thesis statement for stereotypes and prejudice

113 Stereotype Essay Topics & Examples

Looking for good stereotypes to write about? Look no further! This list contains only the best themes about stereotypes in society for your college essay or project. Whether you need research questions about stereotypes, essay writing tips, or free samples, you will find them here.

❓ How to Write a Stereotype Essay: Do’s and Don’ts

🏆 best stereotype topic ideas & essay examples, 👍 good essay themes about stereotypes, 📌 most interesting stereotype topics to write about, 👍 good research topics about stereotype, ❓ research questions about stereotypes.

All people are different, which makes living without some ingrained assumptions difficult. From discrimination to mere harmless expectations, stereotyping plays a prevalent part in people’s interactions, often imposing particular behavior on them.

Thus, writing a stereotype essay is only as simple as recognizing both the every-day and the society-wide patterns of thinking, finding the connections between them, and writing them down.

• Think of a specific topic before you begin writing or outlining your paper. Do so by penning a thesis statement, which will not only help you stick to your central theme but also remove any irrelevant ideas. Since there are multitudes of stereotype essay topics, this action will help you focus your thoughts on a single issue.
• Brainstorm your problem beforehand by drafting an outline. Whether you are writing a stereotype threat essay or creating a comprehensive list of anti-female education beliefs, you should create a smooth narrative that flows with ease from one point to another. Furthermore, an outline saves you time, which you would have spent on rewriting those parts of your stereotype essay that are lacking in information or structure.
• Read sample essays. An outstanding stereotypes essay example can act as an excellent incentive to begin writing by demonstrating writing tactics and ways of presenting information to the audience. You may even uplift some of those techniques to your own work to increase the quality of your paper.
• Give your essay an eye-catching title. Stereotype essay titles should not only give the audience a glimpse of what the central theme is but also invite them to read further. The more hooks you have at the beginning of your paper, the higher the possibility of a reader going beyond the first paragraph.
• Generate a comprehensive bibliography. With the number of studies on this topic, there exists a vast amount of book and journal titles that can help you find plenty of interesting themes about stereotypes.
• Pick a broad problem. An essay has a specified word count, and your instructor will not reward writing over the set limit. Choose an issue that you are sure you can adequately cover in the specified pages, and remember to adhere to your received instructions. There is nothing worse than writing an excellent essay and losing marks for not following directions.
• Plagiarize from others’ essay examples. Copying and pasting sentences is an academic offense, as is merely rewording them, and you should avoid discrediting your hard work. Getting your paper disqualified is not worth a small increase in marks.
• Attempt to subvert every stereotype you come across. While deconstructing some issues is a noble endeavor, this work may be extensive and exhausting, as well as not the main point of your paper. Remember your thesis statement, and work in those facts that relate to it.
• Make light of your chosen problem. Just as with your title, your writing should remain respectful and academic, using only credible information and referencing trusted sources. Remember that, as with any humanities issue, stereotypes are a societal byproduct that affects living people, who deserve fair treatment.
• Skip the pre-writing stages. Doing so may lead you to write an essay, which is not only off-point but also overwhelmingly one-sided. Your paper should give adequate attention to different sides of one issue, presenting different viewpoints, studies, and academic opinions, which brainstorming helps achieve.

Need more tips? Let IvyPanda guide your writing process!

• Canadian Stereotypes On the cover of the novel Canadian stereotypes, there will be the image of the maple leaf bag. The image of the maple leaf bag will represent both the flag and the history of the […]
• Stereotypes and Their Effects Three common stereotypes include the perception that Muslims are terrorists, Christians are ignorant, and that women are less intelligent than men.
• Gender stereotypes of superheroes The analysis is based on the number of male versus female characters, the physical characteristic of each individual character, the ability to solve a problem individually as either male or female and both males and […]
• The “Welfare Queen” Stereotype in the US Reagan’s portrayal of these ladies was used to justify real-world policy changes and contributed to the shrinkage of the social safety net.
• Stereotypes in United Kingdom A stereotype is a common or popular belief about certain people or behaviors of certain individuals. People from different cultures have different stereotypes.
• Stereotype of a Black Female In the following paper, three stereotypes that I have faced in my life will be addressed in terms of the reasons for their formation and the mistakes that lie at the heart of these stereotypes.
• To Be Disabled: Stereotype Analysis The purpose of this paper is to examine, how the stereotype is reinforced in the world, and how disabled people experience it.
• The Dynamics of Stereotype Priming and Assimilation The activation of a mental representation of a social group leads to behaviour corresponding to specific attributes of the stereotype. For priming a stereotype some researchers have held that accessibility of the information and the […]
• The Male Bashing Stereotype: Formal Critique All of the mistakes and lack of social molding that they show women during their youth are not the stuff that dictates the kind of men they will be in the future.
• Racism Issues: Looking and Stereotype In order to find the answer to this question, it is important to introduce the concept of ‘looking’ supporting with the writing of Sturken and Cartwright, Hall, Goodwin, and Gooding-Williams.
• Women and the Industry of the Trap Music: Empowering or Succumbing to the Stereotype? Indeed, on the further scrutiny of the problem, one will see that the issue of female DJs in the trap music domain In light of the specified argument, one may infer that abandoning the trap […]
• Stereotype of Video Games Being for Boys In the article author speaks about the problem of different video games that designed for boys and for girls. In this article author explains that gender difference in the video games is a marketing strategy […]
• Stereotype Threats and Social Psychology Pickren defines social norm as “The rules of behavior that are considered acceptable in a group or society”.to the society, it was acceptable to treat the immigrants differently from the rest of the population because […]
• Stereotype of Aboriginals and Alcohol in Canada Therefore, it is necessary to research whether the given prejudice has certain grounds to base on, track the measures that are being currently undertaken to eliminate the stereotype and offer other efficient ideas that will […]
• Stereotype-Conductive Behavior The notion that fat people are lazy is because many of them avoid doing activities that would require them to spend a lot of energy and movement. In many cases, the speed of fat people […]
• Chinese Stereotypes Reflected in Movies The main research objective will be to: “Analyse Chinese stereotypes in movies” The specific objectives will include: To identify the various stereotypical depictions of the Chinese in movies To determine the relationship between Chinese stereotype […]
• White Female Stereotypes in Media In most instances, the images that are in the media are of exceptionally slim white girls and women, and this sends a negative image to those women that have bigger bodies.
• Review of Stereotype Threat and Arousal: Effects on Women’s Math Performance The variables used in the study were gender, difficulty of the tests, and the perception of stereotype threat. The results of the data were that the implication of stereotype threat did in fact negatively affect […]
• Stereotypes in the media Media has continued to group people by their tribes and the effects of the tribal stereotype is mostly felt in the less developed world.
• Influence of activating implicit gender stereotypes in females The results revealed that the participants who were subjected to the gender based prime performed relatively poorly compared to their counterparts on the nature prime.
• “Stereotype Threat: Effects on Education” by Smith, Cary Stacy, and Li-Ching Hung In some cases, only the topic of these sources is similar to that of the article and not their subject matter.
• Stereotypes of Islam and Muslims in the West This was evident after Shadid made analyses of various publications which analyzed the threat of Islam and the Muslim community to the western countries and fashion such stereotypical messages in the realm of myth.
• How Anthropology Helps to Evaluate Stereotypes The recent study on leadership shows that women have been enlightened and they are up to take their positions in leadership.
• Towards Evaluating the Relationship Between Gender Stereotypes & Culture It is therefore the object of this paper to examine the relationship between gender stereotypes and culture with a view to elucidating how gender stereotypes, reinforced by our diverse cultural beliefs, continue to allocate roles […]
• Stereotypes people have toward Chinese Most of these studies focus on the major stereotypes held about the Chinese but forget to address the effects of these stereotypes to the Chinese students especially the ones studying in other countries.
• Importance of Stereotypes in Communication People are eager to use their prior knowledge about different ethnic groups to be ready for communicating, still, the impact of stereotypes cannot be pure negative or pure positive, and this is why it is […]
• How contemporary toys enforce gender stereotypes in the UK Children defined some of the physical attributes of the toys.”Baby Annabell Function Doll” is a likeness of a baby in that it that it has the size and physical features of a baby.
• Stereotype Threat: Women’s Abilities in Math On the other hand, in study 2, they demonstrated that it is possible to reduce the performance differences when elimination of the stereotype that is descriptive of the anticipated performance is done to ensure that […]
• Hoodies and the stereotype. Bad or not? The hoodie marches had a lot of racial undertones, but it is clear that the victim’s piece of clothing was the centre of attention in these campaigns.
• Gender Studies: Gender Stereotypes From what is portrayed in the media, it is possible for people to dismiss others on the basis of whether they have masculinity or are feminine.
• Cross-Cultural Interaction: Prejudices and Stereotypes In this regard, the concept of stereotype also influences social categorization and information sharing in the course of cross-cultural communication. One of the most effective ways to exterminate stereotypic and linear thinking is to change […]
• Aspects of Rhetoric and Stereotype Image It is clear then, that feminists are found to be of negative stereotypes from the start. The stereotypes in this group are a complete revelation of both positive and negative image.
• African-American Students and Mathematics Achievement Gap: Stereotype or Reality? The purpose of this research is to find whether there is the evidence of the math performance gap between Black and White students and, if we find that it exists, to throw ling upon its […]
• Sex, Lies, and Stereotypes: Being Prejudiced Because of Inequalities Is Not Always Correct The exhibition under consideration, Sex, Lies and Stereotypes, is aimed to prove how unfair but still constant discrimination of people is; and several illustrative posters like Women Are Not Chicks or Oh, So That Explains […]
• Learning to Stereotype: The Lifelong Romance One of the most enchanting novels in the American literature, the piece by Cahan offers a plunge into the world of the usual.
• Stereotypes of American Citizens McAndrew and Akande lament that in the United States, African Americans are the most stereotyped due to racial discrimination and the dark history of slavery.
• Gender Stereotypes on Television Gender stereotyping in television commercials is a topic that has generated a huge debate and it is an important topic to explore to find out how gender roles in voice-overs TV commercials and the type […]
• The Stereotype Of A Smart High Achieving Asian American
• Racial Stereotyping : A Stereotype, As Defined By The Merriam
• Prejudice, Stereotype, Discrimination, and In-Group Vs. Outgroup
• The Sports Media and the Marketing Advertisers a Hypermasculine Stereotype
• Think like a Monkey: Borrowing from Animal Social Dynamics to Reduce Stereotype Threat
• The Metamorphosis Of The Schemer Stereotype
• How Stereotype Threat May Cause Poor Performance In Women
• Women Are Worse Drivers than Men Stereotype
• What Is The Function Of Racist Stereotype In Blackface Minstrelsy
• How Race And Stereotype Can Affect Justice Being Served
• The Imposition of Gender Stereotype by Society Today
• Women’s Oppression In Hurston’s “Sweat”: The Stereotype Of Women’s Role In Society
• Understanding the Gender Stereotype of the Macho-Man Myth
• Use Of A Stereotype Cue On The Perceived Level Of Mathematics
• The Stereotype of Women in a Patriarchal Society
• The Stereotype of Female Taming in Shakespeare’s Time in the Taming of the Shrew
• The Stereotype of the Dumb Blonde in Legally Blonde, a Movie by Robert Luketic
• Americanization : The Creation Of The Indian Stereotype
• The Impact of Stereotype Threat on Age Differences in Memory Performance
• Sexually Driven Media Advertisements Objectify And Stereotype
• Advantage and Disadvantage of Fitting Into the Stereotype
• An Analysis of the Stereotype of Masculinity in the Early 1800s
• Analyzing How a Conventional or Stereotype Character Functions to Achieve Authors Purposes
• Perspective and Stereotype in Western Detective Novels
• The Stereotype Of Criminally Disposed People In Poverty
• Women ‘s Portrayal Of Women Essay – Brand, Marketing, Stereotype, Gen
• Feminine Autonomy and Erasing the Male Stereotype in Juno and the Paycoc
• The History of Chief Illiniwek as a University of Illinois Mascot and Racist Stereotype
• Women ‘s Role For Society ‘s Stereotype Towards Women
• Why Stereotype Based on Blood Type Genotype or Body Type?
• Do Television Advertisements Stereotype the Roles of Men and Women in the Society
• An Analysis of Stereotype Italian American in Sopranos the Cable Show in United States
• Women: Does Stereotype Threat Affect Their Ability?
• American Cheerleader: The Icon, The Stereotype, And The Truth
• Alice Sebold And The Stranger Stereotype
• An Analysis of the Negative Stereotype of the Jewish Race in Jewbird and The Last Mohican
• The Stereotype African Characters in Heart of Darkness by Joseph Conrad
• The Impact Of The Violent African American Stereotype In Rap Music
• The Teenage Driver Stereotype in Society
• Breaking the Stereotype: Why Urban Aboriginals Score Highly on Happiness Measures
• An Analysis of the Macho-Men Stereotype Plaguing Today’s Man
• The Problems of the Aboriginal People and the Average Media Stereotype
• How Racialized Stereotypes Determine a Community’s Value?
• What Is a Cultural Stereotype?
• How Advertising Reinforces Gender Stereotypes?
• How Stereotypes for Women Came to Be?
• How Do Contemporary Toys Enforce Gender Stereotypes?
• What Are Social Stereotypes?
• Are Continuum Beliefs About Psychotic Symptoms Associated With Stereotypes About Schizophrenia?
• How Do Hispanic Bilinguals’ Cultural Stereotypes Shape Advertising Persuasiveness?
• How Do Racialized Stereotypes Determine a Community’s Value?
• How Does Drag Affect Stereotypes About Gay Men?
• How Refugee’s Stereotypes Toward Host Society Members Predict Acculturation Orientations?
• Why Are Female Stereotypes in Advertising Still Effective?
• Can Gender Quotas Break Down Negative Stereotypes?
• Does Mainstream Media Have a Duty to Challenge Gender Stereotypes?
• How Have Gender Stereotypes Always Been a Part of Society?
• How Do Attitudes and Stereotypes Develop?
• Are Sexist Attitudes and Gender Stereotypes Linked?
• Are Gender Stereotypes Perpetuated in Children’s Magazines?
• What Are Gender Stereotypes?
• How Gender Stereotypes Warp Our View of Depression?
• How Are Class Stereotypes Maintained in the Press?
• How Can Bob Dylan and Wolf Biermann Be Employed to Make Students Aware of Stereotypes and Prejudice?
• How Do Racial Stereotypes Affect Society?
• How Did Photography Reflect the Values and Stereotypes That Underlay European Colonialism?
• How Can Stereotypes Contribute to Inequality?
• What Makes People Have Certain Stereotypes?
• How Can Stereotypes Negatively Affect Listening?
• Why Are Stereotypes Dangerous and What Can Be Done to Reduce Them?
• How Are Stereotypes Used to Racially Profile People?
• How American Minorities Are Stereotypes in American Drama Series?
• Sociological Imagination Topics
• Orientalism Titles
• Racism Paper Topics
• Sociological Perspectives Titles
• Tolerance Essay Ideas
• Sexism Essay Ideas
• Gender Roles Paper Topics
• Self-Concept Questions
• Chicago (A-D)
• Chicago (N-B)

IvyPanda. (2023, October 26). 113 Stereotype Essay Topics & Examples. https://ivypanda.com/essays/topic/stereotype-essay-examples/

"113 Stereotype Essay Topics & Examples." IvyPanda , 26 Oct. 2023, ivypanda.com/essays/topic/stereotype-essay-examples/.

IvyPanda . (2023) '113 Stereotype Essay Topics & Examples'. 26 October.

IvyPanda . 2023. "113 Stereotype Essay Topics & Examples." October 26, 2023. https://ivypanda.com/essays/topic/stereotype-essay-examples/.

1. IvyPanda . "113 Stereotype Essay Topics & Examples." October 26, 2023. https://ivypanda.com/essays/topic/stereotype-essay-examples/.

Bibliography

IvyPanda . "113 Stereotype Essay Topics & Examples." October 26, 2023. https://ivypanda.com/essays/topic/stereotype-essay-examples/.

• Tools and Resources
• Customer Services
• Communication and Culture
• Communication and Social Change
• Communication and Technology
• Communication Theory
• Critical/Cultural Studies
• Gender (Gay, Lesbian, Bisexual and Transgender Studies)
• Health and Risk Communication
• Intergroup Communication
• International/Global Communication
• Interpersonal Communication
• Journalism Studies
• Language and Social Interaction
• Mass Communication
• Media and Communication Policy
• Organizational Communication
• Political Communication
• Rhetorical Theory
• Share This Facebook LinkedIn Twitter

Article contents

Culture, prejudice, racism, and discrimination.

• John Baldwin John Baldwin School of Communication, Illinois State University
• https://doi.org/10.1093/acrefore/9780190228613.013.164
• Published online: 25 January 2017

Prejudice is a broad social phenomenon and area of research, complicated by the fact that intolerance exists in internal cognitions but is manifest in symbol usage (verbal, nonverbal, mediated), law and policy, and social and organizational practice. It is based on group identification (i.e., perceiving and treating a person or people in terms of outgroup membership); but that outgroup can range from the more commonly known outgroups based on race, sex/gender, nationality, or sexual orientation to more specific intolerances of others based on political party, fan status, or membership in some perceived group such as “blonde” or “athlete.” This article begins with the link of culture to prejudice, noting specific culture-based prejudices of ethnocentrism and xenophobia. It then explores the levels at which prejudice might be manifest, finally arriving at a specific focus of prejudice—racism; however, what applies to racism may also apply to other intolerances such as sexism, heterosexism, classism, or ageism.

The discussion and analysis of prejudice becomes complicated when we approach a specific topic like racism, though the tensions surrounding this phenomenon extend to other intolerances such as sexism or heterosexism. Complications include determining the influences that might lead to individual racism or an atmosphere of racism, but also include the very definition of what racism is: Is it an individual phenomenon, or does it refer to an intolerance that is supported by a dominant social structure? Because overt intolerance has become unpopular in many societies, researchers have explored how racism and sexism might be expressed in subtle terms; others investigate how racism intersects with other forms of oppression, including those based on sex/gender, sexual orientation, or colonialism; and still others consider how one might express intolerance “benevolently,” with good intentions though still based on problematic racist or sexist ideologies.

