Hostname: page-component-74656f9884-c52kr Total loading time: 0 Render date: 2026-11-09T10:06:08.068Z Has data issue: false hasContentIssue false
  • English
  • Français

Racial Priming Revived

Published online by Cambridge University Press:  26 February 2008

Tali Mendelberg
Tali Mendelberg
Affiliation:
Princeton University (talim@princeton.edu)
Rights & Permissions [Opens in a new window]

Abstract

This comment addresses the growing controversy over the effects of implicit racial messages in politics. Many scholars find evidence that these implicit messages work and that they have racializing effects. However, the biggest study to date finds that racial messages—implicit or explicit—have no effects. In this paper I conduct a thorough review of several relevant literatures in order to adjudicate between these competing claims. I find that the large study's null findings conflict with 17 public opinion experiments involving over 5,000 subjects, 2 aggregate studies, and a large social psychology literature. Using different methods, samples, and settings, these studies show that racial cues do in fact racialize opinion. I explain the large study's null results by noting that its participants perceived only small differences across messages, that racial predispositions were measured just before exposure to the ad, thereby neutralizing the effect of the ad's racial cue, and that WebTV studies such as this one have failed to provide many subjects with their assigned ad. Thus, the weight of the evidence heavily favors the racial effect of racial cues and messages. I offer several directions for future research on racial communication and politics.Tali Mendelberg is Associate Professor of Politics at Princeton University (talim@princeton.edu). She is the author of The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality (Princeton University Press, 2001). She wishes to thank Oleg Bespalov and Dan Cassino for research assistance, Adam Berinsky, Claudine Gay, Martin Gilens, Vince Hutchings, Jon Krosnick, Shawn Rosenberg, and Nick Valentino for helpful feedback, and the Bobst Center for Peace and Justice at Princeton University and the Center for Advanced Study in the Behavioral Sciences for research support.

Information

Type
EXCHANGE
Information
Perspectives on Politics , Volume 6 , Issue 1 , March 2008 , pp. 109 - 123
Copyright
© 2008 American Political Science Association

For some time now, scholars have been documenting the power of whites' negative racial predispositions. Dozens of studies, conducted with surveys, lab experiments, or observational methods, have found that white Americans' political preferences are shaped by predispositions that characterize African Americans as lazy, welfare-dependent, violent, or demanding special favors.1

See Sears, Sidanius, and Bobo 2000 for a review.

As Hutchings and Valentino sum up, “racial attitudes, broadly conceived, represent one of the fundamental influences on contemporary mass political attitudes”.2

Hutchings and Valentino 2004, 392.

More recently, this literature has asked how the information environment reinforces or primes negative attitudes about groups with lower life chances in American society.3

Berinsky and Mendelberg 2005; Clawson and Trice 2000; Entman and Rojecki 2000; Gilens 1999; Huddy 1994; Kellstedt 2003; Kinder and Sanders 1996; Nelson and Kinder 1996; Sniderman and Piazza 1993.

Particular attention has been paid to the racial priming hypothesis, which predicts that cues in the information environment activate or deactivate citizens' racial predispositions, with consequences for citizens' preferences about policy and vote choice.4

Scholars also investigate how negative views diminish or are neutralized (e.g., Gilliam, Valentino, and Beckmann 2002; Glaser 2002; Hurwitz and Peffley 1997a; Kellstedt 2003; Mendelberg 2001; Peffley and Hurwitz 1993).

However, some scholars posit that the U.S. has experienced enough progress in race relations that negative references to African Americans are scarce and often marginalized beyond the ability to carry an impact. For example, Thernstrom writes, “the American political scene is filled with fringe candidates peddling lines to which no one listens, and in many communities racists operate on the fringe.5

Thernstrom 1987, 204.

According to Thernstrom, implicit appeals do not work either, since they are too subtle: “subtlety in campaigns risks political ineffectiveness; only overt appeals are reliably heard”.6

Thernstrom 1987, 202.

A recent study appears to lend support to her argument. Huber and Lapinski have conducted by far the largest study of racial messages, and find that no racial message, whether overt or subtle, has much effect.7

Huber and Lapinski 2006.

They conclude that people “reject explicit appeals outright” while implicit ones “are no more effective than explicit ones in priming racial resentment in opinion formation”.8

Huber and Lapinski 2006, 421.

This claim matters because it overturns the current understanding of the lingering effects of negative racial predispositions in politics. If racial messages evoke racial predispositions, then we can conclude that one of the important reasons that race continues to shape politics is the communication process from elites to citizens. That is, citizens respond to elite-generated or media-conveyed messages that maintain or enhance African Americans' disadvantage. Conversely, we can conclude that if elites abandon negative racial messages, then citizens' racial views will adjust accordingly, with racial progress following as a result. If racial messages do little to cue racial predispositions, however, then we are left without the key political explanation of the continuing effects of racial orientations in the political life of citizens. The effects of racial predispositions, then, appear impervious to influence from political forces, with less potential for racial progress.

As important as is the dynamic of racial politics in the U.S., the issues at hand are broader than that. Scholars have now extended the investigation of communication and group divisions to a variety of social groups in the U.S. and across countries.9

For U.S., see Berinsky and Mendelberg 2005; Fraga and Leal 2004; Hutchings et al. 2004. For across countries, see Chandra 2004; Dickson and Scheve 2006; Gibson and Gouws 2002.

At stake, then, is the general question of how and why, as well as whether, elite communication leads to more or less social conflict in politics.

The Controversy Outlined

To study the impact of negative racial cues, investigators first assess citizens' exposure to these messages, with most studies using experimental design to randomly assign a viewer either to a stereotypical racial cue (e.g., a black criminal) or to a counter-stereotypical or a nonracial cue (e.g., a white criminal, or a cue on an issue unrelated race, such as the environment). They then measure viewers' subsequent preferences for relevant candidates or policies, and test for evidence that the racial cue caused a racialized reaction. The racialized reaction can take several forms. In a common version of this design, investigators assess citizens' level of longstanding negative racial predispositions (stereotypes, resentments, etc.), and test for evidence that the racial cue caused racial priming, defined as the increased impact of negative racial predispositions on relevant candidates or policy opinions.10

Put differently, in this version, researchers test whether the negative racial cue especially influences people already inclined to stereotype or resent African Americans.

This general approach to the impact of negative racial cues applies to studies of implicit racial cues or messages as well. Implicit messages are distinguished from explicit messages, which use “racial nouns or adjectives to endorse white prerogatives, to express anti-black sentiment, to represent racial stereotypes, or to portray a threat from African Americans.”11

Mendelberg 2001, 8.

Implicit messages have a similar content but use a more subtle and indirect communication style by omitting racial nouns and adjectives such as “blacks” or “racial”. The racial message thus appears peripheral and less objectionable. The message works implicitly either because it uses negative images of the target group that appear coincidental to the main message, which is delivered without racial words; or because it uses words that have racial associations but are not racial nouns or adjectives and therefore do not appear to focus on race.12

See Mendelberg 2001, ch. 1.

Examples are a message about a criminal that includes an image of a black man with criminal traits, or the phrase “inner city,” which has been paired often with predominantly black, run-down residential areas. The claim of the implicit literature is that egalitarian norms now prohibit the use of any communication that would be perceived as racist.13

Berinsky 2002; Kuklinski et al. 1997.

Elite communication of racial stereotypes thus takes place in a more subtle fashion.14

Edsall and Edsall 1991; Jamieson 1992; Kinder and Sanders 1996; Reeves 1997.

These more subtle messages work because their racial cues can operate outside the perceiver's awareness; the audience attends to the racial content just enough to process it, but does not attend to the fact that the content is racial.

This implicit/explicit model has been most closely tested with a research design that contrasts implicitly racial messages against explicitly racial messages (which make the racial content obvious by using categorical racial nouns), nonracial messages devoid of any racial association, or counter-stereotypical messages that cast whites in a negative light or blacks in a positive one. This model is perhaps most elaborated and most thoroughly tested in Mendelberg, the subject of Huber and Lapinski's critique.15

Mendelberg 2001.

Mendelberg conducted three experiments and a survey study and found consistently that implicitly racial messages were much more effective than any other type at eliciting racial priming.16

Mendelberg 1997; 2001.

