C. Neil Macrae
Galen V. Bodenhausen
Alan B. Milne
A pervasive problem in mental life is that of exemplar selectivity or how one isolates specific category members from other instances of a class. This problem is particularly pronounced in person perception, where perceivers may routinely want to personalize selected individuals while continuing to respond towards other members of the category in a stereotype-based manner. To realize these flexible effects, we hypothesized that, when perceivers encounter a group member, they inevitably encode an exemplar-based representation of the individual in mind. Part of this representation, moreover, is information signaling the personās goodness-of-fit with respect to his or her salient group memberships. When the representation is activated on a subsequent occasion, these inferences of category fit moderate the extent of stereotypical thinking. The results of two studies provided converging evidence for this analysis of stereotype function. Exemplar typicality moderated both the accessibility of stereotypic knowledge (Study 1) anxd the extent to which perceivers used a stereotype to organize information about a target (Study 2). We consider the theoretical and practical implications of these findings for our understanding of the role of stereotypes in person perception.
Dealing with Disconfirmation
Conventional wisdom deals with the above problem in a relatively straightforward manner. When category disconfirmation occurs, perceivers are believed to search long-term memory for alternative knowledge structures that can successfully accommodate the attributes of deviant group members (Brewer, 1988; Fiske and Neuberg, 1990). For example, if oneās initial classification of Malcolm as a ātraditional maleā were challenged by the discovery that he enjoys cooking, knitting, and child rearing (i.e., low goodness-of-category fit), then one would implement a process of ārecategorizationā in an attempt to preserve a stereotype-based conception of him. This renewed attempt to find an adequate categorical solution can prompt a variety of outcomes. As Fiske and Neuberg note:
Recategorization is an attempt to find a different category that can be interpreted as adequately organizing the bulk of current information. It may entail accessing a subcategory ... an exemplar... a self schema ... or recategorization may entail accessing an entirely new category.... If the perceiver can successfully recategorize the target, then the perceiverās affect, cognitions, and behavioural tendencies in relation to the target are likely to be those relevant to the new category. (1990)
The message emerging here is an instructive one: if at first you donāt succeed, try again until you locate an alternative knowledge structure that can capture a targetās idiosyncratic constellation of attributes. Once activated, it is business as usual as this cognitive representation guides oneās impressions, recollections, and evaluations of the target.
The flexibility of categorization and stereotype activation processes that we have just described is a compelling notion in the abstract, but the underlying cognitive mechanisms that provide this adaptive flexibility are currently not well specified. Consider, for instance. the previous example of Malcolm, the ostensibly non-traditional male. Given his poor fit to the general category stereotype (after all, real men donāt knit; they repair cars and play football), perceivers are believed to search memory for alternative knowledge structures that can accommodate his deviant characteristics. Thus knowledge of his atypical attributes may prompt perceivers to re-classify Malcolm as a āmodernā man (i.e., a category subtype); as someone who resembles Uncle Bob (i.e., an exemplar-based representation); or else in terms of an entirely different social category, such as āhippie.ā Despite the flexibility of the recategorization process, however, it cannot alter a simple fact of life: Malcolm remains a man, albeit a relatively atypical member of the category. One may reasonably speculate, therefore, on what would happen if perceivers were to re-encounter Malcolm in a situation where his gender categorization was the entirely appropriate classification to make, as would surely be the case in a variety of social settings (e.g., when looking for a male tennis partner). Under these conditions, how would oneās previous experience of Malcolm (as an atypical male) shape the course and products of the categorization process? In particular, what would be the fate of the ill-fitting initial categorization?
The issue of category disconfirmation, of course, raises another thorny issue. When perceivers encounter individual exemplars who seem to be atypical of their social group, they clearly face something of a dilemma. Obviously, they must modify any stereotypic impression of the atypical target that has formed, but do they also need to revise their stereotypic preconceptions about the targetās social group as a whole? After all, a central premise of contemporary models of category representation is that, among other things, social perceivers store information about specific group exemplars in memory (Elio and Anderson, 1981; Smith and Medin, 1981; Lingle, Altom and Medin, 1984; Medin et al., 1984; Brewer, 1988; Judd and Park, 1988; Smith and ZĆ”rate, 1992; Bodenhausen et al., 1995). If these exemplars constitute an important basis for category representation, what then is the impact of atypical exemplars on peopleās general categorical beliefs? An inspection of the available literature on this topic confirms that people are quite facile at finding reasons for viewing disconfirming instances as exceptions to a generally valid rule. Accordingly, they see little reason to generalize the attributes of deviant exemplars to the wider social group (Allport, 1954; Brewer et al., 1981; Weber and Crocker, 1983; Wilder, 1984; Rothbart and John, 1985; Kunda and Oleson, 1995). Herein, however, lie several important theoretical puzzles. For example, exactly how is it that perceivers can construe some exemplars in a non-stereotypic manner while continuing to respond towards other members of the group in a predominantly stereotype-based fashion? What are the underlying cognitive processes that make these disparate information-processing outcomes possible? Moreover, at a wider representational level, what sort of cognitive architecture can sustain the flexible stereotyping of group members? Our intuition is that an episodic retrieval model of person perception may provide some useful initial insights into these as yet unresolved theoretical issues.