• discrimination
• intolerance
• heterosexism
• stereotypes
• ethnocentrism

Introduction

One of the causes that gave rise to the postmodern revolution in France in 1968 was the failure of modern science and philosophy—liberalism, social science, reason, and so on—to remedy problems of war, poverty, and intolerance (Rosenau, 1992 ). As we look around today at the world in general, or even within specific nations, we continue to see a wide range of prejudice, from the 1994 genocide of Tutsis (and many Hutus) by Hutus in Rwanda to the mass killing of 70 people, mostly youths, at a Utøyan youth camp in Norway by Anders Behring Breivik. At this writing, a major refugee problem exists from people fleeing Middle Eastern countries where a strong ISIS influence is leading to the killing of gays, Christians, and Muslims from rival belief systems. In many European countries, hate groups and right-wing politicians are gaining ground. The Southern Poverty Law center tracks 1,600 hate groups within the United States (“Hate and Extremism,” n.d. ), classifying 784 that were active in 2014 (“Hate Map,” n.d. ), and the FBI reports nearly 6,000 hate crimes in the United States, with the greatest numbers due to race (48.5%), religion (17.4%), and sexual orientation (20.8%; FBI, 2014 ). These statistics reveal some interesting things about intolerance. For example, the “race”-based hate crimes include crimes based on anti-white sentiment as well as against people of color; and about 61% of hate crimes based on sexual orientation target gay males.

Both the international events and the statistics relevant to any specific nation prompt difficult questions about intolerance. In a white-dominant society, can or should we call anti-white crimes by people of color “racist”? If someone commits a hate crime based on sexual orientation, why are gay men more often the target than lesbians? Would hate crimes in other countries reflect the same axes of difference, or might hate crimes be based differently? German hate crimes might be based more on ethnicity (e.g., against Turkish immigrants, who by most racial classifications would be Caucasian). Why do people commit such acts at all?

One mistake we often make is thinking of prejudice and discrimination only in extreme terms such as genocide and hate crimes. In many countries and cultures, where overt expression of racism (and other intolerances) has become socially unacceptable, intolerances have gone “underground,” hidden in subtle forms. Further, intolerance can rely upon a wide variety of identity groups, including some that are (supposedly) biologically based, like racism, or based on other aspects, such as political party, fan status, or membership in some perceived group such as “blonde” or “athlete.” In sum, we must consider the relationship between different forms of intolerance, including but not limited to prejudice, racism, and discrimination; but these must always be understood within specific cultural contexts.

Culture and Intolerance

(re)defining culture.

As we look to the cultural influence on intolerance, we must first consider the definition of culture. The study of culture has deep roots in anthropological and linguistic research, especially as seen in the work of Franz Boaz and his students Margaret Mead, Ruth Benedict, and Edward Sapir, as well as in the early work of Edward Tyler, itself based on earlier traditions of ethology (Darwin) and social evolution (Marx). This work influenced the work of anthropologist E. T. Hall (Rogers & Hart, 2002 ) and others who laid the groundwork for the study of intercultural communication (Leeds-Hurwitz, 1990 ). Scholars have debated whether culture is a shared mental framework of beliefs, norms for behavior (i.e., the expectations for behavior rather than the behaviors themselves), values, and worldview, or whether culture should also include actual behaviors, texts, and artifacts of a group. In 1952 , A. L. Kroeber and Clyde Kluckhohn synthesized over 150 definitions of culture into a single definition that focuses on “patterns, explicit and implicit, of and for behavior,” along with “ideas and especially their attached values” (p. 181). These are influenced and created through symbolic behavior, action, and other aspects of the environment (history, geography). The definitional dimensions of culture described by Kroeber and Kluckhohn explained well many of the definitions of culture up until the 1980s. After that time, some scholars (especially in communication) began to treat culture more as a set of symbols and meanings. Others framed culture as a process of constructing social meanings and systems through communication. As people sing, speak, play, tell jokes, and conduct business, they are constantly (re)creating their culture—both relying upon it and changing it.

More pertinent to the study of intolerance is a new approach to culture that sees culture neither as “suitcase” of things (be those beliefs and values or texts and artifacts) passed down from one generation to the next nor as a neutral process of mutual symbolic creation through time, but as having vested power interests that seek to influence what is seen as accepted or normal within a culture. For example, Moon ( 2002 ) defines culture as a “contested zone”:

Thinking about culture as a contested zone helps us understand the struggles of cultural groups and the complexities of cultural life … If we define culture as a contested zone in which different groups struggle to define issues in their own interests, we must also recognize that not all groups have equal access to public forums to voice their concerns, perspectives, and the everyday realities of their lives” (pp. 15–16).

That is, every cultural manifestation, such as the framing of Australian culture as “individualistic” or saying that “Australian men have such-and-such characteristics,” highlights what one should not be within that culture and establishes bounds for group-based intolerance.

With this diversity of definitions in mind, one is not sure what to think culture is or should be. Baldwin, Faulkner, Hecht, and Lindsley ( 2006 ) present a series of essays on the definition of culture by authors from six different disciplines (e.g., multicultural education, anthropology, political science), as well as 313 definitions of culture from an even greater number of disciplines, which they analyze. While they are reluctant to settle on a single definition of culture, this definition embraces most trends:

The way of life of a group of people, including symbols, values, behaviors, artifacts, and other shared aspects, that continually evolves as people share messages and is often the result of a struggle between groups who share different perspectives, interests, and power relationships (Baldwin, Coleman, González, & Shenoy-Packer, 2014 , p. 55).

This definition of culture, like most definitions that take a symbolic, process, or critical approach, does not treat cultures as “nations,” but as people groups who share symbolic or speech codes, with multiple cultural groups—defined not by demographic constitutions such as race, sex, or age, but by shared communicative realities—sharing single geographic areas. It is in the creation and defending of cultures—from countries to local and virtual communities—that intolerance often becomes apparent.

The Role of Culture in Prejudice

Of various schools of thought about the nature and origins of intolerance, only one approach suggests that intolerance is biological or in some way inherited, and that is sociobiology, or evolutionary theory. This approach suggests that intolerance is based on such things as preservation of the purity of the gene pool of one’s group, an inherent fear of strangers, or an inherited need for group identity. But even evolutionary theorists cannot explain all intolerance based on a theory of inherited impulse. Meyer ( 1987 ) argues:

Xenophobia and ethnocentrism as extreme forms of this search for identity cannot be attributed to [human] biology … Their very existence is a result of [human’s] attempts towards understanding the world, and [their] strong affective need to delimit a cosmos of conspecifics with whom [they] can share interpretations of [their] socially construed world (p. 93).

Research on intolerance in 90 preindustrial societies suggests that, when there are clearly psychological causes for intergroup conflict, groups ultimately use communication to create who the enemy is and how one should demonstrate or show intolerance (Ross, 1991 ). In sum, there is a strong cultural component determining which intolerances are felt or expressed in a given place or time.

Culture, however one defines it, can affect tolerance. Culture might be a set of values and beliefs, such as the value of loyalty to one’s group, combined with a belief that people who belong to a particular group have particular characteristics, are unlikeable for some reason, or merit mistreatment and the application of a different set of standards than we apply to ourselves (Opotow, 1990 ). If culture is a process, then we might look at how a culture creates both identity and intolerance through the ongoing structures of language, including word choices (“babe,” “hunk,” “faggot”), conversational structure (interruptions, etc.), joke- and storytelling, and so on. For example, West and Zimmerman’s ( 1987 ) notion of “doing gender” (i.e., gender as an everyday accomplishment of language) has led to countless studies of gender construction in several nations, as well as a focus by others on how we also “do race” and other identities. The way that we construct our identities through communication is inherently linked to how we construct the identities of those in outgroups, as we shall see; but they are also linked to behavior within our group. Social constructionist approaches to culture thus often become critical in their focus on power relations. Critical approaches look at how cultures, through communication, architecture, law, literature, education, and so on create a sense of the “other”—and of the self—that constrains us and pits us against one another in group conflict.

“Culture”-Based Prejudices: Ethnocentrism, Xenophobia

The purpose of this article is primarily to look at racism and discrimination as forms of prejudice; however, these cannot be understood without a larger understanding of prejudice in general and other forms or types of prejudice. Allport ( 1979 ) defines prejudice as an antipathy one has or a tendency to avoid the other, based on the other person’s group. For Allport, prejudice is a cognitive or psychological phenomenon:

Prejudice is ultimately a problem of personality formation and development; no two cases of prejudice are precisely the same. No individual would mirror his [or her] group’s attitude unless he [or she] had a personal need, or personal habit, that leads him [or her] to do so (p. 41).

Based on the Greek word that means “fear of strangers,” xenophobia refers to “the fear or hatred of anything that is foreign or outside of one’s own group, nation, or culture” (Herbst, 1997 , p. 235). The idea is frequently applied to a mistrust or dislike (rather than merely fear) of outgroups or those perceived to be different, especially in national terms. While the Greek translation suggests the psychological component of fear, recent researchers have treated the concept in behavioral or message terms. Historical research on xenophobia links it to anti-Semitism and, more recently, to Islamophobia, though it does not have as clear a historical trajectory as ethnocentrism; many more recent studies look at South Africa as a model nation in attempting to strategically reduce xenophobia. Researchers use a variety of methods to look at xenophobia, depending on their research assumptions and background disciplines. Rhetorical media research, for example, analyzes how Czech newspapers code anti-Roma sentiment through subtle terms such as “inadaptable citizens” ( nepřízpůsobivý občan , Slavíčková & Zvagulis, 2014 , p. 159); and psychological survey research investigates how, among Southern California students, ethnocentrism is positively associated with both language prejudice and feelings of being threatened by immigrants (Ura, Preston, & Mearns, 2015 ).

Van Dijk ( 1993 ) notes how groups can use language such as hyperbole of differences to marginalize immigrants, often through appeals to so-called democratic values. He notes that in some countries, such as in Central Europe, where claims of racism are often forcefully resisted due to conceptual ties of the term to Hitler’s Holocaust, Ausländerfeindlicheit (fear of foreigners) takes its place, though this fear of foreigners is frequently aimed at Turks and other (often darker-skinned and religiously different) people who resist adoption of traditional Germanic culture.

Ethnocentrism

Some types of prejudice relate specifically to the larger and more traditional notion of culture (i.e., cultures as nations). Ethnocentrism gained prominence as an area of research following sociologist Robert Sumner’s 1906 definition of the term as gauging others in reference to one’s own culture ( 1975 ), though other sociologists soon began to distinguish between this notion of “centrality” and the idea of “superiority”—that one’s culture or group is superior to those of others. If one sees ethnocentrism strictly as a feeling of superiority, nationalism (or school spirit, or religious loyalty, etc.) might not in and of itself be ethnocentric if it focuses only on being loyal to or highlighting the benefits of one’s own group, without denigrating others, though some might argue that it is impossible to feel pride in one’s own group without, at some level, disdaining or thinking less of other groups. The possibility of an ethnocentric bias in research led many early anthropologists to suggest ethnography—spending extended time within a culture to see things from cultural members’ point of view—as a way to reduce ethnocentrism in research.

A consideration of ethnocentrism has implications for other forms of bias as well, as the factors that predict national cultural ethnocentrism—and solutions that address it—could apply equally to one’s perception of life within one’s own community. The Hmong-descended people of the Pacific Northwest in the United States will likely feel that their ways are superior to those of Moroccan- or Guatemalan-descended peoples, as well as to those of the dominant culture. Auestad ( 2013 ) presented a series of essays on the rise of political discourses across the world that highlighted elements of national security and identity (tradition), as well as the building of cultures of fear by focusing on the negative aspects of foreigners or those of different religious groups within single countries. Some elements of the U.S. presidential race rhetoric of 2015–2016 exemplified this xenophobic and ethnocentric trend.

Within the field of intercultural communication, at least two lines of research have focused on ethnocentrism. The first is by Jim Neuliep, who, with colleagues, has revisited the measurement of ethnocentrism in the classic 1950 work by the Frankfurt School, The Authoritarian Personality , with a new measure of ethnocentrism. After applying the measure to white Americans, Neuliep ( 2012 ) continues to test the relationship of ethnocentrism to other important intercultural variables, such as intercultural anxiety and communication satisfaction. The second is Milton Bennett’s ( 1993 ) consideration of ethnorelativism. In this approach, a range of attitudes reflects either ethnocentrism or ethnorelativism. Ethnocentric stances include denial (e.g., indifference toward or ignorance of any difference at all), defense (traditional ethnocentrism of denigrating the culture of the other or feeling one’s own culture is superior, but also in “going native”), and minimization (focusing on similarities and ignoring differences, by claiming “color blindness,” or focusing on how we are all the same, be that as “God’s children” or in the Marxist struggle against oppression; 43). As one grows more “ethnorelative,” or accepting of difference, one exhibits one of three stages: acceptance (being respectful of and even appreciating the value and behavioral differences of others), adaptation (actually adopting behaviors or views of other groups), or integration (adopting a worldview that transcends any single culture). This approach has gained ground around the world and in different disciplines, from Finland to Iran, with applications from cultural sensitivity to interreligious tensions.

One of the difficulties of discussing prejudice is the conceptual overlap between terms (e.g., xenophobia conflates with racial or ethnic prejudice; ethnocentrism might refer to any people group, such as ethnic groups, and not just nations). At the root of our understanding of prejudice is the very goal of “tolerance.” In fact, the notion of tolerance for diversity may be limited: It is often treated merely as “the application of the same moral principles and rules, caring and empathy, and feelings of connections to human beings of other perceived groups” (Baldwin & Hecht, 1995 , p. 65). That is, it is similar to Bennett’s ( 1993 ) notion of acceptance, of respect for difference, though that respect sometimes (a) occurs at a difference and (b) sometimes exists in behavioral form only, but is not internalized. Communication of tolerance is a worthwhile pursuit in our behavior and research; however, we argue that we can go beyond tolerance to appreciation—even to the behavioral and attitudinal integration of elements of the other culture (Hecht & Baldwin, 1998 ). There is a danger of such appreciation, as borrowing (e.g., “cultural hybridity”) occurs within power relations. We are not talking about a dominant group borrowing from subordinate or subaltern groups in a colonizing or folklorizing way, but about cultural learning and dialogue.

Limited Perspectives of Prejudice

That consideration of tolerance/prejudice should be treated as a dichotomy or a range is only one of the difficulties that has haunted the study and conceptualization of prejudice. Debates have swirled around the nature of prejudice, the causes of prejudice, and the “locus” of certain prejudices (such as racism or sexism), among other things. Allport ( 1979 ) suggests that prejudice is a “generalized” attitude—that if one is prejudiced, say, toward Jewish people, she or he will also be prejudiced toward communists, people of color, and so on. It is possible, however, that one might be prejudiced toward some groups, even in some contexts, but not toward other groups (Baldwin & Hecht, 1995 ).

The nature of prejudice

Allport ( 1979 ) defines prejudice as “an avertive [i.e., avoiding] or hostile attitude toward a person who belongs to a group, simply because he [or she] belongs to that group, and is therefore presumed to have the objectionable qualities ascribed to the group” (p. 7). By this definition, prejudice is an aspect of affect , or feeling toward a group, though it is closely related to cognitions , or thoughts about the group, referring to stereotypes. Also, prejudice is inherently negative, following the primary definition common in modern dictionaries, though a secondary definition includes any sort of prejudgment based on group belonging, such as prejudice toward one’s own group. Most dictionary definitions follow the attitudinal approach, though in common usage, people often use the term to refer to things like racism, which carry behavioral and even policy implications that are not strictly attitudes. By strictest definition, prejudice is an attitude that favors one group over another, based on or related to cognitions, and both leading to and influenced by behaviors (including communication), texts (e.g., media, rhetoric), and policies (following the notion of structuration, in which social structures guide social behavior, but social behavior in turn creates and changes social structures).

Causes of prejudice

Allport ( 1979 ) recognized a series of influences that impact a particular incident of prejudice, such as police brutality based on racial group/social class divisions or anti-Islamic bullying in secondary schools around the Western world. These include historical, sociological, situational, psychodynamic, and phenomenological (i.e., perceptual) influences. But ultimately, for Allport, a social psychologist, prejudice is “a problem of personality formation and development” (p. 41). For Althusser ( 1971 ), a Marxist philosopher, prejudice would likely, in the last instance, be an issue of economic and social class considerations. Ultimately, a cross-disciplinary perspective is more useful for understanding a complex phenomenon like prejudice (Hecht & Baldwin, 1998 ). A broader consideration should consider multiple causes (Baldwin, 1998 ), including evolutionary causes, psychological causes (both psychodynamic and perceptual), sociological causes, and rhetorical causes. Communication and behavior become central in each of these causes, highlighting the need for a communicative understanding of prejudice.

Evolutionary causes, often referred to under the rubric of sociobiology, focus on the way in which prejudice might be an inherited trait, possibly even genetic (see, e.g., essays in Reynolds, Falger, & Vine, 1987 ). This approach includes the idea that groups seek to preserve themselves (e.g., by preservation of a supposedly pure gene pool or because of fear of the stranger), the ethnocentrism already noted. Behaviors that exclude have a sense of “naturalness” in that they help a group to survive, and such exclusion of strangers may help to preserve a group’s existence. Some scholars have criticized this approach as a rationale for conservative politics that create a notion of “us” and “them” as natural and that exclude the other, often in racial or religious terms, in order to preserve the way of life of a dominant group within a culture or nation.