The Huber and Lapinski article, by contrast, finds no evidence that any race cues have any statistically significant effects. It concludes that implicit appeals are no more effective at priming racial predispositions than are explicit appeals, and that neither one is effective relative to a nonracial control message. According to Huber and Lapinski, there is a small exception among low education respondents, whose resentment is slightly activated by both implicit and explicit messages (although this result is not statistically significant).17

Huber and Lapinski 2006, 436.

But the authors conclude that “more generally, across all four policy areas and for both education levels, there are no instances in which the opinions of those viewing the implicit and explicit appeals are distinguishable” and that “one cannot evoke more conservative responses to opinion questions by covertly appealing to underlying anti-black predispositions”.18

Huber and Lapinski 2006, 436.

They explain the discrepancy with Mendelberg by referring to her smaller, less representative, and geographically narrower sample, in contrast with Huber and Lapinski's large, representative national sample.19

Mendelberg 2001.

They further argue (in error) that Mendelberg did not use a nonracial control condition.20

Huber and Lapinski 2006, 436.

I make two main points in response, each of which I will elaborate below. The first point is that Huber and Lapinski's findings—that racial cues do not generate any racial effects—do not square with several large, even massive, literatures showing that racial cues do affect opinion. These literatures also repeatedly replicate the more specific claim rejected by Huber and Lapinski—that racialization often works implicitly. Huber and Lapinski's null findings are a large anomaly, and require explanation.

The second point is that Huber and Lapinski's evidence shows that their experimental manipulations did not work. This is true not only in the sense of null results, but in the sense that subjects do not appear to be able to tell Huber and Lapinski's messages apart clearly. In addition, there is reason to believe that many, perhaps most, of their subjects in fact were not exposed to any message. If two-thirds of responses on the key causal variable consist of measurement error, then the variable's impact may be severely under-estimated. Moreover, Huber and Lapinski's is the only study that measures racial predispositions just before exposing subjects to ads; in doing so it likely primed subjects in every condition and neutralized the distinctive effects of the implicit racial ad.

Evidence for a Dynamic View of Race

Social Psychology Studies

The largest literature in conflict with Huber and Lapinski's null findings is found in social psychology. Hundreds of studies demonstrate that race cues are primed by racial stereotypes. Moreover, this literature also makes the specific point critiqued by Huber and Lapinski—that racial cues often work in an implicit way, that is, without awareness. This is an important development because it comes on the heels of studies showing a dramatic drop in explicit measures of racial stereotypes and prejudice, with social desirability a heavy influence.21

Devine and Elliot 1995; Devine et al. 1991, Study 3; Dovidio and Gaertner 1986; Kinder and Sanders 1996; Schuman et al. 1998; Sears et al. 2000.

In counterpoint to that decline, hundreds of studies have shown that less overt measures—measures not easily available for self-regulation—reveal high levels of racial stereotyping and bias. The conclusion is thus not that racial stereotyping and bias have disappeared, but rather that they continue to operate in a more subtle fashion.

There are many implicit cue designs, and space permits me only a partial review here. One common type of implicit priming study exposes subjects to a subliminal racial cue, such as the words “black” or “white.” The exposure is too quick to register consciously; that is, when asked, subjects are not able to say that they saw these words. Nevertheless, subjects exposed to the word “black” and who are asked to indicate whether subsequent letter strings represent real words or not are quicker and more accurate in classifying negative than positive stereotypical words; the reverse is true for those exposed to “white”.22

Wittenbrink, Judd, and Park 1997.

In a related design, subjects cued with images of blacks are more likely to misidentify innocuous objects as weapons, while white images lead to mistaking weapons for innocuous objects.23

Payne 2001; Payne, Lambert, and Jacoby 2002.

In turn, subliminally cuing objects or concepts associated with black stereotypes leads to more attention to black faces, with the relationship strongest for more stereotypical black faces.24

Eberhardt et al. 2004.

Despite declines in overt stereotyping and Americans' near-universal endorsement of equal treatment under the law, exposure to black faces leads to discrimination in murder cases when race is cued by black-on-white murder (that is, when race is salient because of the apparently racial nature of the crime). In actual, real-world cases where this race cue is present, defendants with more stereotypical black faces are sentenced to die at more than twice the rate of defendants with less stereotypical black faces (58 percent versus 24 percent), controlling on mitigating and aggravating circumstances and murder severity.25

Eberhardt et al. 2006.

These effects are for real juries deciding real cases and sentencing real defendants.

Perhaps the most famous of these designs is the Implicit Association Test (IAT).26

Greenwald and Krieger 2006.

Subjects are asked to distinguish among black and white faces, and among positive and negative words, by pressing different computer keys. They are then asked to press a single key for a face paired with a word. Bias is measured by the difference between the speed of response to the black-positive versus white-positive pairings, and to white-negative versus black-negative pairings. All demographic groups show much more racial bias on this measure than they reveal in explicit measures, with faster reactions to white-positive and black-negative.27

Greenwald and Krieger 2006.

Sixty-one studies have shown an association between the IAT and relevant behaviors. Of particular interest is its high predictive validity with relevant behaviors outside people's normal awareness or control, such as brain scans and automatic eye movements or body language.28

Poehlman et al. 2005.

Implicit stereotyping has been replicated over time, situation, stimuli, samples and target groups.29

Banaji and Greenwald 1994; Bargh 1996; Bargh and Chen 1997; Bargh, Chen, and Burrows 1996; Correll et al. 2002; Devine 1989; Dovidio, Evans, and Tyler 1986; Dovidio et al. 1997; Fazio et al. 1995; Fiske et al. 1999; Krueger 1996; Macrae, Milne, and Bodenhausen 1994; Olson and Fazio 2004; Payne 2001.

These results are not an artifact of student samples or artificial decisions. The findings have been replicated or validated with, for example, judgments by police officers, juries, and judges, with some studies examining real-world decisions.30

Blair, Judd, and Chapleau 2004; Eberhardt et al. 2004; Eberhardt et al. 2006; Graham and Lowery 2004; Plant and Peruche 2005.

So the evidence suggests that racial cues racialize people's responses even though—and especially when—people are unaware that they have been exposed to racial cues, and that their decisions are shaped by racial stereotypes or attitudes. These findings, along with the dramatic decrease in explicit stereotyping, lend support to the claim rejected by Huber and Lapinski: that racial messages are likely to work, and to work implicitly. However, because these studies tend to deal with cues and responses that are often removed from political decisions, their evidence is insufficient. I therefore turn to studies of public opinion.

Studies of Media and Public Opinion

The specific literature of greatest relevance is the set of studies of the impact of racial media messages on public opinion (see table 1 for a summary of sample characteristics and other relevant information).31

In all the experiments described here subjects were randomly assigned to a condition.

The implicit cueing results reviewed above are also obtained with realistic news messages. For example, Johnson et al conducted an experiment to test for implicit stereotyping in criminal justice judgments.32

Johnson et al. 1997.

They first exposed subjects to newspaper stories featuring either violent crimes or crime-irrelevant stories, with suspects' race unspecified. Then subjects read a description of a violent crime. The investigators used photos to manipulate the defendant's race to be black, white, or unspecified (no photo). The violent media stories increased internal attributions of the defendant's behavior for black but not white defendants, increasing the likelihood of guilty verdicts. The violent crime stories primed racial stereotypes that were applied in judgments of black defendants.

Public opinion studies finding racial effects of race cues

A more political application of the implicit stereotyping literature is found in Terkildsen, whose findings support the notion that racial predispositions can be activated out of awareness through an implicit process.33

Terkildsen 1993.

