Implicit Attitudes and Stereotyping

Saera R Khan. Handbook of Racial & Ethnic Minority Psychology. Editor: Guillermo Bernai. Sage Publications. 2003.

Over the past several decades, psychologists have attempted to discern the relation between the conscious and unconscious and how these systems create and maintain categories for efficiently processing information (for further information, see Lakoff, 1987). Social cognition researchers are especially interested in documenting the process by which an activated construct produces social behavior. This assessment of the relation between implicit and explicit attitudes holds great promise in providing answers for many social problems. In particular, the past century has borne witness across the globe to both large- and small-scale acts of violence stemming from ethnic hatred. The recent Bosnian Muslim and Rwandan genocides are stark reminders of how negative prejudice can quickly escalate to violence. Psychologists want to understand how the seeds of hatred, prejudice, and stereotypes influence judgment and, in turn, behavior.

Although U.S. race relations have appreciably improved over the past several decades, both racial tension and prejudice continue to exist inside and outside of the United States (Dovidio & Gaertner, 1986; Kinder, 1986; Nier, Mottola, & Gaertner, 2000). From a researcher’s perspective, understanding the structure of racial attitude and its underlying factors may reveal viable solutions for reducing racial conflict. Unfortunately, theories or models of racial attitude and it relationship to behavior have become increasingly complex with improved research tools and techniques. Since 1990, at least 11 different racial attitude measures have been published in the United States (Biernat & Crandall, 1999). Many of these scales are constructed to measure at least one of the dimensions of racial attitude (e.g., endorsement of stereotypes, support or opposition to federal aid policies).

Furthermore, researchers must now reckon with the difficulties of appropriately measuring explicit racial attitudes while also documenting the role of implicit attitudes (e.g., Dasgupta, McGhee, Greenwald, & Banaji, 2000; Greenwald & Banaji, 1995). The rapid growth in technology and methodology in the past 10 years has increased our appreciation of the prominence of implicit attitudes and stereotyping within the mind’s toolbox. Exploring the structure of their relation is yielding insights into this phenomenon. However, many questions remain. Understanding the strength and shortcomings of current cross-cultural research in this area will inspire future research. This chapter seeks to provide an overview of the research on implicit racial attitudes. Also, unresolved issues, such as the relation between implicit and explicit attitudes, and the need for a unifying theory as well as future directions for research will be discussed.

Overview of Explicit Attitudes

Before discussing racial attitudes in depth, an extremely brief overview of attitude is needed. Attitudes are most generally referred to as one’s enduring evaluation of a particular entity. In particular, one’s affective reaction underlies the evaluation. This definition has been disputed; there is a debate over the unidimensional versus multidimensional structure of attitude (for a review, see Eagly & Chaiken, 1993). Although research has been equivocal on this issue, understanding racial attitudes from both perspectives has yielded valuable insights. From the unidimensional perspective, racial attitude can be thought of as one’s favorable or unfavorable evaluation of a particular racial/ethnic group. From the tripartite perspective, attitudes are classified into three components: cognition, affect, and behavior (Katz & Stotland, 1959; Rosenberg & Hovland, 1960; see also Eagly & Chaiken, 1993). Responses stemming from a person’s favorable or unfavorable beliefs, thoughts, or knowledge about the attitude object form the cognitive component. Affective responses stem from positive or negative feelings or emotions toward the attitude object. There is some disagreement in the literature regarding the need to make these components of attitude conceptually distinct (e.g., Fishbein & Azjen, 1975; Rosenberg, 1960; Zajonc & Markus, 1982; Zanna & Olson, 1994). Nevertheless, recent empirical work has been successful in distinguishing between affective and cognitive components of attitude (e.g., Breckler & Wiggins, 1989; Crites, Fabrigar, & Petty, 1994; Jackson et al., 1996; Stangor, Sullivan, & Ford, 1991). For instance, Haddock, Zanna, and Esses (1993) found that distinguishing between stereotypic beliefs, affect, symbolic beliefs, and prior personal experience significantly predicted both high and low authoritarians’ explicit attitudes toward homosexuals. In most current research, both implicit and explicit racial attitudes are conceptualized as unidimensional. Nevertheless, for both models, stereotypes consist of trait associations with a particular group. Although an immense amount of information has been uncovered through the unidimensional framework, revisiting the tripartite model might shed some light on several unresolved issues, such as the link between implicit and explicit attitudes. This point will be revisited in the conclusion. Prior to this, however, distinctions must be established between implicit and explicit attitudes.