Psychological explanations of prejudice fall into at least two major divisions. The first, psychodynamic, suggests that prejudice serves as a mechanism for individuals to meet psychological needs. Thus researchers have long linked it to things such as ambivalence toward parents, rigid personality structure, and a need for authority (Allport, 1979 ; Adorno et al., 1950 ). We see this indirectly through Kenneth Burke’s ( 1967 ) approach to rhetoric in his analysis of Hitler’s campaign against Jewish people as a means to divert negative emotions related to economic and political difficulties from the mainstream German people to Jews, and in Edward Said’s ( 2003 ) Orientalism , which notes how Medieval Europe cast negative images of lust and vice on Middle Easterners that the Europeans did not see in themselves.

A second aspect of the psychological approach concerns perception or cognition. This contains a range of possible influences on prejudice, including such things as selective attention, perception, and recall of the negative behavior of outgroup members, or the notion of attributional biases that impact how we give meanings to the behavior of those of our ingroup and those of outgroups. At the center of many of these explanations is the notion of categorization of people (i.e., dividing them into cognitive groups such as ingroups and outgroups). Social identity theory (Tajfel & Turner, 1986 ) suggests that we cannot think of ourselves apart from the groups to which we belong; we engage in intergroup comparison as a means to make us feel better about our group; and, if our group does not compare well to a group we admire or must rely on in some way—often the dominant group—we engage in strategies to reclaim a sense of pride for our group or distance ourselves from it.

Categorization, in social identity theory, is not a form of prejudice—it is simply the mental placing of people (or things, actions, characteristics, etc.) into mental boxes. However, those boxes are closely related to the stereotypes that cling to groups. Stereotypes are overgeneralizations we make about groups that we apply to individuals in those groups (Herbst, 1997 ). Although these stereotypes provide a mental shortcut for processing information about others, they interfere with our encoding, storage, and recall of information about members of our own group and other groups (Stephan, 1985 ). Countless studies of stereotypes suggest that stereotypes, like ethnocentrism, can serve positive ingroup functions, that they sometimes have at least some basis in an actual behavior or custom (a “kernel of truth”), and that we stereotype both our own group and other groups. Devine (e.g., Devine & Sharp, 2009 ) has found that even people who report lower prejudice, if mentally occupied, still rely on stereotypes, suggesting that everyone is aware of societal stereotypes toward certain groups (e.g., the elderly, athletes, the deaf). It is likely that if we are on auto-pilot or in a state of mindlessness, we will resort to stereotypes. But individuating people (i.e., taking them out of the group we perceive them to be in and treating them as individuals; Dovidio, Gaertner, & Kawakami, 2003 ) may require deliberate cognitive effort.

Group-based, or sociological, approaches, like psychological approaches, are varied. These include Marxist approaches, which are themselves varied in form (see various essays in Rex & Mason, 1986 ). Some hold tightly to a “vulgar” vision of Marxism, framing intolerance like racism as a creation of the elite to divide the working classes and distract them from revolution through “false consciousness.” Few Marxists take such a severe approach, choosing to see looser relations between capital and the construction of intolerance, but in the “last instance,” seeing intolerance as linked to social class and economic systems. “Capitalism, colonialism, and patriarchal social systems are frequently identified as producing inherent race and gender inequalities which, in various ways, serve the needs of the systems they perpetuate” (Knowles & Mercer, 1992 , p. 110). Weberian approaches see a wider variety of classes than workers and elite, with prejudice linked not just to labor forces but to the struggle over goods, services, and prestige (Gerth & Mills, 1946 ). Other group-based factors also impact prejudice, such as perceived group competition for jobs and resources in times of economic upheaval (e.g., the 1970s oil crisis in the United States), known as realistic group conflict (Bobo, 1983 ); immigration reasons (refugees versus those seeking economic opportunity, patterns of settlement; Omi & Winant, 1986 ); and historically developed class statuses between groups that link immigrants or members of a minority group to a certain class (Wilson, 1978 ), such as the Gastarbeiter (guest-worker) Turks in Germany or the Algerian-descended French.

In a classic “chicken-egg” argument about which came first, it is fruitless to debate whether psychology leads to sociological causes or vice versa, and, in turn, whether these lead to the communicative expression of intolerance, or whether it is the communicative construction of group identities and intolerance that creates the attitudes (Ruscher, 2001 ). It is more likely that mental structures and communicative practices co-create each other, through forms we shall examine in more detail. One possible metaphor for understanding these influences, the impact of historical situations (such as the longstanding antipathy between Turkish and Greek Cypriots, Broome, 2005 ), and specific incidents (such as the attack on the World Trade Towers in New York City in 2001 ), is as layers building upon one other, or even as a hologram, in which we can imperfectly see some semblance of a complex prejudice through a single image—an experimental study on racial perceptions and media use, an analysis of an anti-Irish speech or a pro-nationalist song, or interviews with women who are victims of catcalling (Hecht & Baldwin, 1998 ). But, as a complete hologram provides the most faithful image, the most complete view of an intolerance will come through multiple views (e.g., disciplines), using multiple methods.

Racism: A Case Study in Prejudice

Racism as a specific type of prejudice is one of the most hotly discussed and debated sites of intolerance in contemporary times in the United States and beyond. Even countries that once imagined themselves as “racial democracies” in which racially different people lived side by side (like Brazil) are now admitting the harsh reality of entrenched and historic racism. Even though many there argue that class, not race, is the primary social distinction, as racism has become officially illegal, forms of overt racism, from social media to abuse and killing of unarmed blacks by police continue to receive recent focus in U.S. news.

Racism is a form of intolerance that is based on the supposedly biological distinction of race, but many authors today argue that race is a social construct, sometimes defined differently from country to country and even over time within a single country. Different authors have outlined the history of the notion of race in the English language, noting that at different times, it has referred to an ancestral clan (the race of Abraham), to supposed biological differences, and, more recently, to culture (Banton, 1987 ; Omi & Winant, 1986 ). Those who see a biological component cannot agree on how many races there are and, historically, politics and rhetoric have done as much to construct who belongs in a particular race as biology (e.g., in the early U.S., the Irish were considered “colored”). In the United States, race was based on racist assumptions, on one having even a small degree of colored blood in one’s ancestral lineage; in other cultures, race is based strictly on physiological features, regardless of lineage. Ethnicity , in contrast, is related more to the cultural origins of one’s background or ancestry, sometimes linked to a specific time and place. To emphasize its social constructedness, many authors bracket “race” with quotation marks.

Who Can Be Racist? The Locus of Racism (and Other Intolerances)

Can minority members be “racist”.

Beyond the nature of race itself, researchers and educators debate the very nature of racism. Some contend that racism is an intolerance based on the construction of race that is perpetrated and held by the support of the dominant system. For example, Malott and Schaefle ( 2015 ) define racism as “a system of oppression, whereby persons of a dominant racial group (whites in the United States) exercise power or privilege over those in nondominant groups” (p. 361). According to this argument, only whites can be racist in a white-dominated system (whether that dominance is by numbers or in political and social power). Others contend that racism is any system of beliefs—“held consciously or otherwise”—that treats members of a group that is different on supposedly biological grounds as “biologically different than one’s own” (Herbst, 1997 , p. 193). By this definition, anyone who sees another race group as inferior would be racist.

The locus of racism: Individual or structural?

This distinction in racism also applies to definitions of sexism or to the delineation between homophobia as a personal dislike or fear of LGBT individuals and heterosexism as a social structure that reinforces prejudice against them (Nakayama, 1998 ). The debate is similar to the definitional debate of prejudice in general—is it something that is strictly an individual trait, or is it something that is socially built into the structures of society—the laws, the media, the educational system, the church, and so on? Associated with this question is the nature of what racism is: The “individual-level” definition treats racism as a system of beliefs (i.e., a psychological construct), and the other treats it as a system of oppression that goes beyond individual psyche and personality to consider racism embedded within social structures. The question of where we see racism (and other intolerances) is vitally important. Those who see racism and other intolerances as primarily individual-level (stereotypes, personal dislikes, etc.) tend to address intolerance through training and educational programs in organizations and schools; those who see it as systemic believe that such approaches ignore larger issues of policy, law, segregation, discrimination, and media/rhetoric that produce and reproduce racist beliefs or create an environment that makes them grow. We see this tension, for example, in Rattansi’s ( 1992 ) discussion of the debate between multicultural education—an educational solution to tolerance focused on educating about differences—and antiracism, which addresses political and social structures that propagate and support racism.

Racism: Defined by intent or result?

A related definitional distinction regarding racism concerns whether an intent of harm or exclusion is necessary to define thoughts or actions as racist. Miles ( 1989 ) criticizes earlier notions of racism, largely in that they re-inscribe the notion of race as if it were a concrete reality rather than a social construction. He weaves together a new approach to racism that begins with discourses that serve to exclude the “other” (based on supposed biological differences); for Miles, “the concept of racism should refer to the function, rather than the content of the discourses” (p. 49), allowing racism to include things that may not sound racist but still seek to exclude the other. Miles differentiates racism from racialization , the categorization of people based on supposed biological differences. He argues against the use of racism and disagrees with a stance that would have only whites being racist, such that “all ‘white’ people are universally and inevitably sick with racism” (p. 53), as this concept may ignore the specifics of racism in particular countries, cultures, or circumstances; however, he notes the need to consider institutional racism—racism built into organizational, legal, and social structures—that does favor whites in many countries. By this, one could speak of racism as something any person could hold or express, but institutional racism would be reserved for a group that has power in a particular context. Finally, he bases racism not on the intent of an action, but on the result. He argues that racism is an ideology, based on differentiation, that leads to “exclusionary practices” (pp. 77–78), such as differential treatment or allocation of resources and opportunities, regardless of one’s intent or even awareness of the ideological underpinnings of one’s actions. Goldberg ( 1993 ) argues that we should allow racism to include either intent or result.

Including resulting exclusionary practice in our definition of racism has implications for redressing or addressing racism. First, it suggests a limitation in addressing overt racist thoughts and stereotypes only through education, as policies, laws, and social structures foster an environment for the presence of such thoughts and their communication. Miles ( 1989 ) advocates that “strategies for eliminating racism should concentrate less on trying exclusively to persuade those who articulate racism that they are ‘wrong’ and more on changing those particular economic and political relations” (p. 82). A second implication is that, even as we seek to address racism through everyday interactions and social media, because racism is such a charged topic, we will advance our cause little by calling an action, a joke, or a Facebook or Twitter posting “racist.” The poster, holding a more traditional view of racism as intentionally harmful in some way, will deny racist intent, and a charge of racism will move the discussion into the original communicator’s attempts to avoid the charge of racism (or sexism, etc.), rather than addressing the specific policy, image, or statement. Instead, we might discuss and demonstrate through evidence the way that the policy or image excludes others based on race. Without invoking the “r-word,” we may have a better chance at engaging in dialogues about policies, laws, and communicative behaviors that exclude others.

Intersectionalities of Racism

As we have begun to notice, one thing that complicates the concept of racism is its overlap with other terms, such as prejudice (with racism being a subset of prejudice). So, although xenophobia and ethnocentrism are distinct and separate from racism, the “other” within these concepts is often articulated or perceived in terms of race. A focus on racism and antiracism, unfortunately, often excludes other bases of intolerance that may be even more prominent within a given area, such as religious intolerance, sexism, or heterosexism. At the same time, it is useful to see how racism intersects with and sometimes leads to other intolerances, all of which have received much thought in recent years.

In some cases, feminists and antiracists have been at odds, proponents of each claiming that their sphere of oppression is the one that merits the most attention. Feminism is defined as “the belief that men and women are equal and should have equal respect and opportunities in all spheres of life—personal, social, work, and public” (Wood, 2008 , p. 324). Feminist communication research seeks to make the voices of women heard, to highlight their experiences within the social construction of gender, and “their experiences of oppression and of coping with and resisting that oppression” (Foss & Foss, 1994 , p. 39). Recent feminists consider how patriarchy, or male power or hegemony over the realities and voices of women, is not something maintained only by men nor is it deliberate. Rather, it is held in place by systems often beyond the awareness of men and women, and consented to and participated in by women themselves (Zompetti, 2012 ). Each of these ideas could also apply to racism, revealing a similarity between sexism and racism. But racism and sexism are also joined in the experiences of women of color, whose specific life situations are not fully addressed by either antiracist efforts or feminism. Collins ( 1990 ), for example, argues that African American women in the United States live in a site of triple oppression—by race, sex, and class, with these oppressions articulated by both the dominant white community and within the black community.

Queer theory

Queer theory seeks to challenge the way in which society passes on heterosexuality as the norm. Warner ( 1991 ) sees oppression of gays and lesbians in every aspect of society and in “a wide range of institutions and ideology” (p. 5). But even more so, he feels that the academy’s silence regarding oppression of sexual identity participates in that oppression. Chávez ( 2013 ) supports this claim, noting that at the writing of her article, no major journal in the National Communication Association had devoted a full issue to queer studies. Again, recent scholars have been looking at the intersection of race and sexual orientation (Yep, 2013 ), such as the representations and experiences of older gay male adults, Latina lesbians, and transgender blacks.

Whiteness studies

Based on the early writings of Richard Dyer ( 1997 ) and Ruth Frankenberg ( 1993 ), researchers have highlighted the notion of whiteness —a hidden system of ideology and social structure that maintains whites in a position of advantage—but one that is often invisible to, and yet defended by, whites (Wander, Nakayama, & Martin, 1999 ). Whiteness studies call attention to areas of white privilege. “By exposing the ‘invisibility’ of whiteness, the study of whiteness helps us understand the way that white domination continues” (p. 22). A current search for “whiteness” in a communication library search engine reveals over 800 articles on the topic. Many of these are media studies on how whiteness is promoted and/or challenged in a wide variety of texts, including South Park , the Rush Hour movies, The Hunger Games , and Glee . But whiteness is also analyzed in areas of education, everyday language, and health and organizational communication, as well as in many different countries.

Orientalism/postcolonialism

whiteness studies owe part of their heritage to postcolonialism, which has its own roots in the conceptualization of Orientalism by Edward Said ( 2003 ). Said analyzes European art and literature to reveal the construction of the Arab or Middle Easterner as “other.” He notes how the Western ideology of the East (referring to the Middle East) folklorizes and sexualizes Middle Easterners, treating them as backward, in a way that justifies European colonization and paternalism. Thousands of books now deal in some way with Orientalism, and Said’s notion of the “other” has become a stock theme in how we consider the racial other. For example, though not framed explicitly in Orientalism, James Baldwin’s famous 1955 essay “Stranger in the Village” talks about the rage of the black man as he confronts white America and the naiveté of whites—a naiveté that they work hard to preserve (thus relating Baldwin’s ideas to whiteness). When whites arrive in Africa, blacks are astonished:

The white man takes the astonishment as tribute, for he arrives to conquer and to convert the natives, whose inferiority in relation to himself is not even to be questioned; whereas I, without a thought of conquest, find myself among a people whose culture controls me, has even, in a sense, created me, people who have cost me more in anguish and rage than they will ever know, who yet do not even know of my existence … The rage of the disesteemed is personally fruitless, but it is also absolutely inevitable: the rage, so generally discounted, so little understood even among the people whose daily bread it is, is one of the things that makes history.

Postcolonialism, building upon Orientalism, considers all locations where one nation or people group has colonized another group, considering the cultural, political, and social ramifications of that colonization and seeking to remedy social ills that it has brought about. Shome and Hegde ( 2002 ) call the approach “interventionist and highly political” (p. 250). Postcolonialism notes how much of the world is forced to work within thought systems created by the Western world (an effect only magnified through the rise of the internet and globalization). Postcolonial writers are often interested in issues such as migration of people groups (including diasporic groups); the hybrid (but power-laden) mixture of ideas, artifacts, and behaviors between cultures; the liminal spaces between cultures; and the imperialism of ideas (Bhabha, 1994 ). Thus, postcolonialism is inherently about prejudice and oppression beyond racism, though it also has links to racism specifically, as authors consider the ways that some have used racial categories to colonize others (e.g., see essays in Nakayama & Halualani, 2010 ).

Discrimination: Considering the Form(s) of Intolerance

As we have seen, it is difficult to discuss prejudice in general or racism specifically without moving into issues of institutionalized prejudice, media representations, school and government policies, and so on. In this sense, both prejudice and racism are intricately intertwined with discrimination. Discrimination specifically refers to “behavior that denies equal treatment to people because of their membership in some group” (Herbst, 1997 , p. 185). It is based on the “beliefs, feelings, fantasies, and motivations of prejudice” (p. 185), but these mental or social concepts are not in themselves discrimination. Discrimination involves behavior.

Institutional Discrimination

When we think of institutional-level discrimination, many examples come to mind. These include things like not allowing certain groups housing or refusing other privileges, resources, or opportunities to them. At the writing of this chapter, a popular U.S. media topic is the county clerk, Kim Davis, who refused to give marriage licenses to gays or lesbians based on her faith, despite a state law that allowed her to do so. The Jim Crowe laws of the United States, which gave unequal educational and public access rights to blacks and whites is a classic example, with many facilities being for “whites only.” The website Global Issues (Shah, 2010 ) details instances of racism and racial discrimination around the world, such as racism against white farmers in Zimbabwe and discrimination against the Dalits—the “untouchables” in India.

Genocide and ethnic cleansing

At the extreme end of discrimination, we have genocide and ethnic cleansing . For example, around 1915 , the Ottomon (Turkish) empire slaughtered 1.5 million Armenians (75% of the Turkish Armenian population). The Turkish government took Armenian (largely Christian) children and converted them, giving them to Islamic families. Even today, Turkey defends this “Turkification” of Turkey as a necessary act of war and has resisted the U.S. and other nations defining it as genocide (Armenian genocide, n.d. ). Other genocides have occurred in Central Europe (the Holocaust) in the 1930s–1940s, Rwanda in 2003 , Cambodia in the 1970s, and the Greek/Pontic genocide of World War I. Extreme discrimination includes hate crimes and overt hate groups. The introduction of this chapter noted the prevalence of hate crimes and hate groups within the United States and other nations.

