1
Responding to the social world:
Attributions and stereotype-based judgments
Lucy Johnston and Lynden Miles
University of Canterbury, New Zealand
The present chapter considers recent research that has investigated how social perceivers respond to stereotype-relevant information. Past research on stereotype change[1] has focused solely on stereotype-inconsistent information and the extent to which perceivers integrate such information into their pre-existing beliefs about a target group. In addition to reviewing this research, the present chapter also describes research that has investigated the impact of stereotype-consistent information on perceptions of the target group. Indeed a major argument put forward in the chapter is that greater attention should be paid to perceivers’ responses to stereotype-consistent information as a potential means of moderating stereotype-based judgments. The presented research considered both direct and indirect measures of the impact of stereotype-relevant information. The direct measures used were stereotype-based ratings of the target group. The indirect measures used related to measurement of the attributions that perceivers offered for stereotype-relevant information. We advocate that such indirect measures provide important insight into the reaction of social perceivers to stereotype-relevant information that may not be captured by direct measures alone.
The assumption of a strong association between negative stereotypes and discrimination (Dovidio, Brigham, Johnson & Gaertner, 1996) has led both researchers and social legislators to pursue attempts to change social stereotypes. Guided by Allport’s (1954) contact hypothesis it has been assumed that providing perceivers with additional information about target groups will increase the accuracy of their beliefs about these groups and hence reduce their reliance on inaccurate group-based stereotypes. Experimental attempts to moderate social stereotypes have, accordingly, involved presenting perceivers with information, including stereotype-inconsistent information, about the target group and then asking them to judge the group on stereotype-based dimensions. This approach has been shown to lead to stereotype moderation under certain conditions (see Hewstone, 1994 for a comprehensive review). A number of processes by which perceivers maintain their stereotypes, even in the face of inconsistent information have also, however, been identified, suggesting that the maintenance of stereotypic beliefs is not simply a function of lack of information about a target group. Indeed research has demonstrated that perceivers are prepared to devote cognitive resources to stereotype maintenance (Yzerbyt, Coull & Rocher, 1999).
One prevalent means by which the impact of inconsistent information is reduced is by isolation, or exclusion, of disconfirming group exemplars. If individuals displaying stereotype-inconsistent behaviour can be considered to be atypical exemplars of the target group they are subtyped (Brewer, Dull & Lui, 1981; Johnston & Hewstone, 1992; Kunda & Oleson, 1995, 1997). That is, they are considered to be a distinct subcategory of the superordinate target group. A link between the subordinate and superordinate group is acknowledged but members of the subtype are considered to possess certain stereotype-inconsistent characteristics in addition to the stereotype consistent characteristics that they share with the superordinate group members. Although the subtyped exemplars are judged in a less stereotype based manner as a consequence of their stereotype-inconsistent behaviour, this reduction in stereotyping is not generalized beyond the subcategory to the group as a whole. Subtyping can explain why extremely disconfirming exemplars (eg. Margaret Thatcher, Jesse Jackson) have little impact on group-based beliefs (of women and African-Americans respectively). However, many mildly disconfirming group exemplars, who cannot be readily subtyped, are encountered in everyday situations. How is it that perceivers maintain their stereotype-based beliefs in the presence of such exemplars displaying stereotype-inconsistent behaviours?
Attributions and Stereotype Moderation.