She exposed subjects to a manufactured newspaper story describing a fictional gubernatorial candidate. Inserted was a photo of the supposed candidate, a male “morphed” to appear either dark-skinned black, light-skinned black, or white. All aspects of the candidate other than race and skin color were identical. Terkildsen found that both of the black candidates elicited less support than the identical white candidate. In addition, consonant with the hypothesis of implicit racial priming, the effect of negative racial predispositions on candidate support was strongest with the implicit cue, that is, the light skinned candidate. However, among subjects low in self-monitoring—people less concerned about social norms—the effect of negative racial predispositions on candidate evaluation was stronger for the dark- than the light-skinned candidate. That is, in a process much like the self-censorship documented by the psychology studies, self-monitors completely repressed the impact of their negative racial predispositions with exposure to a dark skinned candidate. In other words, light skin serves as an implicit racial cue that avoids conscious detection and circumvents whites' self-censorship. This study thus provides evidence that race cues matter: candidate skin color acts as an effective racial cue and results in racial priming. Furthermore, we also have evidence that the effect of race cues is regulated by sensitivity to norms, which represses racial predispositions. More implicit cues—light skin—can better operate outside awareness, in line with the implicit stereotyping model.34

Citrin, Green, and Sears 1990 conducted a survey study of an actual black candidate and did not find racial priming, but their analyses of pre-election polls are contaminated by omitting the extensive “don't know” and refusal responses to questions about the black candidate, which are likely to hide rejection of that candidate (Berinsky 1999), and by miscoding “don't know” and refusal responses together with the most egalitarian responses on the stereotype scales.

The most direct test of the hypothesis that implicit appeals prime racial predispositions, other than Mendelberg's, is Valentino, Hutchings, and White's study.35

Valentino, Hutchings, and White 2002.

Conducted during the 2000 campaign, the study used a fictitious 30-second ad by George W. Bush promising to reduce Democrats' spending of “your tax dollars” on “wasteful government programs” and to “reform an unfair system that only provides health care for some, while others go without proper treatment.” In the “neutral visuals” condition, the investigators inserted nonracial visuals. A second condition, “race comparison,” adds images of blacks (paired with the negative statements) and whites (paired with positive statements). A third condition, closely resembling Mendelberg's implicitly racial messages, used the same black images but omitted the white images (“undeserving blacks”). Two other conditions are counter-stereotypical, switching images to create “deserving blacks” or “undeserving whites.” The authors test for racial priming by contrasting the impact of several racial predispositions on candidate choice in the no-ad control condition against their impact in the various ad conditions, including statistical controls for education, gender, and nonracial ideology.

The results clearly support the racial priming hypothesis. The implicit, “undeserving blacks” condition primed the racial predisposition of racial resentment in vote choice more powerfully than did the other conditions. It also primed the related racial predispositions of laissez-faire racism and perception of blacks' undue influence.36

Bobo and Smith 1998.

Racial predispositions increase from no significant effect in the control condition to a large effect—a 62-point effect for racial resentment—in the implicit, “undeserving blacks” condition. Also replicating Mendelberg, while the implicit condition primed racial predispositions, it did not prime nonracial predispositions. Moreover, following the implicit stereotyping literature, the accessibility of racial predispositions (measured by reaction time to race-relevant words such as “black”) increased after the implicitly racial cue relative to the control condition, with “race comparison” second and “neutral visuals” third. Finally, also replicating Mendelberg, counter-stereotypical messages neutralized the impact of racial predispositions (though this holds for the “deserving blacks” but not the “undeserving whites” messages). Thus, this study closely replicates Mendelberg in design and result.37

Mendelberg 1997; 2001.

These findings are all the more remarkable considering the disproportionately Democratic sample exposed to pro-Bush ads. The power of elites to promote or deflate racial politics is strong and consistent.

These authors conducted a limited replication with a probability sample of the Detroit area, using face-to-face at-home interviews. They found that the samples in their previous study, which was conducted in a lab in Ann Arbor, Michigan, and the Detroit at-home study, behaved according to a similar pattern.38

Valentino, Traugott, and Hutchings 2002.

The similar findings matter because the Detroit area sample is much closer to national samples on relevant variables.

Nelson and Kinder also found that implicit racial cues prompt racial priming in attitudes about affirmative action.39

Nelson and Kinder 1996.

They exposed subjects to negative (stereotype-consistent) or positive (stereotype-inconsistent) images of blacks or to irrelevant images of whites. Racial resentment had virtually no effect with white images, a moderately sized but statistically insignificant effect after positive images of blacks, and a very large impact with negative images of blacks.

Especially noteworthy is Hurwitz and Peffley's study, which reinforces the finding that implicit cues produce racial priming, this time with a national sample.40

Hurwitz and Peffley 2005.

They conducted an experiment with a nationwide random digit dialing sample of 290 in which they asked non-Hispanic whites to respond to the following question: “Some people want to increase spending for new prisons to lock up violent [inner-city] criminals. Other people would rather spend this money for antipoverty programs to prevent crime. What about you?”41

Hurwitz and Peffley 2005, 102–103.

One random half of the sample got a version of the question that omits the phrase “inner-city”, while the other half got a version with the phrase “inner-city” inserted as indicated above. Hurwitz and Peffley found that racial attitudes (anti-black stereotypes and perceptions of racial fairness, which resemble racial resentment and laissez faire racism) influence punitive preferences on crime policy—but only when violent criminals are identified as “inner-city.” With that implicit racial cue, racial predispositions became more powerful than variables measuring attitudes toward the general fairness of the justice system, equality, fear of crime, political ideology, party identification, education, gender, age, or region. The fact that racial cues caused racial priming with a national sample means that Huber and Lapinski's discrepant findings cannot be explained by their use of a national sample.

Valentino also found racial priming with implicit crime-based race cues.42

Valentino 1999.

He conducted an experimental study of the impact of race cues embedded in actual crime news stories on support for Bill Clinton. The racial cue was a five-second mug shot of two suspects, both of the same race. One group viewed the crime story with no images, while four other groups viewed white, black, Asian, or Hispanic male suspects, and a control group saw no crime story. The three minority conditions were pooled. The news stories with minority suspects reduced Clinton's lead over Dole by 22 percentage points over the no-crime control condition. The white suspects had no effect. In addition, the impact of perceived Clinton performance on the issue of welfare (a race-coded issue) on evaluations of Clinton obtained only after exposure to the minority suspects. Finally, only the race cues caused concern over Clinton's treatment of whites to shape Clinton feelings.

Gilliam and Iyengar conducted the largest of the existing experiments of racial cues, also using implicit cues.43

Gilliam and Iyengar 2000.

Subjects were exposed to an actual, typical television news story about crime featuring either a mug shot of a black or white male suspect, or no image. A control group saw an unrelated story. Nearly half the subjects exposed to the crime story without an image misremembered seeing a black suspect where there was none. This finding makes sense in light of the documented heavy dose of racial crime coverage on local TV news and studies in psychology showing that people fill in missing information from a mental “script.” Unsurprisingly, then, the black image and no image conditions had similar results, in contrast to the white image and the unrelated story. Among whites, exposure to the race cue drove dispositional attributions of and punitive remedies to crime upward by 4 to 6 percentage points. In addition, race cues increased racial resentment. So, a thousands-strong study also replicates the basic finding that implicit messages racialize whites' political preferences.

Peffley, Shields, and Williams employed a design without a no-story control condition, but their findings replicate these studies in other respects.44

Peffley, Shields, and Williams 1996.

They exposed whites to a local news story of a violent crime, inserting an image of a white or black male suspect. Only exposure to the violent black image—the implicit race cue—elicited a strong effect of anti-black stereotypes on judgments of guilt, punishment severity, predicted recidivism, and fear and anger.

Gilliam, Valentino, and Beckmann replicated Gilliam and Iyengar, finding that exposure to implicit cues consisting of black violent criminals in local news stories increased negative racial stereotypes by 6 points, feelings of distance from blacks by 4 points, dispositional causal attributions by 4 points, and punitive crime solutions by 11 points among whites living in homogenous racial areas (as the majority of whites do).45

Gilliam, Valentino, and Beckmann 2002; Gilliam and Iyengar 2000.

Domke extended these studies on the impact of implicit race cues with two new measures.46

Domke 2001.

One experimental group received a news article with racially coded words such as “gangs, inner city, and crack cocaine,” the other group a version without. The study asked for open-ended responses to crime. Thirty-four percent of those cued with racial code words mentioned these and related terms, more than double the percentage in the no-cue condition. Further, stereotypes of blacks and Hispanics were moderately correlated with nonracial political ideology (conservatism or liberalism) after the racial cues only.

Not only do negative messages work this way, but even messages advocating more spending to assist programs that benefit blacks can backfire and cause racial priming if they cue negative racial stereotypes. White conducted two experiments using such messages.47

White 2007.