Measurement of Explicit Attitudes

Explicit attitudes refer to one’s conscious beliefs. Self-report measures are thought to capture one’s attitude toward a particular object or group. In early attitude research, respondents reported their evaluation or feelings on a Likert-type scale ranging from strongly agree to strongly disagree. For example, questionnaires reflecting what is referred to as “old-fashioned racism” asked respondents to what extent they agreed with statements such as, “Blacks are generally more aggressive than Whites” (McConahay, 1986; see also Lambert, Cronen, Chasteen, & Lickel, 1996).

Assessing racial attitude with blatant measures became problematic as society experienced a shift in values. In the United States, the advent of the civil rights era increased social consciousness and encouraged egalitarian norms that rendered simple self-report measures (face-valid scales) as outdated and inaccurate. Many viewed segregation and other racist practices as antiquated and inconsistent with their egalitarian values. Still others, despite holding negative racial attitudes, were unwilling to report them due to normative pressures to appear nonprejudiced (Dovidio & Fazio, 1992). In response to this concern, instruments such as the Modern Racism Scale (MRS) were developed to unobtrusively measure racial prejudice (McConahay, 1986). The questions were worded so that people unknowingly revealed their true beliefs. According to this measure and surveys, there has been a dramatic decline in racial prejudice and endorsement of stereotypes over the past two decades (Devine & Elliot, 1995; Judd, Park, Ryan, Brauer, & Kraus, 1995). Unfortunately, other research suggests that the decline in prejudice may be related in part to faulty measures of racial attitude (e.g., Crosby, Bromley, & Saxe, 1980; Fazio, Jackson, Dunton, & Williams, 1995). For example, the MRS is not unobtrusive as once thought and has met with accusations that it confounds political conservatism and racism (Sniderman & Tetlock, 1986a, 1986b). In other words, respondents who are simply conservative can be classified as racist by this measure. Also, if the measurement of explicit attitudes via self-report were truly reliable, then we might happily conclude that a vast majority of Americans no longer hold any negative prejudices against ethnic minorities. Of course, the continuing humiliating acts of prejudice (Feagin, 1991) experienced by many ethnic minorities along with shocking reports of violent racial incidents remind us that the eradication of racism is not complete. The knowledge that self-report measures were not solely sufficient motivated researchers to develop novel research techniques for measuring attitudes.

Overview of Implicit Attitudes and Stereotypes

Interest in implicit attitudes has experienced a revival, with mounting evidence from cognitive, memory, and social research suggesting a distinction between implicit and explicit attitudes (Jacoby, Yonelinas, & Jennings, 1997; Shiffrin & Schneider, 1977). Implicit attitudes are defined as “introspectively unidentified (or inaccurately identified) traces of past experience that mediate favorable or unfavorable feeling, thought, or action toward social objects” (Greenwald & Banaji, 1995, p. 8). Research regarding automatic processing is analogous to implicit processing of information (e.g., Bargh, Chaiken, Govender, & Pratto, 1992; Fazio, Sanbonmatsu, Powell, & Kardes, 1986; Greenwald, McGhee, & Schwartz, 1998). Implicit or automatic processes are characterized as unconscious, uncontrollable, unintentional, and efficient. In contrast, explicit or controlled processes are described as conscious, controllable, intentional, and effortful (Bargh, 1989, 1994; Wegner & Bargh, 1998). The precise relation between these two types of attitude is disputed. However, recent evidence suggests there is flexibility within and perhaps overlap between both types of attitude (Bargh, 1989, 1994, 1996; Zbrodoff & Logan, 1986).