One possible process through which perceivers can maintain their stereotype-based beliefs in the face of inconsistent information is through the attributions made for that information. Compared to stereotype-consistent or expected behaviours, unexpected or inconsistent behaviours are more likely to result in spontaneous attribution or explanation by perceivers (Hastie, 1984; see also von Hippel et al., this volume). Hewstone (1989) proposed an attributional model of stereotype change which predicted that stereotype-inconsistent information would only lead to moderation of group-based beliefs if that behaviour was attributed to stable dispositional factors of a typical group exemplar. An internal stable attribution indicates that the stereotypically inconsistent act is representative of the actor’s usual behaviour leading to the actor being perceived in a non-stereotypic manner (Krueger & Rothbart, 1988; Locksley, Borgida, Brekke & Hepburn, 1980). Since the actor is also perceived to be a typical group exemplar his/her behaviour is considered to be representative of the target group. Hence, this non-stereotypic perception of the target individual should generalize to the group as a whole, resulting in stereotype moderation. In contrast, if the target is considered to be an atypical group member, generalization would not occur as his/her behaviour is not considered to be representative of that of other group members, and hence the exemplar could be subtyped. In a similar manner, situational attribution for inconsistent behaviour also leads to the prediction of no generalization to group-based perceptions, as the behaviour is not considered to be predictive of the target’s usual behaviour (Jackson, Sullivan & Hodge, 1993). Wilder, Simon and Faith (1996) provided empirical support for Hewstone’s model. Participants were presented with a stereotype-inconsistent behaviour for which they were provided with either a situational or a dispositional attribution. Subsequent judgments of the target group were less stereotype-based than those of control, no information, participants only when the exemplar was considered to be a typical group member and the behaviour had been attributed to dispositional causes. Judgments of the group did not differ from baseline when the inconsistent behaviour was attributed to situational causes. Attributing stereotype-inconsistent behaviour to dispositional causes can then result in moderation of stereotype-based beliefs. However, stereotype-inconsistent behaviour is usually attributed to situational rather than dispositional factors (Bodenhausen & Wyer, 1985; Duncan, 1976; Evett et al., 1994; Macrae & Shepherd, 1989), attributions that do not lead to revision of stereotype-based judgments (Wilder et al., 1996).
Rather than provide participants with attributions for a stereotype-inconsistent behaviour, in two experiments we asked them to rate the extent to which the behaviour was caused by both situational and dispositional factors. Participants then evaluated the individual target and the stereotyped group on stereotype-based dimensions (Johnston, Bristow & Love, 2000; Expt’s 2 and 3). Consistent with past research (Bodenhausen & R.S. Wyer, 1985; Duncan, 1976; Evett et al., 1994; Macrae & Shepherd, 1989) the stereotype-inconsistent behaviour was attributed more strongly to situational than to dispositional factors, although some participants (approximately 30%) did show the opposite pattern of attributions, attributing the inconsistent behaviour more strongly to dispositional than to situational factors. This finding is discussed in more detail later. The stereotype-inconsistent behaviour resulted in the target individual being evaluated in less stereotype-based terms but this did not generalize to the group as a whole. Consistent with the stronger attribution to situational causes, there was no reduction in stereotype-based judgments of the target group relative to control, no information, participants. In order to further investigate the relationship between the attributions made for the target’s behaviour and stereotype-based ratings, an attribution index was calculated for each participant. Ratings for dispositional causes were subtracted from those for situational ratings; a positive index indicates stronger situational than dispositional attributions and a negative index stronger dispositional than situational attributions (Wittenbrink, Gist & Hilton, 1997). Correlations were computed between the attribution index and stereotype-based ratings of both the target individual and the group as a whole. There was only one significant correlation, between the index and ratings of the target in Experiment 3. The higher the index (relatively stronger situational ratings) the more strongly the target was evaluated in stereotype-based terms. There was no correlation between the attribution index and judgments of the group, or between judgments of the target individual and the group however; less stereotype-based judgments of the target were not generalized to judgments of the group as a whole. Although attributing stereotype-inconsistent behaviour to dispositional causes can lead to stereotype moderation (Wilder et al., 1996), such attributions, and hence stereotype moderation are rare. Indeed, perceivers may use situational attributions to reduce the impact of stereotype-inconsistent information on pre-existing group-based beliefs.
The research discussed this far, and that reviewed by Hewstone (1994) has all involved participants being presented with a pre-packaged set of information, including stereotype-inconsistent information, about members of the target group. This experimental approach allows researchers to investigate the impact of inconsistent information on perceivers’ judgments of the group but, at the same time it fails to capture an important feature of the social perceiver. Social perceivers are, by nature, active gatherers, not passive recipients, of information (Yzerbyt & Leyens, 1991). Using information seeking methods (e.g., a bulletin board; question asking) allows perceivers to control, to some extent, the nature and amount of information they receive about members of a target group, whilst still allowing some experimental control over the information received by participants. Using such methods we have demonstrated that perceivers show a bias toward stereotype-consistent information and avoid stereotype-inconsistent information (Johnston, 1996; Johnston & Macrae, 1994). In addition the impact of stereotype-inconsistent information received under such situations is reduced relative to traditional information-given research. Perceivers who received stereotype-inconsistent information under information-gathering conditions did not moderate their stereotype-based judgments, relative to controls. Perceivers who received the same amount of stereotype-inconsistent information under information-given conditions, however, did show stereotype moderation. Perceivers in the information-gathering conditions did attend to the inconsistent information, but its impact was reduced (Johnston, 1996). In more realistic, information gathering situations, therefore, perceivers may avoid inconsistent information and the impact of any such information may be less than in traditional research (Hewstone, 1989). Caution needs, then, to be exercised in generalizing findings of stereotype moderation from the research laboratory.