In one experiment subjects were exposed to a newspaper article about a politician who opposes the war in Iraq because it discriminates against African Americans (explicit), or because it prevents spending on programs to assist the poor (implicit), or for a nonracial reason (nonracial), or were exposed to an off-issue control story. Among whites, only the implicit story increased the effect of anti-black feelings on opposition to the war relative to the off-issue control condition. Among blacks, the response consisted of a stronger effect of pro-black feelings on opposition to the war, and it was strongest in the explicit condition, in line with the expectation that more explicit pro-black messages would better mobilize blacks' in-group solidarity. The second experiment found similar results on the issue of funding for food stamps, with whites racializing their opinion on food stamps only in the implicit “inner city families” condition and not in the explicit “black and Hispanic,” implicit “poor,” or nonracial “working American” conditions. Blacks were again more responsive to the explicit than the other messages, but also showed a negative response to the stigmatized subgroup represented by “inner city.” Thus White replicates the finding that whites respond only to implicit cues to negative racial stereotypes and not to explicit cues, even in the more demanding case of a message that advocates more rather than less assistance to African Americans. He also finds that blacks respond favorably to the more explicit appeal to blacks' needs, because the more explicit reference to blacks evokes their sense of racial solidarity in the face of racial disadvantage and they do not perceive that reaction as a violation of the norm of racial equality. This finding is the mirror image of the finding about whites' rejection of explicit messages. The two mirror reactions are consonant with the argument that racial responses are tailored to racial norms, with whites perceiving that their anti-black reaction to an explicit cue is a violation of the norm and blacks perceiving that their own pro-black reaction is not such a violation. Finally, the study also demonstrates that even blacks are vulnerable to implicit anti-black appeals.

Aggregate Studies

Studies of aggregate opinion do not tell us about the mechanism, but they do find that racial cues in the media shape collective opinion on race. Kellstedt finds that aggregate public opinion on race policy shifts predictably with the negative or positive balance of media coverage of blacks.48

Kellstedt 2003.

In times when the balance of media coverage portrays blacks as having been treated unequally and there is less emphasis on the deficiencies of blacks, public opinion moves in a more liberal direction on race policy; conversely, when the balance tips away from blacks' unequal treatment and toward blacks' shortcomings, public opinion turns in a more conservative direction on race policy. Kellstedt covers the four decades from the middle of the twentieth century on, uses Newsweek coverage coupled with 43 pooled national samples asked 19 different survey questions on a variety of racial policy areas, and accounts for a variety of alternative explanations. It is unclear how implicit the cues are. But the main point is that an aggregate design replicates the general conclusion that the extent to which political communication about race is negative or positive corresponds to negative or positive opinion on race policy.

Gilens provides aggregate evidence on the impact of race cues on the issue of welfare.49

Gilens 1999.

In the mid-1960s and 1970s, as the media increasingly exaggerated the link between poverty and African Americans and increasingly portrayed them as undeserving, public opposition to generous welfare policies grew. Conversely, as the economy worsened in the 1980s, the media provided poverty coverage that featured more whites and portrayed them as more deserving, and whites' opposition to generous welfare benefits decreased significantly. Gilens specifically analyzes the images in media messages, so his results speak to the effects of racial cues. Through this analysis and by ruling out alternative explanations, Gilens shows that race cues in the media appear to racialize the white public's response.50

See also Entman and Rojecki 2000 for content analyses of various media. Winter 2006 also finds that discourse about social security racializes opinion on that issue, associating it with whiteness in a mirror image of the findings on welfare.

Summary of the Literatures

In sum, several large literatures based on various methods and samples lead to two conclusions. First, public opinion can be more or less racialized depending on the messages in the political environment. Second, in all but the aggregate studies (in which the nature of the cue is unclear), these messages appear to work implicitly, that is, they conform to Mendelberg's definition of implicit messages. These studies are sufficiently different in demographics, method, time frame, and the nature of the race cues, that their consistent conclusion is difficult to refute by reference to any one of these dimensions. A study that finds no evidence of racial response to race cues stands out from not one but several literatures showing otherwise.

Null Results and Manipulation Checks

The question then is not the one posed in Huber and Lapinski's paper, namely why Huber and Lapinski failed to replicate one experiment in Mendelberg's chapter 7.51

Mendelberg 2001.

Rather, the question is why Huber and Lapinski did not replicate the accumulated findings of entire literatures.

To start, I briefly summarize Huber and Lapinski's design. Huber and Lapinski assigned subjects in their experiment A to view an implicit or explicit anti-welfare issue ad or an off-issue nonracial control ad. The implicit and explicit ads each feature an image of a black woman on welfare; the explicit condition in addition verbally targets blacks (“too many welfare recipients, especially blacks, take advantage of our tax dollars”). Subjects in experiment B were assigned to one of four anti-welfare messages: either the implicit or explicit ads from experiment A, a counter-stereotypical ad replacing the image of the black welfare woman with a white counterpart, or a fourth ad designed to check for accurate audio reception that showed the white woman but used the verbal content of the explicit anti-black ad. Before the ad, they assessed subjects' predispositions, and after the ad, they asked subjects about the ad assigned to them and about their preferences on various race-related policies. The authors then tested whether the implicit message causes a greater increase in the effect of negative racial predispositions on race policies.

As a first step, I turn to Huber and Lapinski's table 4, the key table in the paper. This table tests one of the basic hypotheses of the race cues literature in general, and the key hypothesis of the racial priming literature in particular: that racial cues strengthen the impact of racial predispositions on relevant opinion. Table 4 is striking because it consists entirely of null results; there is not a single statistically significant effect of racial messages.

When experimental conditions fail to have effects, the next logical step is to conduct a manipulation check. Huber and Lapinski's manipulation check rests on a question about whether “issue advertisements were good for democracy”, worded as: “Do you think that it is good for democracy that groups run these types of ads?”52

Huber and Lapinski 2006, 424, 427.

(This wording does not refer to “issue ads”, but Huber and Lapinski write that the question asked about “issue ads.”) Answers are in five categories coded 0 to 1, with 1 indicating disagreement. The logic of using this question is to assess “both agreement with a message's policy content (stronger welfare work requirements) and the acceptability of its racial content”.53

Huber and Lapinski 2006, 427.

The measure is problematic in several respects as a measure of whether people viewed and heard the message as intended. It is much too broad, asking about ads rather than the message at hand; it asks about the irrelevant category of issue ads and asks for a judgment about the vague concept of “good for democracy” rather than the specific and relevant category of a racial ad; and it is designed to tap evaluations of the issue content of the ad along with perception of the ad's racial/nonracial cues. Nevertheless, Huber and Lapinski conclude that the messages were received as intended because there are statistically significant differences between conditions on this measure.

However, with over 6,300 respondents, even a small difference of means is likely to reach statistical significance, so the question is not significance alone but magnitude. In fact, the key contrast, which is between implicit and explicit conditions, amounts to a difference of only 7 percentage points in experiment A and 6 percentage points in experiment B. Also problematic for the authors' claim that the explicit message was perceived as clearly racial and therefore that the treatment was received as intended is the small difference between the explicit message and the two types of messages that did not have any stereotypical racial cues—12 percentage points in experiment A, and 8 percentage points in experiment B.

Two plausible possibilities suggest themselves to explain the apparent failure of Huber and Lapinski's experimental manipulation: 1) that many subjects did not receive their assigned message (and the faulty manipulation check leaves open the possibility that most subjects did not get their message), and 2) that racial predispositions were measured just before the messages were sent and therefore likely washed out the effects of the message. I take up each explanation in turn below.

Explaining Null Results I: WebTV and Exposure

First, unlike the many successful race cue studies, the Huber and Lapinski study used a WebTV survey administered via the Internet to a sample of Knowledge Networks participants in their homes. The study to which Huber and Lapinski refer readers for details of their experimental procedure is Clinton and Lapinski,54

Clinton and Lapinski 2004.

which examines the impact of (nonracial) negative campaign ads.55

Huber and Lapinski 2006, 423, n. 5. This is one of only two published studies of political ads using WebTV, and the study using by far the largest number of respondents. The other study is Gale et al. 2005. A recent KN study by Brooks and Geer of negative campaign appeals used brief written statements rather than ads that must be downloaded, and this difference may have been important; in any case, it does not seem to have suffered the problem of indistinguishable treatments, with a manipulation check yielding percentages of 64 percent, 44 percent and 23 percent for the three basic conditions (Brooks and Geer 2007, 6).