Redlining and racial profiling

In many countries, overt forms of discrimination for many (but seldom all) groups have been outlawed. Institutional discrimination itself may take forms that are harder to name and prove, such as redlining , the process by which banks give fewer mortgages to people of color, based on the belief that they are less able to repay loans. Some real estate agents may steer people of color away from rentals in upscale neighborhoods; school advisers may tell people of color that their children are more suited for trade school rather than college or graduate school. In the United States in 2014–2015 , there was a spate of cases surrounding potential police brutality against unarmed black men, leading to the “Black Lives Matter” movement. There is also racial profiling , such as when police pay more attention to people of color, stopping and/or searching them more frequently than they do whites (what some people of color call “DWB” or “driving while black”). A growing and complex array of academic studies examine whether or not profiling exists and, if so, what its nature is (e.g., is it pro-white, or does it depend on the race of the officer?). A similar phenomenon experienced by many people of color is being followed through stores by security guards, regardless of their attire or appearance. Notably, some aspects of discrimination, such as redlining, might be done, at least in the minds of the banker, real estate agent, or high school counselor, without a notion of racial discrimination; but here, Miles’s ( 1989 ) notion of racism defined by exclusionary outcome would classify the behaviors as racist, as they exclude based on supposed biological differences.

Intolerant Communication

Redneck racism/prejudice.

Central to our discussion is the way that discrimination and racism can occur through communicative behavior. Brislin ( 1991 ) outlined several forms of discriminatory communication. In addition to hate crimes and ethnic cleansing, he mentions redneck racism —the expression of blatant intolerance toward someone of another race. He applies these categories to racism, but we can apply them to any group. These might include jokes, statements (e.g., about the inferiority or backwardness of a group), or slurs or names for people of another group (also called ethnophaulisms ). Conventional wisdom, for example, suggests that there are many more slurs for women then there are for men, and most of these have some sexual connotation.

Sometimes, the intolerance is slightly veiled though still present, as when we resort to “us/them” language or talk to someone from another group about “your people.” Brislin’s ( 1991 ) notion of arm’s-length prejudice occurs when someone voices tolerance for a group, typically of being accepting of them in the neighborhood or workplace, but wants to restrict them from closer relationships, such as marrying a family member (related to Bogardus’s notion of social distance ; Allport, 1979 ). Prejudice might manifest in statements like “She’s very smart for an ‘X’” or “I have a friend who is a ‘Y,’ and he is very articulate,” since such statements assume that most Xs are not smart and most Ys are not articulate.

Prejudiced colloquialisms

Prejudice also manifests in our use of colloquialisms that play upon a particular aspect of identity or ability, such as calling something “lame” or “retarded.” Both the harm and use of such phrases has been established. For example, one study found that hearing the phrase “That’s so gay” made gays and lesbians feel less accepted in the university setting and, to a lesser degree, increased reported health problems. Over 45% of the participants had heard the word “gay” linked to something “stupid or undesirable” (Hall & LaFrance, 2012 , p. 430) ten or more times within the last year. Hall and LaFrance ( 2012 ) find a complex interplay between identity—males’ endorsement of gender identity norms andthe desire to distance themselves from homosexuality, as well as the social norms around them, and their likelihood to use the expression.

Prejudice built into language

We might well say that intolerance can be embedded in every level of language. In one classic study, men interrupted women much more than women interrupted men. If women overlapped men, men continued their turn speaking, but if men interrupted women, women yielded their turn speaking (Zimmerman & West, 1975 ). Coates’s ( 2003 ) analysis of narratives told by men in mixed company (such as around the family dinner table) notes that men are both the target and subject of most stories, with dinner table discussion typically centering on patriarchal authority. Research has explored prejudice through verbal and nonverbal behaviors toward people of different ages, people with disabilities, people with different languages or dialects, and other groups, including much theory and research on how we adjust or do not adjust our behavior toward those we perceive to be of different groups (communication accommodation theory; Gallois, Ogay, & Giles, 2005 ) or how minority members must negotiate their communication with dominant group members because of contexts of power and prejudice (co-cultural theory; Orbe & Spellers, 2005 ).

Bar-Tal ( 1990 ) and Zur ( 1991 ) note the way that we use rhetoric to create a sense of others (i.e., to create the identity of the enemy in a way that then justifies discrimination) resonates with Burke’s ( 1967 ) analysis of Hitler’s rhetorical construction of the Jewish people. Collins and Clément ( 2012 ), summarizing research from a special 2007 issue of Journal of Language and Social Psychology on language and discrimination up to the present, summarize the role of language as it pertains to prejudice:

Language is the primary means through which prejudice can be explicitly and implicitly communicated and is, therefore, a major contributor to its transmission and maintenance. But language can also play a more rooted and integral role in prejudice: changing perceptions by distorting the information it carries, focusing attention on social identities, and being a factor in the definition of group boundaries (p. 389).

Intolerance gone underground: Subtle forms of prejudice

As early as the mid-1980s, authors began to argue that in Western societies, racism and other forms of intolerance were going underground (i.e., aware that the redneck varieties of intolerance were socially unacceptable, people expressed less overt intolerance but continued to show intolerance through racism in ways that were “subtle” and “everyday”—a new and modern racism). People might express such forms of racism (and by extension other intolerances) through nonverbal behaviors, such as placing change on the counter instead of in an outgroup member’s hand, or through subtle sayings and word usages that exclude or put down the other person in some way that is not clearly distinguishable as prejudice. In the new racism, minority groups are not spoken of as inferior but as “different,” “although in many respects there are ‘deficiencies,’ such as single-parent families, drug abuse, lacking achievement values, and dependence on welfare and affirmative action—‘pathologies’ that need to be corrected” (van Dijk, 2000 , p. 34). Today, researchers and social activists refer to these subtle manifestations of prejudice as microagressions .

Symbolic racism is similar to subtle racism (Sears & Henry, 2005 ), though it relates more to political attitudes. Researchers have framed symbolic racism to include elements of anti-black sentiment hidden by political attitudes (e.g., that affirmative action has gone too far, that blacks are demanding too much; McConahay, 1986 ). Political research has a corollary in communication in that often, as whites talk about economic or political issues, there is at least a mental if not an explicit verbal coding of race or ethnic “othering.” International ownership of business becomes an issue when Japanese or Chinese companies start buying U.S. businesses, regardless of the large and long-term Dutch and English business holdings in America; discussions about welfare, gangs, and urban decay are often subtly about race. Similar verbal coding may also hold true with other identity groups.

Finally, in terms of face-to-face communication, researchers have explored the notion of “benevolent” intolerance. Discussions of things such as benevolent racism or sexism are often based on a larger notion of benevolent domination, whereby one nation or group seeks to dominate another, supposedly in its best interests (based on Rudyard Kipling’s notion of the “white man’s burden”). For example, Esposito and Romano ( 2014 ) contrast benevolent racism to other forms of post-U.S.-civil-rights forms of racism, such as laissez-faire racism, symbolic racism, and color-blind racism. Each might oppose affirmative action, for example, but for different reasons. Laissez-faire would oppose it based on ideas of meritocracy and free enterprise, blaming blacks themselves for lack of economic progress. Symbolic racism would hold that “the United States is a fair and equitable society where everyone has ample opportunity to succeed through hard work and talent” (p. 74), and that blacks who use the “race card” are hypersensitive—they are “too pushy, too demanding, too angry” (McConahay & Hough, 1976 , p. 38). Color-blind racism starts with what seems to be a reasonable assumption, that all people are the same, but then moves to assume that lack of progress of minority members is due to their personal choices, low work ethic, or lack of ability, and ignores structural support for inequalities.

Benevolent racism has a long history, even into slavery, a time in which some whites felt they were doing blacks a favor by controlling them and “providing” for them. More recently, it involves a seemingly positive attitude toward blacks that then opposes any social reforms like affirmative action as belittling blacks and working against their natural progress as citizens (Esposito & Romano, 2014 ). Benevolent sexism holds the same basic idea: Rather than sexism being based on anti-woman attitudes, it can also be supported by putting women “on a pedestal,” characterizing them as “pure creatures who ought to be protected, supported, and adored, and whose love is necessary to make a man complete” (Glick & Fiske, 2001 , p. 109). Extensive research has linked such benevolent ideas about women to negative outcomes for them.

Intolerance in the media and on the internet

Finally, many volumes have been written on the issues of stereotypes and intolerance in the media. This includes both social scientific work, such as the cultivation theory research that analyzes both representation of minorities in the media in different countries and the research that considers the effects of such representation. It also includes a wide array of critical and cultural analyses from the cultural studies school. Many of these analyses use the principles discussed—feminism, postcolonialism, critical race theory, whiteness, and so on. They work to demonstrate how the media systematically ignore, oversimplify, or negatively represent particular groups. One line of research in this field is the focus on the symbolic annihilation of race (Coleman & Chivers Yochim, 2008 ), which notes how, unlike stereotypes in the media that focus on the presence of some characteristic associated with a group, symbolic annihilation also considers “the meanings associated with absence, omission, or even inclusion that is not so obviously problematic (negative)” (p. 2), in terms of what such absences and seemingly benign images mean.

With the growth of the internet and video gaming, a final area of importance in understanding, researching, and working against prejudice includes all new media. The internet gives impetus for new research to understand hate groups on the media, flaming (e.g., in comments on video-hosting websites such as YouTube), and social media. We see examples of the use of social media for racist purposes in the flurry of racist twitters that followed the crowning of Nina Davuluri, an American of East Indian descent. Research considers both the presence of stereotypes in such media, as well as their effects.

The potential of communication

Unlike some early critical writers, who felt that media imagery (including new media) only produce and reproduce prevailing (prejudiced) ideologies, we must also consider the potential of face-to-face, mediated, and new media as places to challenge oppression. In terms of face-to-face communication, we can work through education to dispel stereotypes. That education can be simply on cultural differences and accomplishments, though changing cognitions alone may not change deeply felt affective prejudice, and only time (as more tolerant individuals assume positions of leadership) will lead to changes in discriminatory social structures. This is why some advocate for political education that addresses both personal and structural prejudice more directly, as well as political action and intervention in media systems.

Many scholars represent interpersonal contact as one of the best ways to address prejudice. Contact theory holds great potential for the planning of interventions to reduce intergroup tensions, as it describes how interpersonal contact with people from outgroups under the right conditions can work by changing both attitudes and affect, especially if people can see the other person as both a member of a new group while still recognizing their original group identity (Dovidio et al., 2003 ). Thomas Pettigrew ( 2016 ) outlines the history of research on authoritarianism (the desire and support for strong authority structures) and relative deprivation (the feeling that one’s group is disadvantaged in comparison to another group) as two of the main predictors of intergroup prejudice. He notes how, while personality factors like authoritarianism and cognitive rigidity are related to greater intolerance and make the likelihood of meaningful intergroup contact more unlikely, even in the presence of these variables, contact programs can have a positive effect for people with prejudice A meta-analysis of 515 contact studies suggests that contact works specifically by increasing knowledge of the other group, decreasing anxiety when one is with the other group, and increasing empathy for the other group (Pettigrew & Tropp, 2008 ).

In terms of media, we see both a growth in the production of media that challenges and resists stereotypes, rigid gender constructions, and so on, as well as a growth of grassroots efforts to highlight such oppression. One such effort is the website Fat, Ugly, or Slutty , a site composed of posts contributed by women who are stereotyped or verbally assaulted by men in video gaming websites, usually when the women have beaten them. The women are able to post comments made by other players, their own avatars, and even videos that the men sometimes send them. Efforts like these highlight forms of oppression that occur throughout the internet, but they also highlight the potential of the internet for addressing these forms of oppression in creative ways.

Conclusions

We have seen throughout this article that culture, prejudice, racism, and discrimination are related in complicated ways. Some people even see the characteristics of a particular culture (e.g., mainstream America’s conception of male and female beauty, the definition of a “good” education, or the focus on individualism) as negotiated between people with economic and power interests. Cultures (using the term much more widely than “nation”) are always ethnocentric, with individuals sometimes being xenophobic. But these forms of intolerance are frequently linked to other forms of intolerance—religious, racial, ethnic, and otherwise. Prejudice, most technically, is an affect—a desire to avoid someone because of her or his group, as opposed to stereotypes, which are more cognitive associations with a group—and efforts to reduce prejudice should focus on both affect and cognition. But intolerance is also clearly linked to higher-order manifestations of prejudice, such as discrimination through legal and organizational policies, symbolic annihilation of groups in the media, and everyday forms of discrimination, be they overt or subtle. More likely, communicative and policy forms of prejudice (and their manifest effects in terms of housing, education, job opportunities, and so on) “create” prejudicial perceptions, which in turn create the conditions of discrimination. Racism serves as an example—but only one of many—of the links among attitude, communicative action, policy, and social structure. With this complex view in mind, we can see that any attempts to redress or ameliorate racism or any other intolerance must include not only education, or even merely a wide array of communicative responses (media and face-to-face), but also efforts at addressing social inequalities at the structural and policy levels.