The attention of researchers has, understandably, focused on the impact of stereotype-inconsistent information on stereotype-based beliefs. In doing so, however, we argue that an intriguing, counter-intuitive alternative route to stereotype moderation has been ignored. Perceivers prefer to attend to stereotype-consistent information that is easy to process and to integrate into pre-existing beliefs (Johnston, 1996; Johnston & Macrae, 1994). Such behaviour is typically attributed to internal, stable factors (Bodenhausen & Wyer, 1985; Duncan, 1976; Evett et al., 1994; Macrae & Shepherd, 1989), implying generalization of the behaviour across both time and target group member, and hence stereotype maintenance. In the same way as attributing inconsistent information to dispositional factors in order to moderate stereotype-based beliefs (Wilder et al., 1996), one can ask whether attributing consistent information to situational rather than to dispositional factors might lead to moderation of group-based beliefs. If perceivers believe that individuals perform stereotype-consistent behaviours because of situational pressures rather than because of personal characteristics, they may be less likely to endorse stereotypic beliefs about the target group to which those individuals belong. Such an argument for the impact of situational factors may be easy to sustain. Consider, for example, sex-based stereotypic beliefs. Social role theory (Eagly, 1987; Eagly & Steffen, 1984) argues that sex-based stereotypes developed as a consequence of an unequal distribution of men and women in different social roles, not as a consequence of fundamental differences between men and women. That is, sex-based perceptions are argued to be the result of situational, or role, constraints rather than dispositional factors. Making perceivers aware of the situational causes of stereotypic behaviour may reduce stereotype-based perceptions, just as the reversal of roles sees the disappearance of gender-based stereotypes (Eagly & Wood, 1982).
In three experiments we investigated the relationship between attributions made for stereotype-consistent information and stereotype-based judgments (Johnston et al., 2000). In the first experiment participants were provided with either a dispositional or a situational attribution for a target’s stereotype-consistent behaviour and subsequently asked to judge both the target and the group as a whole on stereotype-based dimensions. When the stereotypic behaviour was attributed to internal causes, there was no moderation of the stereotype-based judgments of either the individual target or the group, relative to control participants. Those participants given a situational attribution for the target behaviour, however, rated the target group in less stereotype-based terms than did the baseline participants. Hence, as predicted, making participants aware of situational constraints on the stereotypic behaviour reduced the strength of stereotype-based judgments of the target group.
Experiment 1 demonstrated that situational attributions for stereotype-consistent information can lead to stereotype moderation. In two subsequent experiments, each using a different group and target behaviour, participants were asked to generate attributions rather then being provided with such. Participants rated the extent to which stereotype-consistent behaviour was a result of both situational and dispositional factors prior to evaluating the target and group on stereotype-based dimensions[2]. The stereotype-consistent behaviour was, in both experiments, attributed more strongly to dispositional than to situational factors, a pattern of attributions consistent with past research (Bodenhausen & Wyer, 1985; Duncan, 1976; Evett et al., 1994; Macrae & Shepherd, 1989) and with stereotype maintenance. In neither study was there moderation of the group-based stereotype relative to baseline participants. These results parallel those seen for the stereotype-inconsistent behaviour. Although attributing stereotype-consistent behaviour to situational causes may lead to stereotype moderation, such attributions, and consequently stereotype moderation, did not occur when participants provided the attributions for the target behaviour themselves. As with the inconsistent behaviour a small number of participants (approximately 10%) did show the opposite pattern of attributions, attributing the stereotype-consistent behaviour more strongly to situational than dispositional factors.