That paper reports that the WebTV format resulted in at least 62 percent of subjects in the initial wave failing to receive any ad because of technical difficulties (4,614 of 7,464). Because the people affected were “selected” out of viewing on roughly random criteria, Clinton and Lapinski were able to treat these unexposed subjects as a control group that did not view an ad (and they included appropriate statistical control variables).56

Clinton and Lapinski 2004, 76, n. 8.

But as it was, they found mostly null results. Since Huber and Lapinski cite this study for their procedures, we can assume that roughly the same proportion of subjects who were unable to receive ads in Clinton and Lapinski's study were similarly unable in Huber and Lapinski's study.57

Knowledge Networks provided information only about general current procedures and without sufficient detail to evaluate these procedures or to know which procedures were followed by Huber and Lapinski (Mike Dennis, e-mail March 3, March 13, 2006). The “download rate” for the most recent studies exposing viewers to ads is 75 percent–78 percent but no academic studies are included in this set so this rate does not apply to Huber and Lapinski.

That almost two-thirds of subjects supposed to get an ad probably did not get it, and yet are apparently miscoded as receiving it, has important implications for the Huber and Lapinski study. It means that the key independent variable has more measurement error than true values. Measurement error in any explanatory variable biases the estimated effects of all the variables in the model.58

Achen 1983.

Bartels analyzed random measurement error in self-reported measures of media exposure. He found that about 25 percent of the observed variance was random noise. In addition, he found that about 40 percent of the variance in the opinion variables usually included as predictors of vote choice consisted of error. Consequently, the uncorrected coefficients “generally understate the effects of … exposure, in some cases by as much as 50 [percent].”59

Bartels 1993, 270, 271.

When accounting for message distinctiveness from exposure, failing to adjust for measurement error produced “results that fall short by … 67 [percent] on average.”60

Bartels 1993, 275.

Huber and Lapinski's measurement error is much greater still, so we would expect their estimates to be depressed by even more than the examples in Bartels. Thus, Huber and Lapinski's large non-exposure rates go a long way to explain their null results.61

Of course, it is more realistic for study participants to view ads in their own homes. But if the goal is to make strong causal inferences and a controlled experiment is run toward that end which assumes full exposure by every subject, then we need to ascertain that this level of control was in fact exercised.

Skeptics might raise the counter-argument that even if many subjects are not exposed to their assigned message, there is no threat to internal validity, because subjects watch or do not watch to the same extent across conditions. However, while it is true that there is no threat to internal validity here, that is because there is no effect whose validity is threatened. Equally low exposure across conditions would explain the weak effects we observe across conditions.

Skeptics may raise a second counter-argument: that Huber and Lapinski included a condition designed to check that the ads were in fact received. In Experiment B some subjects were assigned to a condition that used the explicit condition's explicit racial audio but included a visual of a white welfare recipient. Huber and Lapinski included this explicit-audio condition to ensure that all subjects heard the audio in whatever message they were assigned. They report finding no difference in the “issue ads are bad for democracy” responses in this explicit-audio condition and the explicit condition. From this similarity in response they conclude that subjects were in fact exposed to their assigned ad. However, as noted, replies to that question do not tell us if subjects were exposed to any ad, since the question does not ask about an ad. Furthermore, the responses to this question do not vary much across the various conditions, as noted above.62

Finally, I consider the possibility that the particular message used by Huber and Lapinski was ineffectively constructed. This is unlikely because in Gilliam's study exposure to the same images used by Huber and Lapinski for the black versus white welfare “queen” produced a 5-point increase in opposition to welfare spending and a 10-point increase in negative views of welfare recipients, so the images could not have been a total wash (Gilliam 1999). I do not include this study in my discussion of the public opinion literature because its full details have not yet been published.

Explaining Null Results II: Unintended Priming Prior to Exposure

A final explanation for the null results rests with question order. More precisely, Huber and Lapinski measured racial predispositions just before exposure to the ad, which likely neutralized the effect of the ad's racial cue.63

Huber and Lapinski 2006, 424, n. 7.

They recognize the problem with this strategy: “We accept the risk of priming racial considerations with the pre-test.”64

Huber and Lapinski 2006, 424, n. 7.

They justify it by trading off the risk of racial priming against the risk that the racial ad will alter responses to the questions about racial predispositions. But the price paid for this design decision is that the racial questions closely preceding the ad likely primed racial dispositions for all subjects thereby muting the effect of subsequent exposure to the racial cues.

A well-established literature documents the pervasive influence of question order in surveys.65

Bradburn 1982; Schuman and Presser 1981; Tourangeau and Rasinski 1988.

For example, asking questions about women's rights—a predisposition relevant to abortion policy—alters answers to policy questions about abortion.66

Tourangeau and Rasinski 1988.

The prevailing theory of the survey response is that people answer survey questions by sampling from accessible considerations. Relevant prior questions shape answers to subsequent questions by making those considerations more accessible. In essence, the prior question itself acts as a cue that primes predispositions.

These findings on question order have transformed our understanding of substance (how the public decides) and method (how to prevent the huge biases introduced by seemingly small changes in question order).67

Bartels 2003; Berinsky 2004; Zaller 1992.

The lesson here is that the close pre-message measurement of racial predispositions increases the accessibility of these predispositions, blurring the distinctions between the subsequent message conditions and undermining the ability of the race cues to elicit distinctive racial effects. If we want to know whether a message primes predispositions, we must measure the predispositions well in advance of the message.68

See Mendelberg 2001.

Why the Reasons Offered by Huber and Lapinski for the Discrepancy Fail to Explain It

We are now in a better position to evaluate the main reasons provided by Huber and Lapinski for the discrepancy between their findings and Mendelberg's. Their first reason is that Mendelberg's study in chapter 7 was regionally specific while their study is national, and therefore more valid. However, Mendelberg replicated her Michigan study in New Jersey, and further corroborated it with the 1988 National Election study (NES), a national sample.69

Mendelberg 2001.

More important, the basic finding of the racial effects of implicit race cues has been replicated in a variety of samples both local and national, reviewed above.

The second reason they provide for the discrepancy is that the Michigan study used only a few hundred cases while theirs used several thousand. However, the basic finding—that cues can racialize public preferences—rests on studies totaling over 5,000 cases, adding across the public opinion studies detailed above. If Mendelberg's results are due to sampling vagaries, then we would expect replications to fail, yet as reviewed above, they do not.

Third, the article claims that Mendelberg had no control group.70

Huber and Lapinski 2006, 424.

In fact, in two of three separate experiments on racial priming conducted by Mendelberg, she found racial priming relative to a nonracial control group.71

One experiment is reported in Mendelberg 2001, ch. 8, and the other in a 1997 article also described on p. 184 of the book. Mendelberg writes of the New Jersey experiment: “the impact of resentment on candidate choice is negligible when the candidate does not engage in a racial appeal. Resentment makes a good deal of difference to candidate choice, but only when candidates communicate racial messages” (2001, 219–220).

In addition, the Michigan study cited in Huber and Lapinski did in fact include a control condition, but as reported there, the random assignment to this condition failed and this group was significantly different from the others on a variety of demographic characteristics and thus could not be included.72

Huber and Lapinski 2006, 197, n. 8.

The main point is that a large accumulation of studies have shown racial priming relative to a nonracial control.

Huber and Lapinski also claim that too few resentful individuals were included in the Mendelberg study and therefore the results are due to chance. But while the Michigan study did have a low resentment mean, the New Jersey study's mean approaches the mean of national samples such as NES or Huber and Lapinski's (0.50 versus 0.59 in Huber and Lapinski). In addition, Mendelberg also established the racial priming effect with the nationally representative 1988 NES, with a resentment mean of .60.73

Mendelberg 2001, 188.

Finally, an abundance of other studies found racial priming, with various levels of racial predispositions.

Another explanation provided by Huber and Lapinski for their discrepant finding is that because their study was a representative national sample, it had a good distribution of education, while Mendelberg's study had a very high education level. They argue that people with more education are especially immune to racial priming. But this argument does not explain the discrepancy. First, the poorly-educated subsample in the Huber and Lapinski study does not behave differently at conventional levels of statistical significance. Second, highly educated samples in other studies have yielded strong racial priming effects.74

Mendelberg 2001 and Valentino, Hutchings, and White 2002; also, many implicit stereotyping studies are conducted on well educated samples.