Implicit Stereotyping

A great deal of research and theory has focused on understanding implicit racial attitudes and its relation to using stereotypes for judgment (e.g., Devine, 1989; Dovidio, Kawakami, Johnson, Johnson, & Howard, 1997; Fazio et al, 1995; Greenwald & Banaji, 1995; Nesdale & Durkin, 1998; von Hippel, Sekaquaptewa, & Vargas, 1997; Wittenbrink, Judd, & Park, 1997). Stereotypes are the cognitive companion of prejudice and are generalized group-based traits used to judge individual members of a group (Ashmore & Del Boca, 1981). (For a detailed review on prejudice, sexism, and racism, please refer to Chapter 10 of this Handbook.) Just as implicit attitudes exist, implicit stereotypes do as well. Implicit stereotypes are defined as “introspectively unidentified (or inaccurately identified) traces of past experience that mediate attributions of qualities to members of a social category”(Greenwald & Banaji, 1995, p. 15). Features or traits consistent with the group stereotype can become unconsciously activated on encountering a member of that group (Greenwald & Banaji, 1995; Hamilton & Sherman, 1994; Smith, Fazio, & Cejka, 1996; cf. Gilbert & Hixon, 1991). Allport’s (1954) definition of prejudice is analogous to racial attitude in that it focuses on the affective reaction toward out-group members. Thus, stereotypes and attitude may be conceptually distinct but are related in influencing overall judgment and behavior (Eagly & Mladinic, 1989; cf. Gardner, 1973).

Devine’s (1989) groundbreaking studies on automatic and controlled processing and their relation to stereotyping provided the catalyst for much of the recent work in this area. Research on automatic and controlled processing has proceeded along the lines of the work on implicit and explicit processing; both seek to distinguish processing that occurs inside and outside one’s awareness and control. According to Devine, both high-and low-prejudiced persons exhibit racial stereotyping at the unconscious level but exhibit differences in judgment at the conscious level. Thus, if a low-prejudiced person is somehow unwilling or unable to exert mental effort when processing information about another person, stereotypes will be used for judgment. Critics of Devine’s work argue that the primes used in her work mingled negative stereotypes with category activation and that stereotypical judgments are not inevitable at the unconscious level for those low in prejudice (Fazio et al., 1995; Lepore & Brown, 1997). Nevertheless, she demonstrated how the activation of implicit stereotypes influenced behavior.

Measurement of Implicit Attitudes

The development of response latency measures by cognitive psychologists introduced a novel way of measuring implicit attitude. In particular, semantic priming procedures have reliably demonstrated that paired associations facilitate responses. People are faster to respond to pairs of words, images, or even concepts that are already associated in memory than those that are not. Activation of implicit attitude occurs too rapidly for it to be solely driven by conscious processes, but it can be measured by these techniques (Banaji & Hardin, 1996; Bargh et al, 1992; Fazio et al., 1986). The logic of using response latency measures is that perceivers typically rely on highly accessible constructs activated from memory for judgment than newly constructed associations not consistent with our attitudes. This method of measurement is powerful because people can be unaware of or reluctant to articulate their beliefs. Thus, when using these measures for assessing racial attitudes, we can indirectly measure people’s underlying attitudes toward concepts without directly asking them.

Research on Implicit Attitudes and Stereotyping

Over the past 15 years or so, numerous studies have been published on implicit attitudes and stereotyping. It is impossible to review all of them, but a few noteworthy studies employing different methods to assess implicit attitudes will be discussed. Special attention will be paid to the development and application of the implicit association test (Greenwald & Banaji, 1995; Greenwald et al., 1998). The Implicit Association Test (IAT) has been used extensively to document racial implicit attitudes (e.g., Greenwald et al., 1998; Neumann & Seibt, 2001; Neumann, Totzke, Popp, & Fernandez, 2000; Ottaway, Hayden, & Oakes, 2001; Rudman, Greenwald, Mellott, & Schwartz, 1999).

Response time measures (Dovidio, Evans, & Tyler, 1986; Gaertner & McLaughlin, 1983) and semantic priming methodology (Devine, 1989) are alternative ways to measure racial attitude. The basic premise of priming studies is that the prior presentation of a stimulus can influence the interpretation of novel information (Bruner, 1957). The primed information (e.g., words or images) activates a construct in memory that remains accessible while perceiving novel information. Higgins, Rholes, and Jones (1977) demonstrated that participants judge a character in a story in more favorable terms if they were recently primed with positive traits than if they were primed with negative traits.

Priming participants with words designed to activate the category for a particular group allows an assessment of implicit racial attitudes. Presumably, not only is the category activated, but the attitude toward that group is as well (Fazio et al., 1986). For example, to activate attitude toward Black people, Lepore and Brown (1997) subliminally presented White British participants with words or group labels associated with the racial category. High-prejudiced participants subsequently rated an ambiguously described person as aggressive and unreliable, whereas low-prejudiced people judged the same person in more favorable terms. Thus, the mere presentation of semantic group labels successfully activates the category and attitude toward that group.