Further Reading

• Chin, J. L. (Ed.). (2010). The psychology of prejudice and discrimination . Santa Barbara, CA: Praeger.
• Crenshaw, K. W. , Gotanda, N. , Pellar, G. , & Thomas, K. (Eds.). (1995). Critical race theory: The key writings that formed the movement . New York: New Press.
• Delgado, R. , & Stefanic, J. (Eds.). (1997). Critical White studies: A look behind the mirror . Philadelphia: Temple University Press.
• Dovidio, J. F. (Ed.). (2010). The SAGE handbook of prejudice, stereotyping, and discrimination . London: SAGE.
• Fishbein, H. D. (2002). Peer prejudice and discrimination: The origins of prejudice . Mahwah, NJ: Lawrence Erlbaum.
• Giles, H. , & Watson, B. (Eds.). (2013). The social meanings of language, dialect, and accent: International perspectives on speech styles . New York: Peter Lang.
• Gioseffi, D. (Ed.). (1993). On prejudice: A global perspective . New York: Anchor.
• Goldberg, D. T. (1993). Racist culture: Philosophy and the politics of meaning . Oxford: Blackwell.
• Harwood, J. , & Giles, H. (Eds.). (2005). Intergroup communication: Multiple perspectives . New York: Peter Lang.
• Hecht, M. L. (Ed.). (1998). Communicating prejudice . Thousand Oaks, CA: SAGE.
• Nakayama, T. K. , & Martin, J. N. (Eds.). (1999). Whiteness: The communication of social identity . Thousand Oaks, CA: SAGE.
• Nelson, T. (Ed.). (2009). Handbook of prejudice, stereotyping, and discrimination . New York: Psychology Press.
• van Dijk, T. A. (1987). Communicating racism: Ethnic prejudice in thought and talk . Newbury Park, CA: SAGE.
• Whillock, R. K. , & Slayden, D. (Eds.). (1995). Hate speech . Thousand Oaks, CA: SAGE.
• Yūki, M. , & Brewer, M. (Eds.). (2013). Culture and group processes . Oxford: Oxford University Press.
• Adorno, T. W. Frenkel-Brunswik, E. , Levinson, D. J. , & Sanford R. N. , with B. Aron , M. H. Levinson , & W. Morrow (Eds.). (1950). The authoritarian personality . New York: Harper.
• Allport, G. (1979). The nature of prejudice . Reading, MA: Addison-Wesley.
• Althusser, L. (1971). Lenin philosophy and other essays B. Brewster (Trans.). London: NLB.
• Armenian genocide (n.d.). History .
• Auestad, L. (Ed.). (2013). Nationalism and the body politic : Psychoanalysis and the rise of ethnocentrism and xenophobia . London: Karnac Books.
• Baldwin, J. (1955). Stranger in the village . From Notes of a Native Son (pp. 159–75). (1984 [1955]). Boston: Beacon Press.
• Baldwin, J. R. (1998). Tolerance/intolerance: A historical and multi-disciplinary view of prejudice. In M. L. Hecht (Ed.), Communicating prejudice (pp. 24–56). Thousand Oaks, CA: SAGE.
• Baldwin, J. R. , Coleman, R. R. M. , González, A. , & Shenoy-Packer, S. (2014). Intercultural communication for everyday life . Malden, MA: Wiley-Blackwell.
• Baldwin, J. R. , Faulkner, S. L. , & Hecht, M. L. (2006). A moving target: The illusive definition of culture. In J. R. Baldwin , S. L. Faulkner , M. L. Hecht , & S. L. Lindsley (Eds.), Redefining culture: Perspectives across the disciplines (pp. 3–26). Mahwah, NJ: Lawrence Erlbaum.
• Baldwin, J. R. , Faulkner, S. L. , Hecht, M. L. , & Lindsley, S. L. (Eds.). (2006). Redefining culture: Perspectives across the disciplines . Mahwah, NJ: Lawrence Erlbaum.
• Baldwin, J. R. , & Hecht, M. L. (1995). The layered perspective of cultural (in)tolerance(s): The roots of a multidisciplinary approach. In R. Wiseman (Ed.), Intercultural communication theory (pp. 59–91). Thousand Oaks, CA: SAGE.
• Banton, M. (1987). Racial theories . Cambridge, U.K.: Cambridge University Press.
• Bar-Tal, D. (1990). Causes and consequences of deligimitization: Models of conflict and ethnocentrism. Journal of Social Issues , 46, 65–81.
• Bennett, M. J. (1993). Towards ethnorelativism: A developmental model of intercultural sensitivity. In R. M. Paige (Ed.), Education for the intercultural experience (pp. 21–71). Yarmouth, ME: Intercultural Press.
• Bhabha, H. (1994). The location of culture . London: Routledge.
• Bobo, L. (1983). Whites’ opposition to busing: Symbolic racism or realistic group conflict. Journal of Personality and Social Psychology , 45, 1196–1210.
• Brislin, R. W. (1991). Prejudice in intercultural communication. In L. A. Samovar & R. E. Porter (Eds.), Intercultural communication: A reader (6th ed., pp. 366–370). Belmont, CA: Wadsworth.
• Broome, B. J. (2005). Building bridges across the Green Line: A guide to intercultural communication in Cyprus . Nicosia, CY: United Nations Bicommunal Development Program.
• Burke, K. (1967). The philosophy of literary form: Studies in symbolic action (2d ed.). Baton Rouge: Louisiana State University Press.
• Chávez, K. R. (2013). Pushing boundaries: Queer intercultural communication . Journal of International and Intercultural Communication , 6, 83–95.
• Coates, J. (2003). Men talk . Malden, MA: Oxford University Press.
• Coleman, R. M. , & Chivers Yochim, E. (2008). The symbolic annihilation of race: A review of the “Blackness” literature. African American Research Perspectives , 12, 1–10.
• Collins, K. A. , & Clément, R. (2012). Language and prejudice: Direct and moderated effects . Journal of Language and Social Psychology , 31 , 376–396.
• Collins, P. H. (1990). Black feminist thought: Knowledge, consciousness, and the politics of empowerment . Boston: Unwin Hyman.
• Devine, P., G. , & Sharp, L. B. , (2009). Automaticity and control in stereotyping and prejudice. In T. Nelson (Ed.), Handbook of prejudice, stereotyping, and discrimination (pp. 61–87). New York: Psychology Press.
• Donald, J. , & Rattansi, A. (1992). Introduction. In J. Donald & A. Rattansi (Eds.), “ Race,” culture and difference (pp. 1–8). London: SAGE.
• Dovidio, J. F. , Gaertner, S. L. , & Kawakami, K. (2003). Intergroup contact: The past, present, and future. Group Processes & Intergroup Relations , 6 (1), 5–21.
• Dyer, R. (1997). White . New York: Routledge.
• Esposito, L. , & Romano, B. (2014). Benevolent racism: Upholding racial inequality in the name of Black empowerment. Western Journal of Black Studies , 38 (2), 69–83.
• FBI . (2014). Latest hate crime statistics report released . FBI.
• Foss, K. A. , & Foss, S. K. (1994). Personal experience as evidence in feminist scholarship. Western Journal of Communication , 58 , 39–43.
• Frankenberg, R. (1993). White women, race matters: The social construction of Whiteness . Minneapolis: University of Minnesota Press.
• Gallois, C. , Ogay, T. , & Giles, H. (2005). Communication accommodation theory. In W. B. Gudykunst (Ed.), Theorizing about intercultural communication (pp. 121–148). Thousand Oaks, CA: SAGE.
• Gerth, H. H. , & Mills, C. W. (Eds.). (1946). From Max Weber: Essays in sociology . New York: Oxford University Press.
• Glick, P. , & Fiske, S. (2001). An ambivalent alliance: Hostile and benevolent sexism as complementary justifications for gender inequality. American Psychologist , 56 , 109–118.
• Hall, J. , & LaFrance, B. (2012). “That’s gay”: Sexual prejudice, gender identity, norms, and homophobic communication . Communication Quarterly , 60 , 35–58.
• Hate and extremism . (n.d.). Southern Poverty Law Center.
• Hate map . (n.d.). Southern Poverty Law Center.
• Hecht, M. L. , & Baldwin, J. R. (1998). Layers and holograms: A new look at prejudice. M. L. Hecht (Ed.), Communicating prejudice (pp. 57–84). Thousand Oaks, CA: SAGE.
• Herbst, P. H. (1997). The color of words: An encyclopaedic dictionary of ethnic bias in the United States . Yarmouth, ME: Intercultural Press.
• Knowles, C. , & Mercer, S. (1992). Feminism and anti-racism: An exploration of the possibilities. In J. Donald & A. Rattansi (Eds.), “Race,” culture, and difference (pp. 104–125). London: SAGE.
• Kroeber, A. L. , & Kluckhohn, C. (1952). Culture: A critical view of concepts and definitions . Cambridge, MA: Harvard University Press.
• Leeds-Hurwitz, W. (1990). Notes in the history of intercultural communication: The Foreign Service Institute and the mandate for intercultural training. Quarterly Journal of Speech , 76 , 262–281.
• Malott, K. M. , & Schaefle, S. (2015). Addressing clients’ experiences of racism: A model for clinical practice . Journal of Counseling and Development , 93 , 361–369.
• McConahay, J. B. (1986). Modern racism, ambivalence, and the modern racism scale. In J. S. Dovidio & S. I. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 99–125). Orlando, FL: Academic Press.
• McConahay, J. B. , & Hough, J. C. (1976). Symbolic racism. Journal of Social Issues , 32 (2), 23–45.
• Meyer, P. (1987). Ethnocentrism in human social behavior: Some biosociological considerations. In V. Reynolds , V. Falger , & I. Vine (Eds.), The sociobiology of ethnocentrism: Evolutionary dimensions of xenophobia, discrimination, racism, and nationalism (pp. 81–93). London: Croom Helms.
• Miles, R. (1989). Racism . London: Routledge.
• Moon, D. G. (2002). Thinking about “culture” in intercultural communication. In J. N. Martin , T. K. Nakayama , & L. A. Flores (Eds.), Readings in intercultural communication: Experiences and contexts (2d ed., pp. 13–21). Boston: McGraw Hill.
• Nakayama, T. K. (1998). Communication of heterosexism. In M. L. Hecht (Ed.), Communicating prejudice (pp. 112–121). Thousand Oaks, CA: SAGE.
• Nakayama, T. K. , & Halualani, R. T. (Eds.). (2010). The handbook of critical intercultural communication . Malden, MA: Wiley-Blackwell.
• Neuliep, J. (2012). The relationship among intercultural communication apprehension, ethnocentrism, uncertainty reduction, and communication satisfaction during initial intercultural interaction: An extension of anxiety and uncertainty management (AUM) theory . Journal of Intercultural Communication Research , 41 , 1–16.
• Omi, M. , & Winant, H. (1986). Racial formation in the United States: From the 1960s to the 1980s . New York: Routledge.
• Opotow, S. (1990). Moral exclusion and injustice: An introduction. Journal of Social Issues , 46 , 1–20.
• Orbe, M. P. , & Spellers, R. E. (2005). From the margins to the center: Utilizing co-cultural theory in diverse contexts. In W. B. Gudykunst (Ed.), Theorizing about intercultural communication (pp. 173–191). Thousand Oaks, CA: SAGE.
• Pettigrew, T. F. (2016). In pursuit of three theories: Authoritarianism, relative deprivation, and intergroup contact . Annual Review of Psychology , 67 , 1–21.
• Pettigrew, T. F. , & Tropp, L. R. (2008). How does intergroup contact reduce prejudice? Meta-analytic tests of three mediators . European Journal of Social Psychology , 38 , 922–934.
• Rattansi, A. (1992). Changing the subject? Racism, culture, and education. In J. Donald & A. Rattansi (Eds.). “Race,” culture, and difference (pp. 11–48). London: SAGE.
• Rex, J. , & Mason, D. (Eds.). (1986). Theories of race and ethnic relations . Cambridge, U.K.: Cambridge University Press.
• Reynolds, V. , Falger, V. , & Vine, I. (Eds.). (1987). The sociobiology of ethnocentrism: Evolutionary dimensions of xenophobia, discrimination, and nationalism . London: Croom Helm.
• Rogers, E. M. , & Hart, W. B. (2002). The histories of intercultural, international, and development communication. In W. B. Gudykunst & B. Mody (Eds.), Handbook of international and intercultural communication (3d ed., pp. 1–18). Thousand Oaks, CA: SAGE.
• Rosenau, P. M. (1992). Post-modernism and the social sciences: Insights, inroads, and intrusions . Princeton, NJ: Princeton University Press.
• Ross, M. H. (1991). The role of evolution in ethnocentric conflict and its management. Journal of Social Issues , 47 , 167–185.
• Ruscher, J. B. (2001). Prejudiced communication: A social psychological perspective . New York: Guilford.
• Said, E. W. (2003). Orientalism . New York: Vintage Books.
• Sears, D. O. , & Henry, P. J. (2005). Over thirty years later: A contemporary look at symbolic racism. Advances in Experimental Social Psychology , 37 , 95–150.
• Shah, Anup . (Oct 1, 2010). Human rights in various regions . Global issues: Social, political, economic and environmental issues that affect us all.
• Shome, R. , & Hegde, R. (2002). Postcolonial approaches to communication: Charting the terrain, engaging the intersections. Communication Theory , 12 , 249–270.
• Slavíčková, T. , & Zvagulis, P. (2014). Monitoring anti-minority rhetoric in the Czech print media: A critical discourse analysis . Journal of Language and Politics , 13 (1), 152–170.
• Stephan, W. (1985). Intergroup relations. In G. Lindzey & E. Aronson (Eds.), Handbook of social psychology (rev. ed.) (pp. 599–658). Reading, MA: Addison-Wesley.
• Sumner, G. A. (1975). Folkways and mores . E. Sagarin (Ed.). New York: Shocken.
• Tajfel, H. , & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel & W. G. Austin (Eds.), Psychology of intergroup relationships (2d ed.) (pp. 7–24). Chicago: Nelson-Hall.
• Ura, M. , Preston, K. S. J. , & Mearns, J. (2015). A measure of prejudice against accented English (MPAAE) . Journal of Language & Social Psychology , 34 , 539–563.
• van Dijk, T. 2000. New(s) racism: A discourse analytical approach. In S. Cottle (Ed.), Ethnic minorities and the media (pp. 33–49). Milton Keynes, UK: Open University Press.
• van Dijk, T. A. (1993). Elite discourse and racism . Newbury Park, CA: SAGE.
• Wander, P. C. , Martin, J. N. , & Nakayama, T. K. (1999). Whiteness and beyond: Sociohistorical foundations of Whiteness and contemporary challenges. In T. K. Nakayama & J. N. Martin (Eds.), Whiteness: The communication of social identity (pp. 13–26). Thousand Oaks, CA: SAGE.
• Warner, M. (1991). Introduction: Fear of a queer plant. Social Text , 29 , 3–17.
• West, C. , & Zimmerman, D. H. (1987). Doing gender. Gender & Society , 1 (2), 125–151.
• Wilson, W. J. (1978). The declining significance of race: Blacks and changing American institutions . Chicago: University of Chicago Press.
• Wood, J. (2008). Critical feminist theories: Giving voice and visibility to women’s experiences in interpersonal communication. In L. A. Baxter & D. O. Braithwaite (Eds.), Engaging theories in interpersonal communication: Multiple perspectives (pp. 323–334).Thousand Oaks, CA: SAGE.
• Yep, G. A. (2013). Queering/quaring/kauering/crippin’/transing “other bodies” in intercultural communication . Journal of International and Intercultural Communication 6 , 118–126.
• Zimmerman, D. H. , & West, C. (1975). Sex Roles, interruptions, and silences in conversation. In B. Thorne & N. Henley (Eds.), Language and sex: Difference and dominance (pp. 105–129). Rowley, MA: Newbury House.
• Zompetti, J. P. (2012). The cultural and communicative dynamics of capital: Gramsci and the impetus for social action. Culture, Theory and Critique , 53 , 365–382.
• Zur, O. (1991). The love of hating: The psychology of enmity. History of European Ideas , 13 , 345–369.

Related Articles

• Race, Nationalism, and Transnationalism
• Cultural Identities
• Prejudiced Communication
• Behavioral Indicators of Discrimination in Social Interactions
• Critical Whiteness Studies
• Culture, Ethics, and Communication
• Diffusion of Concepts of Masculinity and Femininity
• Performance of Race, Culture, and Whiteness
• Anxiety, Uncertainty, and Intercultural Communication

Printed from Oxford Research Encyclopedias, Communication. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 30 August 2024

• Cookie Policy
• Privacy Policy
• Legal Notice
• Accessibility
• [185.147.128.134]
• 185.147.128.134

Character limit 500 /500

• Architecture and Design
• Asian and Pacific Studies
• Business and Economics
• Classical and Ancient Near Eastern Studies
• Computer Sciences
• Cultural Studies
• Engineering
• General Interest
• Geosciences
• Industrial Chemistry
• Islamic and Middle Eastern Studies
• Jewish Studies
• Library and Information Science, Book Studies
• Life Sciences
• Linguistics and Semiotics
• Literary Studies
• Materials Sciences
• Mathematics
• Social Sciences
• Sports and Recreation
• Theology and Religion
• Publish your article
• The role of authors
• Promoting your article
• Abstracting & indexing
• Publishing Ethics
• Why publish with De Gruyter
• How to publish with De Gruyter
• Our book series
• Our subject areas
• Your digital product at De Gruyter
• Contribute to our reference works
• Product information
• Tools & resources
• Product Information
• Promotional Materials
• Orders and Inquiries
• FAQ for Library Suppliers and Book Sellers
• Repository Policy
• Free access policy
• Open Access agreements
• Database portals
• For Authors
• Customer service
• People + Culture
• Journal Management
• How to join us
• Working at De Gruyter
• Mission & Vision
• De Gruyter Foundation
• De Gruyter Ebound
• Our Responsibility
• Partner publishers

Your purchase has been completed. Your documents are now available to view.

How to approach ‘prejudice’ and ‘stereotypes’ qualitatively: The search for a meaningful way

• Magda Petrjánošová

This paper is partly a theoretical and analytical exploration of different ways to do research about stereotypes and prejudice, and partly a confessional tale of my journey. It is a journey that has been about looking for a meaningful and useful way of approaching empirical material collected in different research projects over more than 15 years, in an attempt to say something about how ordinary social actors talk (and possibly think) about prejudice and stereotypes. There is an immense volume of social psychological writing on this topic, and from that I discuss in detail several new(ish) discursive, critical and constructional approaches and the (im)possibility of applying them to my empirical material.

Introduction

Over time, this article has evolved from a short, simple conference presentation focusing on a few interesting extracts concerning prejudice and stereotypes. When I began writing the article in “discussion” with other (cited) authors, looking for potential new ways of analysing the same extracts, it got far more complicated than I had anticipated. It now takes the form of part theoretical and analytical paper about stereotypes and prejudice, and part confessional tale ( Van Maanen, 1988 ) of my journey. It is a journey that has been about looking for a meaningful and useful approach to analysing empirical material I have collected in different research projects over more than 15 years, in an attempt to say something about prejudice and stereotypes.

The study of prejudice became central to social psychology with the work of Allport (1954) . Allport, who focused mainly on negative ethnic prejudice, defined it as “an antipathy based on faulty and inflexible generalization. It may be felt or expressed. It may be directed toward a group or an individual of that group” ( Allport, 1954 , p. 9). As Billig (2012) notes, prejudice was initially defined more broadly but at the beginning of the 20 th century it narrowed to refer to negative opinions and to focus on categories of ethnicity and race [2] . Today (in mainstream social psychology) this concept can mainly be found in the triad of prejudice, stereotypes/stereotyping and discrimination; prejudice is usually defined more specifically as a complex attitude to a specific group, stereotypes/stereotyping as the attribution of specific characteristics to this group and discrimination as a non-neutral behaviour towards this group and its members (e.g. Dovidio, Hewstone, Glick, & Esses, 2010 ). Alternatively, this triad is understood slightly differently in the three-part model of attitudes (sometimes called the ABC model), in which stereotypes are seen as the cognitive aspect, prejudice as the affective aspect and discrimination as the behavioural aspect of attitudes towards a group ( Fiske, 1998 ). Moreover, in recent writing numerous more specific concepts have appeared that have a more or less clear connection to the concept of prejudice and that can be used as dependent measures which tap into the ABC model. These include social distance, intergroup trust, perceived threat, and so forth (for an incomplete but rather voluminous and recent overview, see Lášticová & Findor, 2016 ).

Nonetheless there is immense variation in the social psychological work done on prejudice. As Condor and Figgou (2012) summarize more generally, in social psychology, prejudice has been studied as a matter of “instinct, drive, motivation, emotion, categorization, social identity, attribution, personality, executive control or rhetoric” (p. 202). There have also been many different opinions on the reasons for prejudice and the mechanisms by which it comes about. Dovidio (2001) describes how these have changed over time—the early works saw prejudice as a personality fault, later it was viewed as imperfect information processing and recently more and more researchers have focused on unconscious and automatic prejudices.

Again, when we look at how prejudice has been researched, there is great diversity. Very different methods were used at different times and by researchers working within different paradigms. Prejudice has been researched using qualitative (e.g. interviews, focus groups), quantitative (e.g. questionnaires, social distance scales), as well as experimental (e.g. pupil dilation, response latency) methods. The predominant methods, though, are perhaps those that are most easy to use: direct self-report questionnaires (for an excellent overview see Fiske & North, 2014 ). Usually those researched give their views (personal or for their whole group [3] ) on different groups defined for example according to race/ethnicity, age, gender, sexual orientation or other criteria, and researchers measure the attitudes of one group towards the other while working with different components according to the theory they subscribe to—competence, trust, warmth, and so on.

The role of the researchers is omnipotent here—they get to say what is (and is not) prejudice and whether and to what extent someone is prejudiced (or not); in short the researchers are the “standard setters” of truth (Kruglanski, 1989 in Dixon & Levine, 2012a , p. 306). As Durrheim and colleagues argue (2016) , social psychologists have “generally sought to develop authoritative definitions and measures of prejudice” (p. 18) and these from top down definitions “have been superimposed on ordinary people’s attitudes in order to identify prejudiced individuals” (p. 18). Moreover, there is a problem with these definitions. As Condor and Figgou (2012) state, they are not very precise and have not exhibited a “high level of consistency over the past century” (p. 201). Despite social scientists’ claims that they have 1) been making the definition more and more accurate over time in line with scientific progress, and 2) that they are much more specific and precise than ordinary social actors—lay men and women—in fact their definitions are not so very different. However, as Durrheim, Quayle and Dixon (2016) state, in everyday communication what counts or does not count as prejudice is context dependent in the given situation and sometimes even fought over “with considerable passion and no little skill” (p.18), mainly because casting the same idea as prejudice or as a rational and legitimate attitude has very different consequences in the real world. Because of all the real-life complications with a clear-cut definition of prejudice, and because of other more generally critical voices (e.g. Whetherell, 2012), I have placed the terms “prejudice” and “stereotypes” in quotation marks in the title and also in important places in the article as a reminder that these are just labels.