Third, Federico finds that a race cue prompts stronger racial associations among the better educated—it works more powerfully for them, not less.75

Federico 2004. Another possibility for the discrepancy, not suggested by Huber and Lapinski, is that the implicit message used in Mendelberg is too overt for the “post-Willie Horton” period. However, Valentino, Hutchings, and White 2002 found strong racial effects for messages very similar to Mendelberg's during the 2000 campaign, over a decade after the media discussion of Horton, so this, too, fails as an explanation.

Most importantly, the particulars of Mendelberg's study are now virtually irrelevant given the large number of studies that replicate her basic results. Even if Mendelberg's study was set aside entirely and disregarded, its conclusion remains.

Conclusion

Over a decade ago, Larry Bartels noted:

The state of research on media effects is one of the most notable embarrassments of modern social science. The pervasiveness of the mass media and their virtual monopoly over the presentation of many kinds of information must suggest to reasonable observers that what these media say and how they say it has enormous social and political consequences. Nevertheless, the scholarly literature has been much better at refuting, qualifying, and circumscribing the thesis of media impact than at supporting it.76

Bartels 1993, 267.

Bartels faulted, in part, the “limitations of research design.”77

Bartels 1993, 267.

In this paper I find several seemingly trivial but in fact crucial design choices that have produced the wrong conclusion from the largest study to date about the effect of media on racial politics, namely that the media have no effect. The fact that this conclusion is false matters on substance, but also on method: even the small particulars of experimental design can affect our ability to draw accurate causal inferences about public opinion.

The literature review here reveals many studies replicating the finding that implicit race cues racialize public opinion. But beyond the claim that race cues work effectively in an implicit fashion, the issue at hand revolves most centrally around the claim that no race cue of any kind matters. These studies contradict the Huber and Lapinski findings of null results by establishing the potency of racial cues in political judgments. In that sense, the question of implicit appeals becomes secondary; the main issue is Huber and Lapinski's failure to find any statistically significant racialized response.

Huber and Lapinski do raise the useful question of whether there are subgroups that might react favorably to explicit racial messages. It is possible that Mendelberg's conclusion that explicit messages in the post-civil rights era do not work very well is too unqualified. The next step is to discover the circumstances and reasons that make this true.

For example, in the South racial attitudes appear to lean more rightward than in the rest of the country, so explicit racial cues may have some effect there.78

Glaser and Gilens 1997; Kuklinski, Cobb, and Gilens 1997; Kuklinski et al. 1997; Valentino and Sears 2005.

This prediction is supported by Hutchings, Walton, and Benjamin's finding that while Southern white women are fairly resistant to explicit racial messages, Southern white men are receptive.79

Hutchings, Walton, and Benjamin 2005.

This result also suggests that gender may be an important moderator, perhaps especially in the South.80

See also Hutchings et al. 2004; Kuklinski, Cobb and Gilens 1997.

Huber and Lapinski specifically raise the possibility that messages are moderated by education. Their analysis is inconclusive since the messages had no statistically significant effects for either low or high education group.81

Huber and Lapinski 2006, figure 5. See also p. 436. In other analyses of the effect of education they interact education with racial predispositions without including the main effect of education (tables 5 and 6). The results in figure 5 may well derive from a model that omitted this term as well (the model's specification is not reported).

Nevertheless, education deserves more attention, although Federico's findings, mentioned earlier, suggest that the direction runs opposite to that suggested by Huber and Lapinski.82

Federico 2004.

Relatedly, explicit messages may work when certain frames are used. Hutchings and Valentino suggest that messages about policy conflicts between racial groups can be explicit and still work, while messages that convey racial stereotypes are more likely to work if implicit.83

Hutchings and Valentino 2003.

This makes sense if the rejection of explicit messages results from people thinking of them as too racist. People are likely to reject the message if it conveys statements easily recognized as racist, such as categorical derogation of blacks as a group. They are less likely to reject the message if its explicit statements appear to be legitimate policy disagreements with black leaders or activists.84

Relatedly, explicit messages may work when the first black candidate is poised to win an influential office and is perceived as a challenge to whites' interests. Reeves' 1997 experimental study identified the candidates verbally as “black” or “white,” and the black candidate as “seeking to become the city's first black mayor.” Reeves codes this as a subtle appeal, but it appears explicit by Mendelberg's definition. Negative racial predispositions kick in against the black candidate, but only when the disagreement is over affirmative action, not when it is about the environment.

A second question for future research, suggested by the literature on race cues reviewed above, is how the distinction between implicit and explicit works. In past work, I have treated these as a dichotomy, but actually it seems more accurate to think of them as end points on a continuum.85

Mendelberg 2001.

One end of the continuum may be anchored by the subconscious cues in implicit stereotyping studies, where people are unaware that they saw a racial cue. The other may be anchored by categorical negative statements about the group as a whole. In addition, some people, such as the subjects of negative appeals, may have a low threshold of awareness for racial cues while others have a high one. If so, then the questions become 1) why and how people differ in their awareness of and sensitivity to race cues; 2) what characteristics of cues make them increasingly salient and noticeable as racial cues; and 3) how societal or cultural forces and institutions, such as the black church or the civil rights movement, shape perceptions of cues as more implicit or more explicit by socializing or mobilizing their members to be vigilant to anti-ingroup appeals.

A final and important question emerging from the literature I reviewed is how the subjects of these messages perceive and react to them. Gilliam and Iyengar find that race cues in news coverage of crime do not affect African Americans as they do white Americans.86

Gilliam and Iyengar 2000.

This result is replicated and elaborated by White, to date the only study to focus on the comparison of white and black responses.87

White 2007.

On the other hand, Valentino, Hutchings, and White offer suggestive evidence to the contrary (but with too few cases to allow subgroup analyses), and White suggests that blacks are susceptible to some extent to negative messages about black subgroups about whom many blacks may feel ambivalent.88

Valentino, Hutchings, and White 2002.

The social psychology literature on racial priming provides a complex answer to this question. In previous work I argued that the strategy of challenging explicit racial messages was an important legacy of the civil rights movement, which suggests that African Americans will be more vigilant about racial content and reject anti-black implicit messages at higher rates than nonblacks.89

Mendelberg 2001.

However, blacks also tend to be ambivalent about which black subgroup interests deserve to be included in the “black” political agenda and may seek to distance themselves from marginalized black subgroups such as drug users, which may make them susceptible to negative cues about black subgroups.90

Cohen 1999.

More research is needed on how disadvantaged groups, including African Americans, Latinos, Asian Americans, Native Americans, women, and immigrants—all subjects of negative messages in society and politics at some times and places—are affected when receiving these messages. We also need more research on when and how these groups respond collectively and politically to these messages.91

Dawson and Wilson 1991.

Finally, we need to better understand reactions to messages advocating more rather than less government action on behalf of African Americans.

In sum, the weight of the evidence from well over a dozen experiments and two aggregate studies, along with hundreds of implicit stereotyping studies, supports the hypothesis that race cues often racialize white public opinion. Racial messages do shape the political response of white citizens. Conversely, racialized preferences are not immutable; when the cues change, so does public opinion. In addition, much of this evidence also shows that race cues frequently work implicitly. Methodological problems explain why a massive randomized study contradicts this conclusion. The null findings for racial cues are artifacts of flawed methodology. The conclusion remains that the media and politicians strengthen or dampen the impact of race on politics in ways consequential for American democracy. Future research should take that as a given and proceed to investigate its contingencies and complexities.