Along similar lines, Fazio et al. (1995) employed faces instead of labels for their priming task. In addition, the study compared participants’ responses toward African Americans obtained by the Modern Racism Scale and a priming task. White American participants completed the MRS scale and at a later session were presented with Black and White male and female faces on a computer screen. Immediately after the presentation of a face, participants judged the valence of an adjective word flashed on the screen (good vs. bad). The participants showed relatively faster judgments when White faces preceded positive adjectives and Black faces preceded negative adjectives. Also, MRS scores were not correlated with participants’ responses on the priming task and were influenced by the presence of a White or Black experimenter. Most important, this study provided a severe blow to the claims made that the MRS is an unobtrusive measure of prejudice. In another study, Wittenbrink et al. (1997) used a semantic priming task in which White American participants were first subliminally primed with the word Black or White and then asked to judge whether a string of letters presented on a computer screen constituted a real word or not (i.e., word or not word). The actual words were positive and negative stereotypical traits of Whites and Blacks. Participants also completed several explicit measures of prejudice. Participants’ responses were facilitated when negative stereotypical words were preceded by the Black prime and the positive stereotypical words preceded the White prime. The explicit measures were highly correlated with each other and with implicit prejudice. Unlike Fazio et al.’s (1995) study, this experiment provides evidence that implicit attitudes are indeed related to explicit attitudes.

Research Using the Implicit Association Test

Implicit Association Test. Greenwald et al. (1998) developed the IAT as a measure of implicit attitudes. Participants’ attitudes are inferred by their response latencies in categorizing psychologically congruent and incongruent paired target concepts (e.g., White + pleasant; Black + pleasant). Response latencies should be faster when categorizing implicitly associated concepts compared to pairs that are not associated a priori. For the critical trials, two concepts are assigned to a single response key (e.g., Black names and pleasant words), whereas the other two contrasting concepts are assigned to a different response key (e.g., White names and unpleasant words). Participants are asked to quickly press the response key that correctly categorizes the stimulus word presented on the screen. The IAT effect is calculated by subtracting the response latency of the congruent paired concept from the incongruent paired concept. The IAT employs numerous methodological controls for variables that could influence response latencies.

Research Studies Employing the IAT. The first published studies using the implicit association test demonstrated its validity in measuring implicit associations between affect or valence (pleasant vs. unpleasant) and categories (Greenwald et al., 1998). In Experiment 2, Korean American and Japanese American college students categorized typical Korean and Japanese surnames according to their ethnic group by pressing the appropriate computer key that was also used to differentiate pleasant or unpleasant words. Two paper-and-pencil questionnaires were used to assess explicit attitudes. For the feeling thermometer measure, participants rated their warmth or coolness toward the in-group and out-group. Participants also rated each group along a 7-point scale, anchored at each end by polar-opposite adjective pairs (e.g., beautifulugly). Next, participants answered questions assessing cultural immersion in their respective ethnic group. The results revealed IAT effects for both groups. That is, Korean Americans’ responses were faster when the category Korean shared the same response key as pleasant and Japanese with unpleasant. Japanese Americans showed the same effect for their own group. An additional effect was found for cultural immersion. Those participants reporting greater immersion in their culture showed larger IAT effects than those acculturated to the United States. Neither of the explicit measures revealed this difference among participants.

Experiment 3 was conducted similarly to the previous experiment, except that White American male and female college students categorized Black and White names along with pleasant and unpleasant words. Participants showed a clear implicit preference for Whites over Blacks on this measure. Explicit attitude measures did not reveal this preference, nor did these measures predict IAT responses.

Criticisms of Implicit Attitude Research. In response to Greenwald et al.’s study (1998) as well as others using the IAT, critics argued that Whites’ unfamiliarity with African Americans as well as the African American names used in Experiment 3 may account for the large IAT effects obtained. Studies conducted by Ottaway et al. (2001) and Dasgupta et al. (2000) attempt to rule out some of these alternative explanations. Ottaway and colleagues controlled for subjective familiarity and frequency for Black, White, and Hispanic names. Dasgupta et al. (2000) statistically controlled for name familiarity and used pictures of unfamiliar Blacks and Whites as well as names for their stimuli. Both studies reduced the overall magnitude of the IAT effect but still demonstrated Whites’ implicit preference for their own group over Blacks and Hispanics.