As for the inconsistent behaviour, an attribution index was calculated for each participant and correlations between the attribution index and stereotype-based ratings of the target individual and group computed. In contrast to the findings for the stereotype-inconsistent behaviour, this analysis revealed relationships between the attribution index and stereotype-based judgments of both the target individual and the group, as shown in Figure 1. In Experiment 2 there were significant correlations between the attribution index and ratings of both the target individual and the group. The stronger the behaviour was attributed to situational relative to dispositional factors (higher attributional index) the less stereotypic both the individual and the group were perceived to be. In addition there was generalization from stereotype-based ratings of the individual to those of the group. In Experiment 3 there was a significant correlation between the attribution index and ratings of the individual but not ratings of the group as a whole. There was however, strong generalization from stereotype-based ratings of the individual to those of the group. For stereotype-consistent behaviour, therefore, greater recognition of the situational constraints on stereotypic behaviour (a higher attribution index) was related to less stereotype-based ratings. Stereotype moderation was not dependent on higher absolute situational than dispositional attributions. No parallel relationship between the attributional index and stereotype-based ratings was evident for the stereotype-inconsistent behaviour. Hence, moderating stereotype-based beliefs through the presentation of stereotype-consistent information may be more effective than through the presentation of inconsistent information. Given perceivers’ preference for stereotype-consistent information in information-gathering situations, this possible opportunity for stereotype moderation warrants further research attention.
Insert Figure 1 here
For both stereotype-consistent and -inconsistent behaviours there was a small proportion of participants who showed the “opposite” pattern of attributions to those expected. That is, some perceivers attributed inconsistent behaviour more strongly to dispositional than to situational factors or consistent behaviour more strongly to situational than to dispositional factors. Although there were too few individuals in these groups to provide sufficient power for statistical analyses, two interesting trends were evident. First, a greater proportion of participants showed this “opposite” pattern of attributions for inconsistent than for consistent behaviour. This trend is consistent with the general tendency of perceivers to attribute the behaviour of others to internal causes (Ross, 1977). Second, the impact of making the “opposite” pattern of attributions appeared to differ for stereotype-consistent and –inconsistent behaviors. For those who made the “opposite” pattern of attributions for an inconsistent behavior, mean stereotype-based ratings of the target group did not show stereotype moderation relative to control participants. The impact of the attributional pattern for stereotype-inconsistent information was again minimal. For those who made the “opposite” pattern of attributions for a consistent behavior, however, mean ratings of the target group were less stereotype-based than those of baseline, control, participants. Again, the possible role of consistent information in stereotype moderation is highlighted.
Prejudice, Attributions and Stereotype Moderation.
The presence of individuals who made the “opposite” pattern of attributions for stereotype-based behaviour raises the question of whether certain characteristics of the social perceiver are associated with the non-expected pattern of attributions for stereotype-relevant information and under what conditions. The remainder of this chapter describes ongoing research in our labs that addresses this question. There has been little consideration of the possible influence of individual differences on stereotype moderation. Monteith, Zuwerink and Devine (1994) suggested that stereotype change strategies were, as a consequence, often ineffective because they were not targeted appropriately for their specific audiences. Understanding the influence of individual differences on perceivers’ responses to stereotype-relevant information may enable the development of more effective stereotype-change strategies targeted at specific perceivers. An intuitively obvious individual difference factor to consider in this context is prejudice level. High and low prejudiced individuals differ in the extent to which they endorse and use stereotypes (e.g., Devine, 1989) and in their reactions to their own stereotype use (Devine, Monteith, Zuwerink, & Elliot, 1991). Low prejudiced individuals strive against stereotype use and feel guilt if they do (inadvertently) use stereotypes whilst high prejudice individuals do not experience negative self-focused emotions as a consequence of stereotype use (Devine et al., 1991). Low prejudiced individuals are motivated to avoid stereotype use under all conditions. High prejudiced individuals, on the other hand, are happy to use stereotypes unless doing so has negative consequences, such as social disapproval (Plant & Devine, 1998). It would seem reasonable to suggest that low and high prejudiced individuals may display different patterns of attributions for stereotype-related behaviours. High prejudiced individuals are motivated to maintain stereotypes and hence we predicted that they would display a pattern of attributions consistent with stereotype maintenance. In comparison, low prejudiced individuals are motivated to avoid stereotype use and hence it is predicted that they would display a pattern of attributions consistent with stereotype moderation. That is, we hypothesized that those individuals making an “opposite” pattern of attributions, attributions consistent with stereotype moderation, would have lower prejudice toward the target group than those showing the “usual” pattern of attributions, attributions consistent with stereotype maintenance.