References

Achen, Christopher H. 1983. Toward theories of data: The state of political methodology. In Political Science: The State of the Discipline, ed. Ada W. Finifter. Washington, DC: American Political Science Association.
Banaji, Mahzarin R., and Anthony G. Greenwald. 1994. Implicit stereotyping and prejudice. In The Psychology of Prejudice: The Ontario Symposium, ed. Mark P. Zanna, James M. Olson, and vol. 7. Hillsdale, NJ: Erlbaum.
Bargh, John A. 1996. Principles of automaticity. In Social Psychology: Handbook of Basic Principles, ed. E. Tory Higgins and Arie W. Kruglanski. New York: Guilford.
Bargh, John A., and Mark Chen. 1997. Nonconscious behavioral confirmation processes: The self-fulfilling consequences of automatic stereotype activation. Journal of Experimental Social Psychology 33 (5): 541–60.Google Scholar
Bargh, John A., Mark Chen, and Lara Burrows. 1996. Automaticity of social behavior: Direct effects of trait construct and stereotype priming on action. Journal of Personality and Social Psychology 71 (2): 230–44.Google Scholar
Bartels, Larry M. 1993. Messages received: The political impact of media exposure. American Political Science Review 87 (2): 267–84.Google Scholar
Bartels, Larry M. 2003. Democracy with attitudes. In Electoral Democracy, ed. Michael MacKuen and George Rabinowitz. Ann Arbor: University of Michigan Press.
Berinsky, Adam J. 1999. The two faces of public opinion. American Journal of Political Science 43 (4): 1209–30.Google Scholar
Berinsky, Adam J. 2002. Political context and the survey response: The dynamics of racial policy opinion. Journal of Politics 64 (2): 567–84.Google Scholar
Berinsky, Adam J. 2004. Silent Voices. Princeton: Princeton University Press.
Berinsky, Adam J., and Tali Mendelberg. 2005. The indirect effects of discredited stereotypes in judgments of Jewish leaders. American Journal of Political Science 49 (4): 845–64.Google Scholar
Blair, Irene V., Charles M. Judd, and Kristine M. Chapleau. 2004. The influence of Afrocentric facial features in criminal sentencing. Psychological Science 15 (1): 674–79.Google Scholar
Bobo, Lawrence D., and Ryan A. Smith. 1998. From Jim Crow racism to laissez-faire racism: The transformation of racial attitudes. In Beyond Pluralism: The Conception of Groups and Group Identities in America, ed. Wendy F. Katkin, Ned C. Landsman, and Andrea Tyree. Chicago: University of Illinois Press.
Bradburn, Norman. 1982. Question-wording effects in surveys. In Question Framing and Response Consistency, ed. R. Hogarth. San Francisco: Jossey-Bass.
Brooks, Deborah J., and John G. Geer. 2007. Beyond negativity: The effects of incivility on the electorate. American Journal of Political Science 51 (1): 1–16.Google Scholar
Chandra, Kanchan. 2004. Why Ethnic Parties Succeed: Patronage and Ethnic Headcounts in India. Cambridge: Cambridge University Press.
Citrin, Jack, Donald Green, and David O. Sears. 1990. White reactions to black candidates: When does race matter? Public Opinion Quarterly 54 (1): 74–96.Google Scholar
Clawson, Rosalee A., and Rakuya Trice. 2000. Poverty as we know it: Media portrayals of the poor. Public Opinion Quarterly 64 (1): 53–64.Google Scholar
Clinton, Joshua D., and John S. Lapinski. 2004. “Targeted” advertising and voter turnout: An experimental study of the 2000 presidential election. Journal of Politics 66 (1): 69–96.Google Scholar
Cohen, Cathy. 1999. The Boundaries of Blackness. Chicago: University of Chicago.
Correll, Joshua, Bernadette Park, Charles M. Judd, and Bernd Wittenbrink. 2002. The police officer's dilemma: Using ethnicity to disambiguate potentially threatening individuals. Journal of Personality and Social Psychology 83 (6): 1314–29.Google Scholar
Dawson, Michael, and Ernest Wilson III. 1991. Paradigms and paradoxes: Political science and the study of African American politics. In Political Science: Looking to the Future, ed. William Crotty and vol. 1. Evanston, IL: Northwestern University Press.
Devine, Patricia G. 1989. Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology 56 (1): 680–90.Google Scholar
Devine, Patricia G., and Andrew J. Elliot. 1995. Are racial stereotypes really fading? The Princeton trilogy revisited. Personality and Social Psychology Bulletin 21 (11): 1139–50.Google Scholar
Devine, Patricia G., Margo J. Monteith, Julia R. Zuwerink, and Andrew J. Elliot. 1991. Prejudice with and without compunction. Journal of Personality and Social Psychology 60 (6): 817–30.Google Scholar
Dickson, Eric, and Kenneth Scheve. 2006. Social identity, political speech and electoral competition. Journal of Theoretical Politics 18 (1): 5–39.Google Scholar
Domke, David. 2001. Racial cues and political ideology: an examination of associative priming. Communication Research 28 (6): 772–801.Google Scholar
Dovidio, John F., and Samuel Gaertner. 1986. Prejudice, discrimination, and racism. In Prejudice, Discrimination, and Racism, ed. John F. Dovidio and Samuel Gaertner. New York: Academic Press.
Dovidio, John F., Kerry Kawakami, Craig Johnson, Brenda Johnson, and Adaiah Howard. 1997. On the nature of prejudice: Automatic and controlled processes. Journal of Experimental Social Psychology 33 (5): 510–40.Google Scholar
Dovidio, John F., Nancy Evans, and Richard B. Tyler. 1986. Racial stereotypes: The contents of their cognitive representations. Journal of Experimental Social Psychology 22 (1): 22–37.Google Scholar
Eberhardt, Jennifer L., Phillip Atiba Goff, Valerie J. Purdie, and Paul G. Davies. 2004. Seeing black: Race, crime, and visual processing. Journal of Personality and Social Psychology 87 (6): 876–93.Google Scholar
Eberhardt, Jennifer L., Paul G. Davies, Valerie J. Purdie-Vaughns, and Sheri Lynn Johnson. 2006. Looking deathworthy: Perceived stereotypicality of black defendants predicts capital-sentencing outcomes. Psychological Science 17 (5): 383–86.Google Scholar
Edsall, Thomas B., and Mary D. Edsall. 1991. “When the Official Subject Is Presidential Politics, Taxes, Welfare, Crime, Rights, or Values … the Real Subject Is Race.” Atlantic Monthly (May): 53–86.
Entman, Robert M., and Andrew Rojecki. 2000. The Black Image in the White Mind: Media and Race in America. Chicago: University of Chicago Press.
Fazio, Russell H., Joni R. Jackson, Bridget C. Dunton, and Carol J. Williams. 1995. Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology 69 (6): 1013–27.Google Scholar
Federico, Christopher M. 2004. When do welfare attitudes become racialized? The paradoxical effects of education. American Journal of Political Science 48 (2): 374–91.Google Scholar
Fiske, Alan P., Shinobu Kitayama, Hazel R. Markus, and Richard E. Nisbett. 1999. The cultural matrix of social psychology. In The Handbook of Social Psychology, ed. Daniel T. Gilbert, Susan T. Fiske, Gardner Lindzey, and vol. 2. Boston: McGraw-Hill.
Fraga, Luis R., and David L. Leal. 2004. Playing the “Latino card”: Race, ethnicity, and national party politics in 2000 and 2004. Du Bois Review: Social Science Research on Race 1: 297–317Google Scholar
Gale, Kristina, Betsey Gimbel Hawkins, Richard Hawkins, David B. Magleby, J. Quin Monson, and Kelly D. Patterson. 2005. Elections: effects of the stand by your ad provision on attitudes about candidates and campaigns. Presidential Studies Quarterly 35 (4): 771–83.Google Scholar
Gibson, James L., and Amanda Gouws. 2002 Overcoming Intolerance in South Africa: Experiments in Democratic Persuasion. New York: Cambridge University Press.
Gilens, Martin. 1999. Why Americans Hate Welfare. Chicago: University of Chicago Press.
Gilliam, Frank D. 1999. The “welfare queen” experiment: How viewers react to images of African-American mothers on welfare. Nieman Reports 53 (2): 112–19.Google Scholar
Gilliam, Frank D., and Shanto Iyengar. 2000. Prime suspects: The influence of local television news on the viewing public. American Journal of Political Science 44 (3): 560–73.Google Scholar
Gilliam, Frank D., Nicholas A. Valentino, and Matthew N. Beckmann. 2002. Where you live and what you watch: The impact of racial proximity and local television news on attitudes about race and crime. Political Research Quarterly 55 (4): 755–80.Google Scholar