Employing these methodological and statistical controls rules out one version of familiarity, but more studies examining different concepts of familiarity remain necessary. Familiarity with names is not equivalent to familiarity with group members, nor is unfamiliarity with particular White and Black faces equivalent to unfamiliarity with the category and its actual members. It is important to keep in mind that the amount and type of actual social interaction may play a large role in moderating implicit attitudes. Social context must not be neglected. Recently, Dovidio, Kawakami, and Gaertner (2000) reported that intergroup contact along with recategorization into a common group identity reduced implicit stereotyping.

Thus far, it is unclear whether implicit and explicit attitudes are related. Some studies reveal a correlation between implicit and explicit measures of prejudice (e.g., Wittenbrink et al, 1997), but others do not (Fazio et al, 1995; Greenwald et al, 1998). These mixed findings may be due to differences in methodologies employed (for further discussion, see Brauer, Wasel, & Niedenthal, 2000). At this point, we do know that both implicit and explicit racial attitudes do exist and can be measured. According to Dovidio et al. (1997), explicit attitudes should not diminish in relevance, nor should we view implicit attitudes as somehow more “true” than explicit attitudes. Rather, each type of attitude might influence different aspects of behavior. Advancement of research and theory is clearly needed in this area so that we may better understand these relationships as well as the possibility of modifying attitudes and behaviors.

Implicit Learning

Social psychologists have paid much greater attention to establishing the ubiquity of implicit attitudes as opposed to documenting their development. Unlike explicit learning, implicit learning is largely unconscious, unintentional, and not limited by cognitive capacity (Berry & Broadbent, 1988; Broadbent, 1989; Hayes & Broadbent, 1988; O’Bnen-Malone & Maybery, 1998; Reber, 1992). Although the information is acquired unconsciously, whether the conscious has access to implicit knowledge is unclear (Hasher & Zacks, 1979, 1984; Lewicki, Hill, & Czyzewska, 1992; Reber, 1989). Two primary models account for implicit learning. Schneider and Shiffrin’s (1977) model describes how information becomes implicit through frequent activation of associated constructs in memory. In this case, stereotypes become implicit through activation on frequent encounters with a stimulus from a particular category. For example, if a male perceiver associates the trait passive with female, the two concepts of female and passive will eventually become implicitly associated if frequently used when interacting with females.

Alternatively, Hasher and Zack’s (1979) model posits that an explicit learning stage is not always necessary; implicit learning is hardwired and part of our information-processing system. Moreover, these innate processes “should be widely shared and minimally influenced by differences in age, culture, education, early experience, and intelligence” (Hasher & Zacks, 1979, p. 360). Differences in opinion among cognitive psychologists do exist concerning the boundary conditions of this type of implicit learning (for further discussion on this issue, see O’Brien-Malone & Maybery, 1998). In general, this model points to the extreme difficulties in tracing how the actual content or associations forming the basis of implicit stereotypes are developed. In fact, racial and gender stereotypes may be formed early in childhood just as other social categories are created.

What is potentially most disturbing about implicit learning is that we often do not and cannot anticipate what information people use to form these associations. In their essay, Miller and Swift (2000) recount a popular riddle to demonstrate this point:

A man and his young son were in an automobile accident. The father was killed and the son, who was critically injured, was rushed to the hospital. As attendants wheeled the unconscious boy into the emergency room, the doctor on duty looked down at him and said, “My God, it’s my son!” What was the relationship of the doctor to the injured boy?

That people are stumped when solving this riddle reflects the difficulty when thinking of a woman as a doctor. In a cognitive sense, people are implicitly associating doctor and male.

A personal example illustrates this point. While playing with a toy car, my 3-year-old twin nephews stated that dads always sit in the driver’s seat and moms sit beside them. Worried that somehow they were making a larger inference about gender hierarchical relationships, I suggested that Mom can drive the car and Dad can sit in the passenger’s seat. Their refusal to accept my suggestion was especially puzzling given their reality. Eventually, I realized that it was not necessarily the media or another child’s influence that led to their conclusion. Their father always drove the car when both parents were present. Thus, their perception of Mom’s driving as more of an exception to the rule (she drives only when Dad is not around) than the norm is rooted in their experience. What can never be determined is how this information (and other examples we unwittingly provided) entered their unconsciousness and the sorts of implicit associations made.