An Episodic Retrieval Model of Person Perception
Episodic retrieval (i.e., exemplar-based) models of mind posit that information is stored in memory in an instance-based manner. Thus, each experience with a stimulus, be it a hamburger, a bicycle, or a loved one, is located in mind as an independent exemplar-based representation (i.e., episodic trace). Within this cognitive architecture, attention plays a central role in the memorial processes of encoding and retrieval (Hintzman, 1976; Jacoby and Brooks, 1984; Logan, 1988; Smith and ZƔrate, 1992). As Logan puts it:
Encoding into memory is an obligatory, unavoidable consequence of attention. Attending to a stimulus is sufficient to commit it to memory. It may be remembered well or poorly, depending upon the conditions of attention, but it will be encoded. Retrieval from memory is [also] an obligatory, unavoidable consequence of attention. Attending to a stimulus is sufficient to retrieve from memory whatever has been associated with it in the past. Retrieval may not always be successful, but it occurs nonetheless. Encoding and retrieval are linked through attention; the same act of attention that causes encoding also causes retrieval. (1988)
Models based on these structural and processing assumptions have provided impressive explanatory power in cognitive psychology, leading to new insights into topics as diverse as object categorization (Medin and Schaffer, 1978; Jacoby and Brooks, 1984; Hintzman, 1986; Nosofsky, 1986); episodic memory (Hintzman, 1976; Jacoby and Brooks, 1984); and automaticity and skill acquisition (Logan, 1988). It comes as no real surprise, therefore, to learn that comparable accounts of mental life have also illuminated a variety of issues in social cognition, most notably how perceivers categorize and construe others (Gilovitch, 1981; Lewicki, 1986; Baldwin and Holmes, 1987; Andersen and Cole, 1990; Smith, 1990; ZĆ”rate and Smith, 1990; Smith and ZĆ”rate, 1992; Bodenhausen, Schwarz et al., 1995). Extending some of these ideas to the stereotyping domain, we propose that the application of an episodic retrieval model of person representation may also unlock one of the more perplexing mysteries of social perceptionānamely, how it is that perceivers can realize the selective stereotyping of others?
The essence of an episodic retrieval model of person memory is that mere exposure to a target is sufficient to locate an exemplar-based representation of the individual in mind (Smith and ZĆ”rate, 1992). These representations, of course, vary considerably in specificity. While some are detailed, content-rich entities (e.g., friends, lovers, relatives), others are more minimalist in stature, emphasizing only a few attributes or characteristics of the person in question (e.g., the women one passes each Friday while walking to work). Of particular relevance, however, is the form that these stored person-based representations are believed to take. As Smith and ZĆ”rate note, āExemplars in memory are records of the stimulus as interpreted or as processed on the previous occasion, rather than a veridical record of the stimulus information that was presentā (1992). Thus, it is not an exact copy of the person that enters memory, rather it is the person-as-interpreted by the perceiver that forms the basis of an exemplar-based representation (Fazio, 1986; Logan, 1988; Smith, 1989; ZĆ”rate and Smith, 1990).
Given the constructive nature of the person perception process (Bruner, 1957), it is easy to appreciate how factors such as current concerns, temporary task objectives, and prevailing prejudices may shape the contents of peopleās exemplar-based representations. Often we may not represent people in mind as they really are, but rather as we want them or expect them to be. This facet of memory function provides valuable insight into how exemplars may affect the person perception process. If, upon meeting a new neighbor, one infers that he is pretty rude, then this attribute-related information will be encoded in memory as part of the cognitive representation of him (Logan, 1988; Smith and ZĆ”rate, 1992). When, therefore, the representation is retrieved on a subsequent occasion, as indeed would be the case if one were to re-encounter the neighbor (or alternatively just to think about him), knowledge of his rudeness would be highly accessible and hence potentially applicable in any evaluation of the neighbor or his behavior (Smith and ZĆ”rate, 1992; Higgins, 1996). This phenomenon is a basic feature of episodic retrieval modelsātarget-based inferences undertaken at stimulus encoding are furnished to perceivers as an obligatory consequence of the exemplar retrieval process (Logan, 1988; Smith and ZĆ”rate, 1992).