Glaser, James M. 2002. White voters, black schools: Structuring racial choices with a checklist ballot. American Journal of Political Science 46 (1): 35–46.Google Scholar
Glaser, James M., and Martin Gilens 1997. Interregional migration and political resocialization: A study of racial attitudes under pressure. Public Opinion Quarterly 61 (1): 72–86.Google Scholar
Graham, Sandra, and Brian S. Lowery. 2004. Priming unconscious racial stereotypes about adolescent offenders. Law and Human Behavior 28 (5): 483–504.Google Scholar
Greenwald, Anthony G., and Linda Krieger. 2006. Implicit bias: Scientific foundations. California Law Review Symposium on Behavioral Realism and Implicit Social Cognition. Berkeley: California Law Review, Inc.
Huber, Gregory A., and John Lapinski. 2006. The “race card” revisited: Assessing racial priming in policy contests. American Journal of Political Science 48 (2): 375–401.Google Scholar
Huddy, Leonie. 1994. The political significance of voters' gender stereotypes. Research in Micropolitics 4: 169–93.Google Scholar
Hurwitz, Jon, and Mark Peffley. 1997a. Racial stereotypes and whites' political views of blacks in the context of welfare and crime. American Journal of Political Science 41 (1): 30–60.Google Scholar
Hurwitz, Jon, and Mark Peffley. 2005. Playing the race card in the post-Willie Horton era. Public Opinion Quarterly 69 (1): 99–112.Google Scholar
Hutchings, Vincent L., and Nicholas A. Valentino. 2003. “Wedge Politics: The Structure and Function of Racial Group Cues in Contemporary American Politics.” Presented at the Eagleton School, Rutgers University.
Hutchings, Vincent L., and Nicholas A. Valentino. 2004. The centrality of race in American politics. Annual Review of Political Science 7: 383–408.Google Scholar
Hutchings, Vincent L., Hanes Walton, Jr., and Andrea Benjamin. 2005. “Heritage or Hate? Race, Gender, Partisanship and the Georgia State Flag Controversy.” Presented at the 101st Annual Meeting of the American Political Science Association. Washington, DC, September 1–4.
Hutchings, Vincent L., Nicholas A. Valentino, Tasha S. Philpot, and Ismail K. White. 2004. The compassion strategy: Race and the gender gap in campaign 2000. Public Opinion Quarterly 68 (4): 512–41.Google Scholar
Jamieson, Kathleen H. 1992. Dirty Politics: Deception, Distraction, and Democracy. New York: Oxford University Press.
Johnson, James D., Mike S. Adams, William Hall, and Leslie Ashburn. 1997. Race, media, and violence: Differential racial effects of exposure to violent news stories. Basic and Applied Social Psychology 19 (1): 81–90.Google Scholar
Kellstedt, Paul M. 2003. The Mass Media and the Dynamics of American Racial Attitudes. Cambridge: Cambridge University Press.
Kinder, Donald, and Lynn Sanders. 1996. Divided by Color. Chicago: University of Chicago Press.
Krueger, Joachim. 1996. Personal beliefs and cultural stereotypes about racial characteristics. Journal of Personality and Social Psychology 71 (3): 536–48.Google Scholar
Kuklinski, James H., Paul M. Sniderman, Kathleen Knight, Thomas Piazza, Phillip E. Tetlock, Gordon R. Lawrence, and Barbara Mellers. 1997. Racial prejudice and attitudes toward affirmative action. American Journal of Political Science 41 (2): 402–19.Google Scholar
Kuklinski, James H., Michael D. Cobb, and Martin Gilens. 1997. Racial attitudes and the “New South.” Journal of Politics 59 (2): 323–49.Google Scholar
Macrae, C. Neil, Alan B. Milne, and Galen V. Bodenhausen. 1994. Out of mind but back in sight: Stereotypes on the rebound. Journal of Personality and Social Psychology 67 (5): 808–917.Google Scholar
Mendelberg, Tali. 1997. Executing Hortons: Racial Crime in the 1988 Presidential Campaign. Public Opinion Quarterly 61(1): 134–57.Google Scholar
Mendelberg, Tali. 2001. The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality. Princeton, NJ: Princeton University Press.
Nelson, Thomas E., and Donald R. Kinder. 1996. Issue frames and group-centrism in American public opinion. Journal of Politics 58 (4): 1055–78.Google Scholar
Olson, Michael A., and Russell H. Fazio. 2004. Trait inferences as a function of automatically activated racial attitudes and motivation to control prejudice. Basic and Applied Social Psychology 26 (1): 1–11.Google Scholar
Payne, B. Keith. 2001. Prejudice and perception: The role of automatic and controlled processes in misperceiving a weapon. Journal of Personality and Social Psychology 81 (2): 181–92.Google Scholar
Payne, B. Keith, Alan J. Lambert, and Larry L. Jacoby. 2002. Best laid plans: Effects of goals on accessibility bias and cognitive control in race-based misperceptions of weapons. Journal of Experimental Social Psychology 38 (4): 384–96.Google Scholar
Peffley, Mark, and Jon Hurwitz. 1993. International events and foreign policy beliefs: Public response to changing Soviet-US relations. American Journal of Political Science 36: 431–61.Google Scholar
Peffley, Mark, Todd Shields, and Bruce Williams. 1996. The intersection of race and crime in television news stories: An experimental study. Political Communication 13 (3): 309–28.Google Scholar
Plant, E. Ashby, and B. Michelle Peruche. 2005. The consequences of race for police officers' responses to criminal suspects. Psychological Science 16 (3): 180–83.Google Scholar
Poehlman, T. Andrew, Eric Luis Uhlmann, Anthony G. Greenwald, and Mahzarin R. Banaji. 2005. “Understanding and Using the Implicit Association Test: Meta-Analysis of Predictive Validity.” Unpublished manuscript.
Reeves, Keith. 1997. Voting Hopes or Fears?: White Voters, Black Candidates & Racial Politics in America. New York: Oxford University Press.
Schuman, Howard, Charlotte Steeh, Lawrence D. Bobo, and Maria Krysan. 1998. Racial Attitudes in America: Trends and Interpretations, Revised Edition. Cambridge, MA: Harvard University Press.
Schuman, Jean, and Stanley Presser. 1981. Questions and Answers in Attitude Surveys. New York: Academic Press.
Sears, David O., James Sidanius, and Lawrence D. Bobo, eds. 2000. Racialized Politics: Values, Ideology, and Prejudice in American Public Opinion. Chicago, IL: University of Chicago Press.
Sniderman, Paul M., and Thomas Piazza. 1993. The Scar of Race. Cambridge: Harvard University Press.
Terkildsen, Nayda. 1993. When white voters evaluate black candidates: The processing implications of candidate skin color, prejudice and self-monitoring. American Journal of Political Science 37: 1032–53.Google Scholar
Thernstrom, Abigail. 1987. Whose Vote Counts? Affirmative Action and Minority Voting Rights. Cambridge: Harvard University Press.
Tourangeau, Roger, and Kenneth A. Rasinski. 1988. Cognitive processes underlying context effects in attitude measurement. Psychological Bulletin 103 (3): 299–314.Google Scholar
Valentino, Nicholas A. 1999. Crime news and the priming of racial attitudes during evaluations of the president. Public Opinion Quarterly 63 (3): 293–320.Google Scholar
Valentino, Nicholas A., and David O. Sears. 2005. Old times there are not forgotten: Race and partisan realignment in the contemporary south. American Journal of Political Science 49 (3): 672–88.Google Scholar
Valentino, Nicholas A., Michael Traugott, and Vincent Hutchings. 2002. Group cues and ideological constraint: A replication of political advertising effects studies in the lab and in the field. Political Communication 19 (1): 29–48.Google Scholar
Valentino, Nicholas A., Vincent L. Hutchings, and Ismail K. White. 2002. Cues that matter: How political ads prime racial attitudes during campaigns. American Political Science Review 96 (1): 75–90.Google Scholar
White, Ismail. 2007. When race matters and when it doesn't: Racial group differences in response to racial cues. American Political Science Review 101 (2): 1–16.Google Scholar
Winter, Nicholas J.G. 2006. Beyond welfare: Framing and the racialization of white opinion on social security. American Journal of Political Science 50 (2): 400–20.Google Scholar
Wittenbrink, Berndt, Charles M. Judd, and Bernadette Park. 1997. Evidence for racial prejudice at the implicit level and its relationship with questionnaire measures. Journal of Personality and Social Psychology 72 (2): 262–74.Google Scholar
Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.
Figure 0

Public opinion studies finding racial effects of race cues