Implications for Implicit Learning

We can see how associations are made without deliberate intent or awareness by the perceiver or communicator. Economic and social realities might provide one way for the formation of racially implicit associations by majority perceivers. In the United States, Hispanic and African Americans are underrepresented in the middle and upper socioeconomic classes. The majority of middle- and upper-class Whites live in rural or suburban areas away from economically disadvantaged minorities. For many Whites, interactions with African Americans or Hispanics are often in the context in which the minority member is a low-paying service provider. Limited interactions such as these as well as negative media portrayals of minorities might provide the foundation for implicit associations. Unbeknownst to both groups of perceivers, an implicit association might be formed between the racial/ethnic group category and a negative trait, such as poverty. Now, the overall valenced association formed for the minority group is negative or unpleasant. This negatively valenced association may lead to the storage of other unpleasant traits within this category. How negative traits tend to be associated with other irrelevant negative traits is analogous to the halo effect (Thorndike, 1920). The halo effect describes the tendency to form impressions of others by associating positive traits with other unrelated positive traits on the basis of limited information. Just as some people benefit from the halo effect (e.g., physically attractive persons), others may suffer from the opposite effect (for a more complete discussion on the halo effect and implicit attitudes, see Greenwald & Banaji, 1995). Thus, seeing or interacting with out-group members in a limited or impersonal fashion may facilitate undesired implicit attitudes that are subsequently activated on encountering members of the group. In no way is the obvious role of plain racism in the formation of implicit attitudes or stereotypes ruled out. What is being argued, however, is that in addition to racism, well-meaning people do acquire information and form personally deplorable associations without awareness or intent. The extent to which these associations are used for judgment and behavior of any real consequence is debatable.

In-Group Implicit Associations

Extending this argument further, the notion of implicitly learning prejudice is provocative because it implies that all of us, including ethnic minorities, may have learned to associate negative concepts with minority groups and positive concepts for the powerful majority group. Although published research on minority members’ own implicit racial stereotypes does not exist, a closely related study by Jost, Pelham, and Carvallo (2000) provides evidence of out-group favoritism in Hispanics and Asian Americans. In this study, Hispanic and Asian American students preferred to participate in a “getting acquainted” study with a White stranger than someone whom they believed belonged to their own ethnic group. Although this study did not strictly examine implicit processes, their behavioral measurement was unobtrusive and unlikely to have aroused suspicions by participants. Another series of studies examining gender and racial groups demonstrates how group status affects implicit associations made by majority perceivers. In these studies, groups accustomed to having more power (i.e., males and Whites) over their respective out-groups showed varying favorable and unfavorable implicit attitudes as the power dynamic changed in the experiment (Richeson, 2000).

Research examining implicit gender stereotypes reveals different results. Male and female participants associated favorable gender stereotypes for the self and their own gender group (Rudman, Greenwald, & McGhee, 2001). According to this study, our self-esteem needs may protect us from implicitly stereotyping our own in-group. Given these mixed results, a closer look at the social stigma attached to particular group identities and its relation to implicit self-stereotyping is needed (for research on stigma and social groups, see Crocker, 1999; Dovidio, Major, & Crocker, 2000; Swim & Stangor, 1998).

The Role of Language

The acquisition and use of language can be viewed as another primary force by which implicit associations are formed. According to the noted linguist Edward Sapir (1963), language reflects the societal hierarchy and provides psychological status to subgroups. In this case, the powerful group creates terms used to describe less powerful groups. Feminists argue that our use of man to describe all humankind is a blatant reminder of patriarchal structure rather than a term of convenience without cognitive ramifications. In fact, research by Banaji and Hardin (1996) and McConnell and Fazio (1996) demonstrated that gender primes automatically activate stereotyping for both males and females. Thus, the mere presentation of a word, such as nurse, automatically activates the concept of female. Moreover, generic masculine terms such as human also produced an automatic activation of the concept male. Thus, group labels do carry an association with power status and favorability.

Allport (1954) noted the sway of language as well in his chapter titled “Linguistic Factors.” In this chapter, he describes how ethnic labels used by majority perceivers to describe out-group members are often tainted with negative connotations. In fact, he concludes, “Prejudice is due in large part to verbal realism” (Allport, 1954, p. 187). An example of language evolving to reflect a shift in societal values and hierarchy is the ethnic/racial labels used to refer to African Americans. These labels changed as African Americans gained societal power and the rest of society decided that offensive labels are no longer acceptable. Although there is no single agreed-on term, African Americans, Native Americans, and other groups understand that reclaiming their right to choose their group labels is a rejection of subordination. In fact, choosing one’s own group label may have even greater implications than once thought. Labels may serve as markers for which category will be activated and used by the perceiver. For example, in Germany, the label Black does not carry a negative connotation, as does foreigner. Neumann et al. (2000) used the IAT to examine how Germans react to the same exemplars identified by different category labels. White German students show an implicit preference for Whites over Blacks when the category label for Blacks was “foreigners.” However, no such preference was found for the Black versus White IAT when the category label was “Black.” One far-reaching implication of these results is that arguments over the “political correctness” of certain types of language should not be dismissed as petty and trivial. Terms deemed as politically correct may potentially activate different categories than their politically incorrect counterparts.