In this respect, potentially one of the most influential inferences that perceivers make about others concerns the degree to which they are representative of the social groups to which they belong (Bruner, 1957; Brewer, 1988; Fiske and Neuberg, 1990). According to Medin and his colleagues (Medin and Schaffer, 1978; Medin and Smith, 1981; Medin, Dewey and Murphy, 1983), part of each exemplar-based representation in memory is information signaling the categorical typicality of the exemplar. Thus, upon witnessing an honest politician, part of the perceiverās resulting cognitive representation would be information indicating the politicianās low-goodness-of-category fit. This estimate of group representativeness, in turn, would have a critical bearing on his or her subsequent evaluations and treatment of the target. In particular, if the generic stereotype (i.e., politician) is deemed to be an inappropriate basis for characterizing the target, then perceivers are unlikely to apply it in their dealings with the politician. After all, if the shoe doesnāt fit, why wear it? As a result, impressions and recollections of the politician are likely to be decidedly nonstereotypic in flavor (Brewer, 1988; Fiske and Neuberg, 1990). At the same time, however, as goodness-of-category-fit has only been challenged for this particular group member, impressions of other politicians should still reflect the discriminatory elements of stereotypical thinking. Effects such as these (i.e., flexible stereotyping), importantly, sit comfortably within an episodic retrieval account of person representation (Smith and ZĆ”rate, 1992). Having encoded an exemplar as an atypical member of the group, perceivers will inevitably retrieve this information when the representation is activated on a subsequent occasion. Moreover, given that atypicality is tied to a specific exemplar, perceivers can continue to respond towards other members of the group (i.e., other exemplars) in a predominantly stereotype-based manner.
Our episodic retrieval model of person perception makes three basic assumptions. First, the allocation of attention to a group member is sufficient to locate an exemplar-based representation of the individual in mind (Logan, 1988; Smith and ZĆ”rate, 1992). Second, a fundamental component of the resulting representation is information signaling the targetās goodness-of-fit with respect to his or her salient group memberships (Bruner, 1957; Brewer, 1988). Third, subsequent allocation of attention to the individual is sufficient to retrieve the exemplar-based representation and its associated inferences from memory (Logan, 1988; Smith and ZĆ”rate, 1992). Ultimately, it is the categorical typicality of the person that determines the extent to which a stereotype is activated and applied when perceivers think about him or her. When goodness of fit is high, construals of the target will likely be highly stereotypic in implication. This is because goodness-of-category fit moderates the relative accessibility of stereotype-related material in semantic memory. If little information is available regarding a targetās goodness of category fit, then good fit is likely to be assumed by default, and stereotype activation and application will ensue (Fiske and Neuberg, 1990). Moreover, when the available information is ambiguous with respect to goodness of fit, biased assimilation processes may lead to the perception of a relatively high degree of fit, also resulting in stereotype use. However, when goodness-of-category fit it is unambiguously low (i.e., the stereotype provides poor predictive accuracy), then the accessibility of stereotype-related information should be correspondingly impaired.
Interestingly, functionally comparable effects have been observed when researchers assess the relative accessibility of category exemplars in memory. It is well known, for example, that in response to a category label (e.g., bird), perceivers find it easier to bring typical (e.g., sparrow) rather than atypical (e.g., ostrich) exemplars to mind (Rosch, 1978; Rothbart and Lewis, 1988; Rothbart et al., 1996). In this context, goodness-of-category fit moderates the ease of retrieval (i.e., accessibility) of exemplars from mind. Our contention is that similar effects may exist in person perception, with a targetās goodness-of-category fit determining the relative accessibility of generic stereotype-based material in memory. When category-fit is high, stereotypic information should be highly accessible; when category fit is low, it should be relatively inaccessible (Rothbart and John, 1985). In other words, construct applicability should gate the accessibility of stereotypic material in semantic memory (Collins and Loftus, 1975; Higgins, 1996; Neely, 1991).
Driving this analysis of person perception is our belief that stereotypes frequently function as mental tools (Gilbert and Hixon, 1991; Macrae, Milne and Bodenhausen, 1994; Macrae et al., 1997). Like any other tool, however, a stereotype will only...