Inspired by these insights and following my old suspicion of measurement tools, in this article, I am not interested in measuring the extent to which someone (a person or group) could be considered prejudiced, or how that changes after such and such an intervention [4] . Neither am I interested in the content, what exactly the potentially prejudiced opinion is about or which characteristics are ascribed to the group of people in question. What I want to look at are the opinions of the participants, but I plan to approach prejudice from a meta position, so as to better understand what the participants say and possibly think ABOUT prejudice, rather than what their explicit definitions of it might be. Thus in this article I shall attempt an approach to stereotypes and prejudice where I am interested in how the research participants THEMSELVES refer to the existence and validity of prejudice/stereotypes (in themselves and others), what they think about how the stereotypes are shared within their own and other groups, how they personally (dis)agree with them, how carefully they express an opinion that could be socially unacceptable, how they work around this complication, and so on. All this is possible only in contexts where the participants can articulate their opinions (not for example in closed questions in a questionnaire). Thus the methods of empirical material collection already determine which approaches can and cannot be used with the material. Here I am using extracts from interviews, focus groups and (open limit) answers to an open question in a questionnaire. Another possibility would be to use statements that did not originate in a research setting, but that are “natural” [5] —like newspapers articles or political speeches (see e.g. Wodak & Meyer, 2001 ).

I would like to make two small points before I present the empirical material collected and the analytical perspective: First, as mentioned above, implicit definitions of prejudice do not appear to differ so greatly in lay discourses and in scientific writing. However, in my experience, if ordinary social actors use (and often they do not) an explicit term when referring to something they consider prejudiced/one-sided/stereotypical, they tend not to distinguish between the concepts of stereotypes and prejudice but use both terms synonymously, or they sometimes use “prejudice” in reference to negative opinions only, and “stereotypes” for both positive and negative opinions. That is why I use both terms in this article, and, of course, I use the term the participants use in each extract from the empirical material. Second, in this article I deal only with nationally defined groups, often defined on the basis of state citizenship because this was the perspective from which our earlier research projects were conducted. This is despite my agreeing that nationalities are labels and that in real life issues around membership in nationally defined groups is often complicated, complex and not at all clear-cut (see also methodological nationalism , Wimmer & Glick Schiller, 2002 ).

Empirical material and analytical perspective

In 2003–2004 we used semi-structured interviews and commented drawings of the borderland in an Austrian–Slovak project about young adults from the borderland and their perceptions of their own nation and the other nation (32 Slovaks, 32 Austrians, aged 16–24, selected using quota sampling taking into account age, gender, education, size of dwelling and (not) having a better experience of the other nation) (see Spannring et al., 2005 );

in 2005–2007 in a project on the lives, attitudes and feelings of home of Slovaks travelling regularly (mostly daily) to Austria for study or work we used focus group discussions, semi-structured interviews and commented drawings of the borderland (26 Slovaks, aged 24–46, selected using quota sampling taking into account age, gender, education and size of dwelling. We then looked for and added “contrasting cases”) (see Lášticová & Petrjánošová, 2014 );

in 2009–2010 in a project on the everyday lives of a “community” of Slovak short-term migrants to Ireland we used semi-structured interviews with 8 (male and female) Slovaks who had different leadership roles in the Slovak group in Ireland (see Lášticová & Petrjánošová, 2013 );

in 2010–2012 in a research project on intergroup attitudes in central Europe we analysed answers to an open question about experiences of the neighbouring nation. The respondents were Czechs and all their neighbours—Austrians, Germans, Poles and Slovaks (1,260 female and male university students from the respective borderlands) (see Graf, Hřebíčková, Petrjánošová, & Leix, 2015 ).

The theoretical perspective I adopt in this article was inspired and influenced mainly by the traditions of discursive analysis (e.g. Condor, 2011 ), critical discursive analysis (e.g. Wodak & Krzyzanowski, 2008 ; Van Dijk, 1984 ) and rhetorical approach (e.g. Billig, 2012 ). Much of the texts, especially the older ones, deal with the tensions around expressing prejudice in a world and era where there is a broadly shared prejudice against prejudice or a “general cultural norm against ‘prejudice’” ( Billig, 2012 , p. 141) and a tendency to consider prejudiced attitudes to be objectionable on ethical grounds and irrational in nature. [6] Researchers have focused on the micro-level of strategic self-monitoring, self-presentation (relying on Goffman’s work) or more generally on identity management issues in expressing prejudice but avoiding the stigma of being evaluated as prejudiced ( Condor, 2000 ; Augustinos & Every, 2007 ) or even more generally on “careful negotiation and identity construction around the topic of prejudice” ( Wetherell, 2012 , p. 168). For example Van Dijk explored specific semantic, pragmatic and conversation strategies of “adequate self-expression, positive self-presentation and effective persuasion” (1984, p. 116) when formulating “ethnic opinions” (p.116). Recently, the writings have focused even more on the “social” side—not on the verbal acts of individual actors but rather on how prejudicing and stereotyping happens as a result of the joint discursive action of several speakers (e.g. Condor & Figgou, 2012 ) or even of speakers and hearers (as a result of implicit allusions on one side and understanding of contextual information and shared categorical associations, e.g. Durrheim et al., 2016 ). [7]

Thus, when looking now (in 2018) at the older materials, I am mainly interested in how people speak ABOUT “stereotypes” and “prejudice”, and for instance how they assess their accuracy and whether they admit to expressing views that can be considered prejudiced. In connection with the last point I assume that there are many more negative stereotypes around, which did not come up in the research projects, because the participants did not want to “admit” to them, meaning they did not consider it socially desirable to share them with us, the researchers, in that interaction. [8]

In the following part I will show, using specific extracts from the empirical material, examples of different approaches to stereotypes/prejudice, the way the research participants talk/write and possibly think about them. From the wide range of aspects that could be focused on, I shall look in this article, in the following order, at prejudice as a source of knowledge; stereotypes declared as shared within the ingroup and (dis)agreement with them; different discursive ways of dealing with personal experience that contradicts a shared stereotype; and at a declared change in a stereotypical attitude following personal experience.

Prejudice as a source of knowledge about the other group

In the following extract from a semi-structured interview with a young Austrian man from the Slovak–Austrian borderland we can see how he tries to meaningfully answer a question on the differences between his national group and the neighbouring national group. The interview took place in 2003, shortly before Slovak accession to the European Union—a small number of people from Slovakia had been studying in Austria, and a larger number went shopping there or on trips, but the border controls still existed and officially it was impossible for Slovaks to work in Austria. In general there was much less contact in the wider borderlands, including the capitals Vienna and Bratislava, than there is today, in 2018.

Extract 1: I don’t know of any prejudice about Slovaks

Interviewer: and what are the differences between slovaks and austrians.

Participant: (...) perhaps wealth. And if anything, then the lifestyle, they do not have it so far, but are on the way. But I don’t know of any prejudice about Slovaks from which I could infer the differences between them and Austrians [9] (answer in a semi-structured interview, 2003, male participant from Austria, age 25, completed upper secondary school, Vienna)

Unfortunately, the interviewer did not press the participant to explain what exactly he meant by “prejudice” when he used the concept in this rather unusual context. More detailed questioning might have revealed more information on his “ethnotheory of prejudice”, on what he thinks about the accuracy of such information or if and how he goes about verifying it. Here, we ultimately lose the advantage obtained by the researcher not asking directly about prejudice but some participants spontaneously thematizing it, which could hopefully have made the answer less socially desirable than if the question had been ‘are you prejudiced against your eastern neighbours?’ This is an extract from a semi-structured interview covering several topics, and so the researcher moves on to the next topic and we learn nothing more than this interesting fragment of information. We could extrapolate that the participant does not know a lot about Slovaks, perhaps because of the socio-political context mentioned above, that perhaps they do not interest him (as he lives in a higher status state), and that he has almost certainly never visited Slovakia, but that nonetheless he still (unsurprisingly) tries to answer the question. Only very seldom do participants tell us straight away that they do not know. In his answer this young Austrian man begins by mentioning economic differences (the most frequently mentioned specific difference from the Slovak participants, too, see Spannring et al., 2005 ). His reference to ‘lifestyle’ probably means standard of living and without giving the adjective, he means a high, higher or perhaps an average European standard of living, seen from the Austrian side of the border. Then, quite quickly he states explicitly that he does not have enough information about Slovaks to answer the question and suggests the reason is he does not know of any prejudice about Slovaks. This could be viewed as supporting the idea that stereotyping and prejudice are innocent attempts at categorizing the unknown. The problem is that it is not innocent when the prejudice is explicitly negative and when it is the cause of people from different groups having no contact with one another and remaining permanently unknown to each other. It is also interesting to see how he (probably lacking the personal experience) automatically looks for any socially shared information about the other group accessible to him, even if it is just hearsay. Not having any information he logically cannot take the next possible step of testing/questioning the accuracy of such information, in contrast to the participants cited in the next part.

Stereotypes declared within the in-group and (dis)agreement with them

In this part I will look at the stereotypes the participants refer to as known or more or less broadly shared within their own group (in-group). It will then be interesting to see whether they declare an agreement or disagreement with them and the reasons they give.

I would like to add a more general note here—if the empirical material collected allows for a comparative perspective, it is useful to look at how the two sides (two groups, for example members of two nations) see each other and what are the differences. In the fourth cited research project involving Austrians, Czechs, Germans, Poles and Slovaks we had this opportunity and found that interesting asymmetries emerged, including in relation to how many stereotypes each group mentioned in reference to the other groups or what the ratio was of positive to negative stereotypes about the same group. If one group (nationally defined, for example) has more stereotypes about the other group than vice versa, this could be interpreted as indicating the second group is particularly interesting to the first group for some, historical, economical or other, reasons. A big difference of this kind was noticeable in relation to Czechs and Germans for instance, with the first group reporting many more stereotypes about Germans than the Germans did about Czechs. [10] Moreover, where it was possible to guess the emotional valence [11] , clear differences emerged in the ratio of “negative” to “positive” stereotypes. For example, all the Austrian stereotypes mentioned in relation to Czechs were negative, but of the reported Czech stereotypes regarding Austrians half were negative and half positive. As mentioned above, I was not interested in statements like The Czechs are close-fisted, which is a hypothetical statement that could have been evaluated as prejudiced from the researcher’s position. Rather I focused on explicit references to the existence of stereotypes/prejudice like Here they say, that Czechs are close-fisted which is another hypothetical statement where it would be interesting to see for example whether and how the speaker maintains the constructed distances from the stereotype (because it is not ´we´ who is saying it, but it is ´they´) in her/his next sentences.

The next extract introduces the theme of agreeing or disagreeing with a stereotype that seems to be broadly shared within the in-group. Sometimes the participants reported agreeing with such a stereotype or that they had experienced it being validated. More often they mentioned such cases when the stereotype was contested, possibly because in these instances it is easier to recognize that stereotypes shape our thinking.

Extract 2: In contrast to what we say here

(...) In contrast to what we say here about Germans, these two girls were much more spontaneous and friendly than me at the time (answer to a single open question “what is your experience of Germans?” in a questionnaire on intergroup contact and attitudes comprised of closed questions except for this one, 2010, Czech statement about Germans, female participant, age unknown, statement no. 1432).

The participant, speaking about a student exchange some time ago during secondary school, does not explicitly say what they “ say here about Germans ”. Again, if it was in an interview, at least it would have been possible for a vigilant interviewer to ask for more details about what “they” say and who “they” are, and whether the speaker thought so before, too. But we do not have this fuller answer and can only infer—for example, from the context of the positively coloured statement about receiving a friendly welcome while on the exchange, we could assume that this thing that is generally said about the Germans is quite the opposite of the speaker’s experience of the two spontaneous and friendly young German girls. This situation repeated itself several times, and always when the participants did not specify the stereotype or prejudice referred to, from the context it was clear that they were negative.

Different discursive ways of dealing with a personal experience that contradicts a stereotype broadly shared within the in-group

Where personal experience did not confirm a stereotype reported to be shared within the in-group, participants used different discursive strategies to deal with this in a meaningful and logical way. Sometimes they just reflected on the difference, as was the case with the statement in extract 2. In some cases they declared an exception to the rule—someone from another national group who did not act in accordance with the stereotype was declared to be an exception, but the stereotype remained uncontested. [12] Sometimes there were so many exceptions that whole exceptional subgroups of the big national group were declared. An example could be (a fictional) statement like The Austrians who are my friends are ok, but in general it is true that as a nation they are all big-headed. These subgroups could be defined according to knowledge of the person, as in the example, but also according to region of origin, age, gender and so on. Only in a few cases from all the material collected did an “antistereotypical” personal experience lead to an (at least declared) change of opinion or abandonment of the stereotype.

Declared change in stereotypical belief

Where there was a (declared) abandoning of a stereotypical opinion, it was often narrated as a story progressing over time in stages: stereotypical information—personal experience—change of opinion/abandonment of stereotype, and this makes the change of opinion sound reasonable and logical.

Extract 3: Ireland is beautiful.

(…) I have heard that the Irish just drink and take drugs and that Ireland is ugly and it’s always raining, but I came here and they are friendly and Ireland is beautiful.(…) (extract from an interview about experiences in Ireland and the existence of a Slovak “community” there, speaking about the decision to go abroad, 2009, AZ, age 29, male participant, short-term Slovak migrant in Ireland)

The personal experiences required for such a change were often not one-off, but repeated and/or long-term. [13] Personal experience of the members of another national group does not always improve relationships and lead to the stereotypes being abandoned (see also Allport’s famous conditions for positive inter-group contact influence, in Allport, 1954 ). In the empirical material there were several cases mentioned where this reportedly did not work (cf. Paolini et al., 2010 ). For example, in one reported story, following personal contact among Czech and German secondary school students that did not go well a new negative stereotype was created (about what Czech secondary school students are like) and the whole exchange program was stopped.

Concluding remarks

I think stereotypes and prejudice are both a fascinating research issue and a topic with real everyday consequences for all of us. Given my vague suspicion of measurement tools such as direct self-report questionnaires, [14] I felt enlightened and inspired when I discovered the work of several scholars that can be mainly grouped under the discursive and critical approaches to stereotypes/prejudice, [15] who were not interested in how many people in group A would tick negative categorical evaluations of the members of group B. They had found so much more to investigate and problematize!

In this article I wanted to apply what I saw them doing with their extracts to the empirical material we had collected over many years and from many projects. Using a qualitative analysis inspired by the discursive approaches allowed me to observe how participants explicitly talk (and possibly think) about stereotypes—for example, how often and how exactly they mention them, assess their accuracy, (dis)agree with them, explain changes in their own originally stereotypical opinions, explain logically two contradictory assessments of members of the same national group in one short statement. However, I did not look at the most “classic” tension points, at the way speakers mitigate or manage expressions that could be judged as prejudiced in order to avoid being judged as prejudiced themselves. I was more interested in what I could learn about stereotypes/prejudice from the viewpoint of the participants, so in this article I have not used extracts containing prejudice/stereotypes (that I the researcher would have to evaluate as such) but ones ABOUT prejudice/stereotypes. That coincides with the claim of Condor and colleagues (2012) that the research on lay understandings of prejudice is surprisingly sparse, and with Billig’s recommendation (2012) that the research should include what ordinary people understand by “prejudice”, given that the concept is so important in lay discourse.

However, I kept to the individual level, just as the majority of measuring approaches do. Condor and Figgou (2012) criticize methodological individualism, [16] as the main tendency among all the different approaches to prejudice and suggest that an alternative could be to think of prejudice in terms of collaborative cognition. In this approach groups or networks, not individuals, are the units of analysis. They show the construction, expression and suppression of subtle or blatant prejudice in a different light: first, they show the workings of so called “social scaffolding”—the way a more skilled person helps a less skilled person, instructing him/her in and facilitating the production of a logical (in this case racist) narrative. Second, they focus on how the pejorative portrayal of Others can be the result of joint action, where the contributions of each person are contextually important to the contributions of other participants, in this case allowing escalation in the expression of negative opinions. Third, they provide examples of joint inhibition, where one individual relies on others instead of self-monitoring and regulating his own expression of prejudice. Thus the display of prejudice is regulated through the interaction of several people—either through correcting the use of prejudiced categories or by openly admonishing the prejudicial talk of some of them— and not in the individual’s mind.

Moreover, Durrheim (2012) , when writing about implicit prejudice in interaction, demonstrates how “stereotypes are formulated in the context of social interaction and that they can take an implicit form in which the hearer must help to stereotype” (p.190). In the same spirit, Durrheim and colleagues (2016) present an identity performance model of prejudice that focuses attention not only on how the expression of prejudice is responsive to norms and audiences but also how it shapes those norms. They also show how contestation of the very definition of what can and cannot count as prejudice, can be used either to mobilize hatred against out-groups (if their negative opinion of them is presented not as prejudiced but as reasonable), and to cement or change identities and norms. Regarding the latter they give an interesting example of paedophile people attempting to cast themselves as a “minor-attracted sexual orientation group” and the negative attitudes towards them as prejudice, which would result in very tangible real-life changes, in law for example. Further they illustrate how accusations and denials of prejudice “help to preserve categories, meanings and boundaries” (p. 26) and how repression of prejudice “can be viewed as a collaborative identity performance” (p. 29) in which all participants avoid the potential shame associated with uttering or hearing prejudice. Thus denials as well as accusations often remain inexplicit, but still the “work of reproducing the racial order” (2016, p. 29) has been done. Their article is a persuasive plea for a new agenda in social research that would attempt to “understand how the very definition of ‘prejudice’ is jointly defined and negotiated and deployed in social interactions to achieve social and political outcomes” (p. 32).

