). On this basis, we can not assume that the e€ects of incidental manipulations of anxiety will be similar to the e€ects of integral, intergroup anxiety. Compared to the literature on anxiety in general, very little research has been conducted speci®cally on intergroup anxiety. The main evidence comes from two surveys, and is largely consistent with the generic anxiety literature. Stephan and Stephan's (1985) survey of Anglo and Hispanic Americans suggested that intergroup anxiety is correlated with outgroup stereotyping. Islam and Hewstone (1993) obtained a similar correlation between intergroup anxiety and perceived group variability in a survey of Muslims and Hindus. Combined, these results suggest that the relationship between generic anxiety and stereotyping does extend to intergroup anxiety. They also suggest that there may be a relationship between intergroup anxiety and intergroup categorization.* The surveys outlined above were correlational, and it is not possible to draw any strong conclusions about the causal relationship between the variables.Most of the experimental literature already reviewed suggests that it is intergroup anxiety that increases category use.{ However, there is an alternative argument: intergroup categorization could increase intergroup anxiety. Insko and his colleagues (Insko, Pinkley, Harring, Holton, Hong, Krams, Hoyle & Thibaut, 1987; Insko, Schopler, Hoyle, Dardis &Graetz, 1990;McCallum, Harring, Gilmore, Drenan, Chase, Insko& Thibaut, 1985; Schopler, Insko, Graetz, Drigotas, Smith & Dahl, 1993) provide evidence to suggest this e€ect. Insko et al. argue that participants who experience intergroup (as opposed to interindividual) contact may assume that this will be competitive. Since it has also been demonstrated that competition is associated with increased anxiety (Wilder and Shapiro, 1989a), it is conceivable that participants who use intergroup categorization may also experience more intergroup anxiety. The relationship between intergroup anxiety and categorization is of both theoretical and practical relevance, both to draw intergroup anxiety into the contact *This is not to suggest that intergroup categorization inevitably implies stereotype use. However, for a stereotype to be accessible, it is reasonable to suppose that the category must also be accessible. {It might also be suggested that there would be a curvilinear relationship between anxiety and performance, with maximal performance (i.e. most systematic processing) at moderate levels of anxiety. We found no evidence to support this hypothesis. Categorization and intergroup anxiety 507 Copyright # 1999 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 29, 503±521 (1999) literature, and to explore the viability of the intergroup categorization model. To date, however, we know of only one attempt to outline this relationship in any depth. The contact model of Vivian, Hewstone and Brown (1997) explicitly includes a link between intergroup anxiety and intergroup categorization. Vivian et al. argue that (1) intergroup categorization requires outgroup members to be typical and that (2) where there is a history of con ̄ict, the typical outgroup member may be seen as negative. Vivian et al. therefore predict that intergroup categorization may increase intergroup anxiety, but as yet there is no empirical evidence to support this argument. THE CURRENT RESEARCH The research to be presented examined the e€ects of categorization and intergroup anxiety in naturally occurring contact between members of real groups. The groups used were British and Japanese nationals.* The main independent variables were quality of intergroup contact; interpersonal, intergroup and superordinate group categorization; and intergroup anxiety. The dependent variables were ingroup bias and negative a€ect towards the outgroup. There were three main predictions from the contact and intergroup anxiety literature. First, we predicted that there would be a signi®cant negative relationship between quality of contact and both ingroup bias and negative a€ect towards the outgroup. Second, we predicted that there would be a signi®cant positive relationship between intergroup anxiety and bias and a€ect. Finally, there were rival predictions for a relationship between interpersonal, intergroup and superordinate categorization and ingroup bias and a€ect. No speci®c predictions was made in favour of any one of these models. In addition to the main predictions, there was an additional prediction over the relationship between intergroup anxiety and intergroup categorization. We predicted that there would be a signi®cant positive relationship between intergroup anxiety and intergroup categorization. The causal relationship between these two variables was speculative, and no speci®c predictions over causality were made. Data from two studies will be presented. The ®rst consisted of a survey reporting contact between Japanese and British people. The second was a longitudinal study, examining the e€ects of intergroup anxiety and categorization over time.