Social and cognitive psychologists’ interdisciplinary efforts have benefited the entire field of psychology. Theoretical and methodological advancements have provided the foundation for interesting and important research. With these advancements, the primary research goal must remain the development and advancement of theories with these new tools. Novice and experienced researchers should address whether their findings fit into existing theoretical frameworks or whether modifying or creating new theories is required. Along similar lines, the structure of attitude as well as the relation between implicit and explicit attitude needs additional scrutiny. At this point, two contrasting models account for the relation (or lack thereof) between the two types of attitude, and both are empirically supported (Brauer et al., 2000). However, improvement in both types of attitude measures is most likely necessary to form any conclusions. Revisiting the tripartite model for measuring both types of attitude might be one improvement. Conceptualizing both types of attitude as unidimensional may have erroneously led to an emphasis of one component of attitude over the others. For example, some explicit measures tap into more of the cognitive than the affective component, whereas the reverse exists in some implicit measures (Baggenstos, Khan, & Ottaway, 2001).

Finally, the crucial link is establishing the connection between implicit racial attitude and social behavior. Dovidio et al. (1997) showed that implicit activation influenced largely uncontrollable behaviors (e.g., blinking and visual contact). These results encourage further investigation of these types of effects. Establishing the boundary conditions for when implicit and explicit attitudes influence judgment and behavior will provide the foundation for developing successful interventions aimed at reducing prejudice.

It is difficult to form general conclusions about the link between implicit attitudes and behaviors because of the varying methodologies used. The area would greatly benefit from researchers systematically investigating the differences in results obtained by the use of different methodologies. One key difference in the various methodologies employed in these studies is whether or not participants are aware of the activation of implicit attitude. Classic priming studies have documented the effects of awareness on behavior. Awareness of primes can lead participants to adjust and correct for the activation, which can lead to contrast instead of assimilation effects (e.g., Herr, Sherman, & Fazio, 1983). Bargh and others demonstrated that subliminal activation of constructs influenced controllable social behaviors (Bargh, Chen, & Burrows, 1996). In this case, participants did not “correct” for the activation in their behaviors because they were unaware of the attitude activation itself. Studies using the IAT demonstrate how despite awareness of activation, participants’ performance on a largely uncontrollable task (i.e., computer response time task) is still influenced by implicit activation of attitude. In this case, awareness does not help participants to “correct” for the influence of activation in their judgments. Implicit attitude research would benefit from integrating theories and research examining stereotype activation and inhibition (e.g., Bodenhausen & Macrae, 1998; Bodenhausen, Macrae, & Garst, 1998; Lambert, Chasteen, Khan, & Manier, 1998). Specifically, future research needs to examine whether the awareness of the activation of implicit attitudes influences overt and covert controllable behaviors in social interactions.

In any case, understanding prejudice and stereotyping at the implicit level is of incredible import. If implicit attitudes influence judgment and behavior outside the perceiver’s awareness, then perhaps it explains why there is often disagreement between majority and minority perceivers regarding the specific acts and prevalence of racial discrimination today. Members of minority groups continue to feel the brunt of racial discrimination in both subtle and flagrant forms. The continual development of new methods for assessing subtle forms of prejudiced behaviors and attitudes may ultimately reduce the difference between racial groups in their perception of the frequency and extent of racial prejudice and discrimination. Members of the majority racial group often fail to appreciate the full effects of their unintentional racism (e.g., aversive racism) (Dovidio & Gaertner, 1999). For prejudice intervention techniques to succeed, recognition of one’s own prejudices and a commitment to eliminating them must occur (Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000; Moskowitz, Gollwitzer, Wasel, & Schaal, 1999). It is hoped that this program of research will encourage the development of equitable racial attitudes and behavior by capitalizing on people’s strong egalitarian values and their genuine desire to be nonprejudiced.