I consider these recent constructionist and critical psychological approaches to be highly inspirational but realize they cannot be applied to my empirical material—which I had at first hoped to do in order to obtain a kind of higher level analysis. Of course, having the empirical material that would allow for qualitative analysis (e.g. interview transcripts) is here not enough. To be able to “shake-off” the individual focus and to pay attention to the social or interactional, I would need “interactional” material—transcripts of conversations, for example. Ideally if I am to approach real-life situations these should not be elicited conversations (at the researcher’s request or in answer to a direct question) but either “natural” ones (see my explanation above) or ones that do not at least primarily focus on the issue interesting to the researchers (cf. Condor & Figgou, 2012 ). This last approach is exactly what I was trying to apply in this article when I began looking at older empirical material with a new topic and new perspective in mind.

To be more precise and honest, I wanted to put together material from over a long period and from several research projects, look at it with a new focus and then show how it could be analysed on several levels, inspired by the “classic” and more recent discursive analytical writing. Then I realized just how far the newest approaches have moved on and how inadequate my empirical material is for that.

I still think that the original idea of looking at the topic of stereotypes/prejudice using empirical material in which participants mention it spontaneously and not when prompted is a good one. But as became immediately clear, this does not work if there was not opportunity for letting them elaborate on the issue once touched upon. There are several reasons for this—the impossibility of asking further more detailed questions in the case of open questions in a questionnaire; interviews having a different focus at the time they were conducted and researchers wanting to cover too many topics in a single encounter in the case of interviews, and I would now say even the inability to explore the unexpected “jewel” of new and interesting information.

Thus in this article I was only able to go as far as the collected empirical material allowed, but exploring these new approaches has given me some new ideas for research that will be more difficult to realize but that will hopefully prove more helpful in the struggle to understand prejudice/stereotypes in the social reality of everyday life.

1 This work was supported by the Slovak Research and Development Agency contract No. APVV-14-0531; however, the empirical material it is based on was collected as part of several earlier research projects (for details, see part 2. Empirical material). An earlier and much shorter version of this article appeared (in Slovak) in Community psychology in Slovakia: Proceedings from a scientific conference ( Petrjánošová, 2015 ).

Allport, G. (1954). The nature of prejudice . Cambridge, MA: Perseus Books. Search in Google Scholar

Augustinos, M., & Every, D. (2007). The language of “race” and prejudice. Journal of Language and Social Psychology, 26 ,123-141. 10.1177/0261927X07300075 Search in Google Scholar

Billig, M. (2012). The notion of ‘prejudice’: Some rhetorical and ideological aspects. In J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge: Cambridge University Press. 10.1017/CBO9781139022736.009 Search in Google Scholar

Condor, S. (2000). Pride and prejudice: Identity management in English people’s talk about ‘this country’. Discourse & Society , 11 (2), 175-205. 10.1177/0957926500011002003 Search in Google Scholar

Condor, S. (2011). Sense and sensibility: The conversational etiquette of English national self-identification. In A. Aughey & C. Berbech (Eds.), These Englands: A conversation on national identity . Manchester: Manchester University Press. Search in Google Scholar

Condor, S., & Figgou, L. (2012). Rethinking the prejudice problematic: A collaborative cognition perspective. In J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge: Cambridge University Press. 10.1017/CBO9781139022736.012 Search in Google Scholar

Dionigi, R. A. (2015). Stereotypes of aging: Their effects on the health of older adults. Journal of Geriatrics , Article ID 954027, 9 p., available online at http://dx.doi.org/10.1155/2015/954027. 10.1155/2015/954027 Search in Google Scholar

Dixon, J., & Levine, M. (2012b). Conclusions and future directions: the nature, significance and inherent limitations of the concept of prejudice in social psychology in J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge University Press. 10.1017/CBO9781139022736.018 Search in Google Scholar

Dixon, J., & Levine, M.(2012b). Introduction in J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge University Press. 10.1017/CBO9781139022736 Search in Google Scholar

Dovidio, J. (2001). On the nature of contemporary prejudice: The third wave. Journal of Social Issues , 57 (4), 829-849. 10.1111/0022-4537.00244 Search in Google Scholar

Dovidio, J., Hewstone, M., Glick, P., & Esses, V. (2010). Prejudice, stereotyping and discrimination: Theoretical and empirical overview. In J. Dovidio, M. Hewstone, P. Glick, & V. Esses (Eds.), The SAGE handbook of prejudice, stereotyping and discrimination. (pp. 3-29). London: SAGE. 10.4135/9781446200919.n1 Search in Google Scholar

Durrheim, K. (2012). Implicit prejudice in mind and interaction. In J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge: Cambridge University Press. 10.1017/CBO9781139022736.011 Search in Google Scholar

Durrheim, K. Quayle, M., & Dixon, J. (2016). The struggle for the nature of “Prejudice”: “Prejudice” expression as identity performance. Political Psychology , 37 (1), 17-38. 10.1111/pops.12310 Search in Google Scholar

Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (pp. 357-411). New York: McGraw-Hill. Search in Google Scholar

Fiske, S.T., & North, M.S. (2014). Measures of stereotyping and prejudice: Barometers of bias. In G. Boyle & D. Saklofske (Eds.), Measures of personality & social psychological constructs. Cambridge, MA: Elsevier/Academic Press. 10.1016/B978-0-12-386915-9.00024-3 Search in Google Scholar

Fúskova, J., Hargašová, L., & Andraščíková, S. (2017). Why we can’t and shan’t measure gender. Human Affairs: Postdisciplinary Humanities and Social Sciences Quarterly, 27 (2), 205-218. 10.1515/humaff-2017-0018 Search in Google Scholar

Graf, S., Hřebíčková, M., Petrjánošová, M., & Leix, A. (2015). Češi a jejich sousedé: meziskupinové postoje a kontakt ve střední Evropě . Praha: Academia. Search in Google Scholar

Lášticová, B., & Findor, A. (2016). Developing explicit measures of stereotypes and anti-Roma prejudice in Slovakia: conceptual and methodological challenges. Human Affairs: Postdisciplinary Humanities and Social Sciences Quarterly, 26 (3), 233-252. 10.1515/humaff-2016-0022 Search in Google Scholar

Lášticová, B., & Petrjánošová, M. (2013). Sociálne inovácie a sociálne identity: občianske iniciatívy slovenských migrantov a migrantiek v Írsku [Social innovations and social identities: Civic initiatives of Slovak migrants in Ireland]. Slovenský národopis, 61 (3), 273-288. Search in Google Scholar

Lášticová, B., & Petrjánošová, M. (2014). Social identities, societal change, and mental borders: Identity construction strategies of young Slovaks commuting between Bratislava and Vienna. In Culture and political psychology: A societal perspective . (pp. 285-312). Charlotte: Information Age Publishing, Inc. Search in Google Scholar

Leix, A. (2013). I was expecting her to be a fanatic catholic, but she was not: How international exchange programmes reduce prejudice. The New Educational Review, 31 (1), 204-216. 10.15804/tner.13.31.1.18 Search in Google Scholar

Levine, J. M., Hogg, M. A. (Eds). (2010). Encyclopedia of Group Processes and Intergroup Relations. Thousand Oaks, London, New Delhi: Sage Publications. 10.4135/9781412972017 Search in Google Scholar

Paolini, S., Harwood, J., & Rubin, M. (2010). Negative intergroup contact makes group memberships salient: Explaining why intergroup conflict endures. Personality and Social Psychology Bulletin, 36 , 1723-1738. 10.1177/0146167210388667 Search in Google Scholar

Petrjánošová, M. (2015). Ako skúmať stereotypy kvalitatívnymi metódami [How to explore stereotypes using qualitative methods]. In Komunitná psychológia na Slovensku 2015: zborník z 2. ročníka vedeckej konferencie Komunitná psychológia na Slovensku (Bratislava, 30. 11. 2015). Ed. Júlia Sadovská Halamová (Ed.), (pp. 72-78). Bratislava: Ústav aplikovanej psychológie, Fakulta sociálnych a ekonomických vied UK. Search in Google Scholar

Popper, M., & Petrjánošová, M. (2016). Do we know what we are asking?: Individual and group cognitive interviews. Human Affairs: Postdisciplinary Humanities and Social Sciences Quarterly, 26 (3), 253-270. 10.1515/humaff-2016-0023 Search in Google Scholar

Spannring, R., Wächter, N., Datler, G., Bianchi, G., Lášticová, B., & Petrjánošová, M. (2005). Neue Gemeinschaften-neue Identitäten? Eine Studie zu territorialen Identitäten Jugendlicher in der österreichisch-slowakischen Grenzregion [New ….- new identities? A study about teritorial identities of young people in the Austrian-Slovak borderlands] . Unpublished Final Report from the Austrian National Bank Funded Project N. 10690. Wien: OEIJ. Search in Google Scholar

Tileaga, C. (2015). The nature of prejudice . Society, discrimination and moral exclusion. London: Routledge. 10.4324/9780203770696 Search in Google Scholar

Van Dijk, T.A. (1984). Prejudice in discourse. Amsterdam: Benjamins. 10.1075/pb.v.3 Search in Google Scholar

Van Maanen (1988). Tales of the field. On writing ethnography . Chicago: University of Chicago Press. Search in Google Scholar

Vázquez, J. J., Panadero, S., & Zúniga, C. (2017). Content and uniformity of stereotypes and meta-stereotypes of homeless people in Madrid (Spain). Journal of Community Psychology, 45 (1), 128-137. 10.1002/jcop.21836 Search in Google Scholar

Wetherell, M. (2012). The prejudice problematic . In J. Dixon & M. Levine (Eds.), Beyond prejudice . Cambridge: Cambridge University Press. 10.1017/CBO9781139022736.010 Search in Google Scholar

Willis, G. B. (1999). Cognitive interviewing: A how to guide. Available online at http://www.hkr.se/pagefiles/35002/gordonwillis.pdf Search in Google Scholar

Wimmer, A., & Glick Schiller, N. (2002). Methodological nationalism and beyond: nation-state building, migration and the social sciences. Global Networks, 4 (2), 301-334. 10.1111/1471-0374.00043 Search in Google Scholar

Wodak, R., & Krzyzanowski, M. (Eds.) (2008). Qualitative discourse analysis in the social sciences . Basingstoke: Palgrave MacMillan. 10.1007/978-1-137-04798-4 Search in Google Scholar

Wodak, R., & Meyer, M. (2001). Methods of critical discourse analysis . Great Britain: SAGE Publications. 10.4135/9780857028020 Search in Google Scholar

© 2018 Institute for Research in Social Communication, Slovak Academy of Sciences

• X / Twitter

Supplementary Materials

Please login or register with De Gruyter to order this product.

Journal and Issue

Articles in the same issue.

Stereotype, Prejudice, and Discrimination: Changing Conceptions in Theory and Research

Cite this chapter.

• Wolfgang Stroebe &
• Chester A. Insko  

Part of the book series: Springer Series in Social Psychology ((SSSOC))

1358 Accesses

66 Citations

Readers of the chapters on prejudice and discrimination in the three editions of the Handbook of Social Psychology (Harding, Kutner, Proshansky, & Chein, 1954; Harding, Proshansky, Kutner, & Chein, 1969; Stephan, 1985) will be impressed by the reduction in theoretical perspectives which this area seems to have experienced within the space of less than two decades. While the earlier chapters (Harding et al., 1954, 1969) approached prejudice and stereotypes from multiple theoretical perspectives, covering psychoanalytic, sociological, developmental, and personality-oriented explanations, Stephan’s (1985) chapter focuses only on one perspective, the cognitive approach.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Unable to display preview.  Download preview PDF.

Similar content being viewed by others

Prejudice: Intra- and Interpersonal Aspects

Stereotypes and Prejudice in Conflict: A Developmental Perspective

Psychological and Sociological Perspectives on the Acquisition of Ethnic and Racial Prejudice in Children

Adorno, T.W., Frenkel-Brunswik, E., Levinson, D.J., & Sanford, R.N. (1950). The authoritarian personality . New York: Harper.

Google Scholar  

Ajzen, I., & Fishbein, M. (1973). Attitudinal and normative variables as predictors of specific behaviors. Journal of Personality and Social Psychology, 27 , 41–57.

Article   Google Scholar  

Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, 84 , 888–918.

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting behavior. Englewood Cliffs, NJ: Prentice Hall.

Allport, G.W. (1954). The nature of prejudice . Reading, MA: Addison-Wesley.

Allport, G.W., & Kramer, B.M. (1946). Some roots of prejudice. Journal of Psychology, 22 , 9–39.

Anderson, N.H. (1965). Averaging versus adding as a stimulus-combination rule in impression formation. Journal of Experimental Psychology, 70 , 394–400.

Article   PubMed   Google Scholar  

Ashmore, R.D. (1970). The problem of intergroup prejudice. In: B.E. Collins, Social psychology (pp. 246–296 ). Reading, MA: Addison-Wesley.

Ashmore, R.D., & Del Boca, F.K. (1981). Conceptual approaches to stereotypes and stereotyping. In D. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 1–35 ). Hillsdale, NJ: Erlbaum.

Bettelheim, B., & Janowitz, M. (1950). Dynamics of prejudice . New York: Harper.

Bettelheim, B., & Janowitz, M. (1964). Social change and prejudice: Dynamics of prejudice . London: The Free Press of Glencoe.

Billig, M. (1976). Social psychology and intergroup relations . London: Academic Press.

Boume, L.E. (1966). Human conceptual behavior . Boston: Allyn & Bacon.

Breckler, S.J. (1984). Empirical validation of affect, behavior, and cognition as distinct components of attitude. Journal of Personality and Social Psychology, 47 , 1191–1205.

Brigham, J.C. (1971). Ethnic stereotypes. Psychological Bulletin, 76 , 15–38.

Brown, R. (1958). Words and things . Glencoe, IL: Free Press.

Brown, R. (1965). Social Psychology . New York: Free Press.

Campbell, D.T. (1956). Enhancement of contrast as composite habit. Journal of Abnormal and Social Psychology, 53 , 350–355.

Hamilton, D.L. (1981). Illusory correlation as a basis for stereotyping. In D.L. Hamilton (Ed.), Cognitive processes in stereotyping and intergroup behavior . (pp. 115–144 ). Hillsdale, NJ: Erlbaum.

Hamilton, D.L., & Gifford, R.K. (1976). Illusory correlation in interpersonal perception: A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 12 , 392–407.

Hamilton, D.L., & Trolier, T.K. (1986). Stereotypes and stereotyping: An overview of the cognitive approach. In J.F. Dovidio & S.L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 127–158 ). Orlando, FL: Academic Press.

Harding, J., Kutner, B., Proshansky, H., & Chein, I. (1954). Prejudice and ethnic relations. In G. Lindzey (Ed.), Handbook of social psychology (Vol. 2 ) (pp. 1021–1061 ). Cambridge, MA: Addison-Wesley.

Harding, J., & Proshansky, H., & Kutner, B., & Chein, I. (1969). Prejudice and ethnic relations. In G. Lindzey & E. Aronson (Eds.), The handbook of social psychology (Vol. 5 ) (pp. 1–76 ). Reading, MA: Addison-Wesley.

Heider, E (1958). The psychology of interpersonal relations . New York: Wiley.

Book   Google Scholar  

Hyman, H.H., & Sheatsley, P.B. (1954). The authoritarian personality: A methodological critique. In R. Christie & M. Jahoda (Eds.), Studies in the scope and method of the authoritarian personality (pp. 50–122 ). Glencoe, IL: The Free Press.

Insko, C.A., & Nacoste, R.W., & Moe, J.L. (1983). Belief congruence and racial discrimination: Review of the evidence and critical evaluation. European Journal of Social Psychology, 13 , 153–174.

Jonas, K., & Hewstone, M. (1986). The assessment of national stereotypes: A methodological study. Journal of Social Psychology, 126 , 745–754.

Jones, E.E., & Davis, K.E. (1965). From acts to dispositions. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2 ) (pp. 219–266 ). New York: Academic Press.

Karlins, M., & Coffman, T.L., & Walters, G. (1969). On the fading of social stereotypes: Studies in three generations of college students. Journal of Personality and Social Psychology, 13 , 1–16.

Katz, D., & Braly, K.W. (1933). Racial stereotypes of 100 college students. Journal of Abnormal and Social Psychology, 28 , 280–290.

Kirscht, J.P., & Dillehay, R.C. (1967). Dimensions of authoritarianism: A review of research and theory . Lexington, KY: University of Kentucky Press.

Klineberg, O. (1951). The scientific study of national stereotypes. International Social Science Bulletin , 3, 505–515.

Koblinsky, S.G., & Cruse, D.F., & Sugawara, A.I. (1978). Sex role stereotypes and children’s memory of story content. Child Development, 49 , 452–458.

Krech, D., & Crutchfield, R.S. (1948). Theory and problems of social psychology . New York: McGraw-Hill.

Kutner, B., Wilkins, C., & Yarrow, P.R. (1952). Verbal attitudes and overt behavior involving racial prejudice. Journal of Abnormal and Social Psychology, 47 , 649–652.

LaPiere, R.T. (1934). Attitudes versus actions. Social Forces, 13 , 230–237.

LeVine, R.A., & Campbell, D.T. (1972). Ethnocentrism: Theories of conflict, ethnic attitudes, and group behavior . New York: Wiley.

Liberman, A.M., & Harris, K.S., & Howard, S.H., & Griffith, B.C. (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology, 54 , 358–368.

Lilli, W. (1970). Das Zustandekommen von Stereotypen über einfache and komplexe Sachverhalte: Experimente zum klassifizierenden Urteil. Zeitschrift far Sozialpsychologie, 1 , 57–79.

Lilli, W., & Lehner, F. (1971). Stereotype Wahrnehmung: Eine Weiterentwicklung der Theorie Tajfel’s. Zeitschrift fur Sozialpsychologie , 2, 285–294.

Lilli, W., & Rehm, J. (1988). Judgmental bases of intergroup conflict. In W. Stroebe, A.W. Kruglanski, D. Bar-Tal, M. Hewstone (Eds.), The social psychology of intergroup conflict (pp. 29–45 ). New York: Springer.

Chapter   Google Scholar  

Linville, P.W. (1982). The complexity-extremity effect and age-based stereotyping. Journal of Personality and Social Psychology, 42 , 193–211.

Lippman, W. (1922). Public opinion . New York: Harcourt & Brace.

Marchand, B. (1970). Auswirkung einer emotional wertvollen und einer emotional neutralen Klassifikation auf die Schätzung einer Stimulusserie. Zeitschrift für Sozialpsychologie, 1 , 370–376.

McArthur, L.Z., & Friedman, S.A. (1980). Illusory correlation in impression formation: Variations in the shared distinctiveness effect as a function of the distinctive person’s age, race, and sex. Journal of Personality and Social Psychology, 39 , 615–624.

McCauley, C., & Stitt, C.L. (1978). An individual and quantitative measure of stereotypes. Journal of Personality and Social Psychology , 39, 929–940.

McCauley, C., & Stitt, C.L., & Segal, M. (1980). Stereotyping: From prejudice to prediction. Psychological Bulletin, 87 , 195–208.

Miller, N.E., & Bugelski, R. (1948). Minor studies in aggression: The influence of frustrations imposed by the in-group on attitudes expressed toward out-groups. Journal of Psychology, 25 , 437–442.

Minard, R.D. (1952). Race relationships in the Pocahontas coal field. Journal of Social Issues, 8 , 29–44.

Moe, J.L., & Nacoste, R.W., & Insko, C.A. (1981). Belief versus race as determinants of discrimination: A study of southern adolescents in 1966 and 1979. Journal of Personality and Social Psychology, 41 , 1031–1050.

Pettigrew, T.F. (1958). Personality and sociocultural factors in intergroup attitudes: A cross-national comparison. Journal of Conflict Resolution , 2, 29–42.

Pettigrew, T.F. (1959). Regional differences in an anti-Negro prejudice. Journal of Abnormal and Social Psychology , 59, 28–36.

Plotkin, L. (1964). The frequency of appearance of Negroes on television . New York: The Committee on Integration, New York Society for Ethical Culture.

Popper, K.R. (1963). Conjectures and refutations . London: Routledge & Kegan Paul.

Robinson, J.E., & Insko, C.A. (1969). Attributed belief similarity-dissimilarity versus race as determinants of prejudice: A further test of Rokeach’s theory. Journal of Experimental Research in Personality, 4 , 72–77.

Rokeach, M. (1960). The open and closed mind . New York: Basic Books.

Rokeach, M., & Smith, P.W., & Evans, R.I. (1960). Two kinds of prejudice or one? In M. Rokeach (Ed.), The open and closed mind (pp. 132–168 ). New York: Basic Books.

Rosenberg, M.J. (1960). An analysis of affective-cognitive consistency. In C.I. Hovland & M.J. Rosenberg (Eds.), Attitude organization and change (pp. 15–64 ). New Haven: Yale University Press.

Schaller, M., & Maas, A. (1987). Illusory correlations and social categorization . Unpublished manuscript, University of Padova, Italy.

Seago, D.W. (1947). Stereotypes: Before Pearl Harbor and after. Journal of Psychology , 23, 55–63.

Secord, P.F. (1959). Stereotyping and favorableness in the perception of Negro faces. Journal of Abnormal and Social Psychology, 59 , 309–314.

Secord, P.F., & Bevan, W., & Katz, B. (1956). The Negro stereotype and perceptual accentuation. Journal of Abnormal and Social Psychology, 53 , 78–83.

Sherif, M. (1967). Group conflict and cooperation . London: Routledge & Kegan Paul.

Simon, H.A. (1955). A behavior model of rational choice. Quarterly Journal of Economics, 69 , 99–118.

Smedley, J.W., & Bayton, J.A. (1978). Evaluation of race-class stereotypes by race and per- ceived class of subjects. Journal of Personality and Social Psychology, 36 , 530–535.

Stagner, R., & Congdon, C.S. (1955). Another failure to demonstrate displacement of aggression. Journal of Abnormal and Social Psychology, 51 , 695–696.

Stapf, K.H., & Stroebe, W., & Jonas, K. (1986). Amerikaner über Deutschland und die Deutschen: Urteile und Vorurteile . Köln: Westdeutscher Verlag.

Stein, D.D., & Hardyck, J.A., & Smith, M.B. (1965). Race and belief: An open and shut case. Journal of Personality and Social Psychology , 1, 281–289.

Stephan, W.G. (1985). Intergroup relations. In G. Lindzey & E. Aronson (Eds.), The handbook of social psychology (Vol. 2 ) (pp. 599–658 ). New York: Random House.

Stephan, W.G., & Rosenfield, D. (1982). Racial and ethnic attitudes. In A.G. Miller (Ed.), In the eye of the beholder: Contemporary issues in stereotyping (pp. 92–136 ). New York: Praeger.

Tajfel, H. (1957). Value and the perceptual judgment of magnitude. Psychological Review, 64 , 192–204.

Tajfel, H. (1959a). The anchoring effects of value in a scale of judgments. British Journal of Psychology, 50 , 294–304.

Tajfel, H. (1959b). Quantitative judgment in social perception. British Journal of Psychology, 50 , 192–204.

Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues , 25, 79–97.

Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223(5) pp. 96–102.

Tajfel, H. (1981). Human groups and social categories . Cambridge: Cambridge University Press.

Tajfel, H. (1982). Social psychology of intergroup relations. In M.R. Rosenzweig & L.R. Porter (Eds.), Annual review of psychology (pp. 1–39 ). Palo Alto, CA: Annual Reviews.

Tajfel, H., & Jahoda, G. (1966). Development in children of concepts and attitudes about their own and other countries. Proceedings of the 18th International Congress of Psychology, 36 , 17–33. Moscow Symposium.

Tajfel, H., & Turner, J. (1979). An integrative theory of intergroup conflict. In W.G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33–47 ). Belmont, CA: Wadsworth.

Tajfel, H., & Turner, J. (1985). The social identity theory of intergroup behavior. In S. Worchel & W.G. Austin (Eds.), Psychology of intergroup relations (pp. 7–24 ). Chicago, IL: Nelson-Hall.

Tajfel, H., & Wilkes, A.L. (1963). Classification and quantitative judgment. British Journal of Psychology, 54 , 101–114.

Triandis, H.C., & Triandis, L.M. (1965). Some studies of social distance. In I.D. Steiner & M. Fishbein (Eds.), Recent studies of social psychology (pp. 207–217 ). New York: Holt.

Turner, C. (1987). Rediscovering the social group . Oxford: Blackwell.

Weitz, S. (1972). Attitude, voice, and behavior: A repressed affect model of interracial interaction. Journal of Personality and Social Psychology, 24 , 14–21.

Wicker, A.W. (1969). Attitude versus actions: The relationship of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 25 , 41–78.

Zanna, M.P., & Rempel, J.K. (1988). Attitudes: A new look at an old concept. In D. Bar-Tal & A. Kruglanski (Eds.), The social psychology of knowledge (pp. 315–334 ). New York: Cambridge University Press.

Download references

You can also search for this author in PubMed   Google Scholar

Editor information

Editors and affiliations.

School of Education, Tel-Aviv University, 69978, Tel-Aviv, Israel

Daniel Bar-Tal

Psychologisches Institut, Universität Heidelberg, D-6900, Heidelberg 1, Federal Republic of Germany

Carl F. Graumann

Department of Psychology, University of Maryland, 20742, College Park, MD, USA

Arie W. Kruglanski

Psychologisches Institut, Universität Tübingen, 7400, Tübingen, Federal Republic of Germany

Wolfgang Stroebe

Rights and permissions

Reprints and permissions

Copyright information

© 1989 Springer Science+Business Media New York

About this chapter

Stroebe, W., Insko, C.A. (1989). Stereotype, Prejudice, and Discrimination: Changing Conceptions in Theory and Research. In: Bar-Tal, D., Graumann, C.F., Kruglanski, A.W., Stroebe, W. (eds) Stereotyping and Prejudice. Springer Series in Social Psychology. Springer, New York, NY. https://doi.org/10.1007/978-1-4612-3582-8_1

Download citation

DOI : https://doi.org/10.1007/978-1-4612-3582-8_1

Publisher Name : Springer, New York, NY

Print ISBN : 978-1-4612-8165-8

Online ISBN : 978-1-4612-3582-8

eBook Packages : Springer Book Archive

Share this chapter

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research

Prejudice and stereotypes

How do you deal with them?

Everyone is prejudiced and uses stereotypes. That’s normal. But prejudice can lead to unequal treatment. Watch the video to see how prejudice works and what you can do about it. Read the answers to frequently asked questions and watch the personal stories of some young people.

Frequently asked questions

How does prejudice come about?

Is it bad to be prejudiced?

When does prejudice become dangerous?

What can you do against prejudice?

Stereotypes and Prejudice

Stereotypes are widely held generalized beliefs about the behaviors and attributes possessed by individuals from certain social groups (e.g., race/ethnicity, sex, age, socioeconomic status, sexual orientation). They are often unchanging even in the face of contradicting information; however, they are fluid in the sense that stereotypic beliefs do not always come to mind or are expressed unless a situation activates the stereotype. Stereotypes generally serve as an underlying justification for prejudice, which is the accompanying feeling (typically negative) toward individuals from a certain social group (e.g., the elderly, Asians, transgender individuals). Many contemporary social issues are rooted in stereotypes and prejudice; thus research in this area has primarily focused on the antecedents and consequences of stereotype and prejudice as well as the ways to minimize the reliance on stereotypes when making social judgments.

• Related Documents

Diversities in Group Specialty Practice

Chapter 11 details diversities in group specialty practice. Group therapies and other group intervention strategies are uniquely positioned to focus on issues of diversity, those issues of multiculturalism such as race, ethnicity, national origin, gender, sexual orientation, socioeconomic status, and age.

Minority Health and Health Disparities in the 21st Century: A Review

Powerful, complex relationships exist between health and biology, genetics, and individual behavior, and between health and health services, socioeconomic status, the physical environment, discrimination, racism, literacy levels, and legislative policies. These factors, which influence an individual’s or population’s health, are known as determinants of health. Today, health disparity is taking an in depth look at the differences in health status between different social groups, gender, race, ethnicity, education, income, disability, and sexual orientation. While on the other hand, health inequality is looking at the unjust and unfair treatment one gets because of their socioeconomic status and demographic area. Such a wide array of differences in health inequality and disparity is what contributes to the United States ranking in the bottom of industrialized western nations when it comes to life expectancy rate, and infant mortality rate. Even though over the years there have been great improvements and changes, there is still more work to be done to make health and equality for all.

Facial Impressions Are Predicted by the Structure of Group Stereotypes

Impressions of other people’s faces (e.g., trustworthiness) have long been thought to be evoked by morphological variation (e.g., upturned mouth) in a universal, fixed manner. However, recent research suggests that these impressions vary considerably across perceivers and targets’ social-group memberships. Across 4,247 U.S. adults recruited online, we investigated whether racial and gender stereotypes may be a critical factor underlying this variability in facial impressions. In Study 1, we found that not only did facial impressions vary by targets’ gender and race, but also the structure of these impressions was associated with the structure of stereotype knowledge. Study 2 extended these findings by demonstrating that individual differences in perceivers’ own unique stereotype associations predicted the structure of their own facial impressions. Together, the findings suggest that the structure of people’s impressions of others’ faces is driven not only by the morphological variation of the face but also by learned stereotypes about social groups.

Correct and Preferred Usage

The Correct and Preferred Usage chapter of the 11th edition of the AMA Manual of Style includes a large glossary of frequently misused or confused terms. In addition to this updated and cross-referenced glossary, additional examples illustrate other usage problems, such as redundant, expendable, and incomparable words and phrases; jargon; and back-formations. Specific advice is provided on describing the administration of drugs, age and sex referents, anatomy, clock referents, laboratory values, and articles before abbreviations and acronyms and before the aspirate h and nonvocalicy. Guidance is also provided on inclusive language (eg, sex/gender, presenting data in tables, personal pronouns, sexual orientation, race/ethnicity, age, socioeconomic status, and terminology for persons with diseases, disorders, or disabilities). New to this edition is a discussion of spelling and spacing variations, with preferences of the JAMA Network journals indicated.

Reframing Polarization: Social Groups and “Culture Wars”

ABSTRACTRecent analyses of American politics often invoke the term “culture war” depicting sharp and increasing divisions within the American polity. Most of this research defines culture in terms of values and beliefs about social issues and defines polarization in terms of partisan and issue divisions. I evaluate the claim of worsening “culture wars” by using a conceptualization of political culture that focuses on social groups and measuring polarization as both social group members’ attitudes toward their own social in-groups and out-groups, and the effects of group attitudes on partisanship. Analyzing inter-group attitudes from 1964 to 2012 for social group cleavages defined by race, class, age, sex, and religion shows that polarization in attitudes toward social groups is minimal and generally stable, and most group members feel positively toward out-groups. Partisan and issue polarization seen in prior research do not extend to deep or increasing inter-group hostility that could reinforce issue-based and partisan polarization.

Abstract 57: Vulnerable Populations, Disparities, and the U.S. Tobacco Endgame

Background: Since 1964 when the first Surgeon General’s report on smoking was published, the decline in smoking has been steepest among those with the most education. As smoking has become denormalized, it is increasingly concentrated in vulnerable populations that are marginalized by socioeconomic status, mental health, race/ethnicity, or sexual orientation. This analysis uses the National Health Interview Survey (NHIS) to assess current trends in cigarette smoking among vulnerable vs. non-vulnerable populations. Methods: Current cigarette smoking among adult NHIS respondents was tracked from 1999-2018. Vulnerable populations were defined by low socioeconomic status (<100% federal poverty level, <high school education, Medicaid recipient), low mental health (Kessler 6 score≥5), or race/ethnicity with elevated smoking (American Indian, multiracial). Sensitivity analyses added food insecurity and minority sexual orientation (non-straight), available starting 2013, to the definition of vulnerability. Regression analysis projected future smoking trends. Results: Smoking rates in all sub-groups declined from 1999 to 2018. Among the non-vulnerable, smoking declined from 19.2% to 9.7%; among the vulnerable, smoking rates were 10 percentage points higher, declining from 30.5% to 19.5%. Smoking rates among vulnerable populations are now double the rates in non-vulnerable populations. The proportion of smokers classified as vulnerable increased from 47.7% to 58.6%; the proportion of cigarettes consumed by vulnerable smokers increased from 50.1% to 60.4%. Expanding the definition of vulnerability to include those food insecure and in sexual minorities, the proportion of smokers classified as vulnerable (2018) was 61.63%. The most common vulnerabilities in the population (2018) were low socioeconomic status (25.5%) and poor mental health (22.0%); (7.8% were both). In 2018, as the number of vulnerabilities increased (0, 1, 2, ≥3), so did the smoking rate (9.4%, 14.6%, 21.7%, 31.7%, respectively, p<0.001). Using linear regression to extrapolate current trends, non-vulnerable populations will eliminate cigarette smoking in 2037, vulnerable populations in 2060. Conclusion: Existing public health policy and social trends are reducing smoking rates among all smokers, but absent concentrated effort to reduce smoking in vulnerable populations, health disparities from disproportionate tobacco use will increase and reverberate for decades.

Facial Impressions are Predicted by the Structure of Group Stereotypes

Facial impressions (e.g., trustworthy, assertive) have long been thought to be evoked by morphological variation (e.g., upturned mouth) in a universal, fixed manner. However, recent research suggests that these impressions vary considerably across different perceivers and targets’ social group memberships. We investigated whether stereotypes at the intersection of race and gender may be a critical factor underlying this variability in facial impressions. In Study 1, we found that not only did facial impressions vary by targets’ gender and race, but the structure of these impressions was predicted by the structure of stereotype knowledge. Study 2 extended these findings by demonstrating that individual differences in perceivers’ own unique stereotype associations predicted the structure of their own facial impressions. Together, the findings suggest that the structure of our impressions of others’ faces is driven not only by the morphological variation of the face, but also our own learned stereotypes about social groups.

Sexual orientation and sex differences in socioeconomic status: a population-based investigation in the National Longitudinal Study of Adolescent to Adult Health

BackgroundSocioeconomic status (SES) is a fundamental contributor to health; however, limited research has examined sexual orientation differences in SES.Methods2008–2009 data from 14 051 participants (ages 24–32 years) in the US-based, representative, National Longitudinal Study of Adolescent to Adult Health were analysed using multivariable regressions that adjusted for age, race-ethnicity, childhood SES, urbanicity and Census region, separately for females and males. Modification by racial minority status (black or Latino vs white, non-Hispanic) was also explored.ResultsAmong females, sexual minorities (SM) (10.5% of females) were less likely to graduate college, and were more likely to be unemployed, poor/near poor, to receive public assistance and to report economic hardship and lower social status than heterosexuals. Adjusting for education attenuated many of these differences. Among males, SM (4.2% of males) were more likely than heterosexuals to be college graduates; however, they also had lower personal incomes. Lower rates of homeownership were observed among SM, particularly racial minority SM females. For males, household poverty patterns differed by race-ethnicity: among racial minority males, SM were more likely than heterosexuals to be living at >400% federal poverty level), whereas the pattern was reversed among whites.ConclusionsSexual minorities, especially females, are of lower SES than their heterosexual counterparts. SES should be considered a potential mediator of SM stigma on health. Studies of public policies that may produce, as well as mitigate, observed SES inequities, are warranted.

Biopolítica, Governamentalidade Digital e Tanatopolítica: idosos e a pandemia de covid-19

 Este artigo é um ensaio que aborda a pandemia pelo Covid-19, a partir de uma perspectiva biopolítica e biodigital, assinalando aspectos da precariedade acirrada vivida por idosos frente ao contágio pelo novo coronavírus e os efeitos nefastos deste em suas existências. Busca-se pensar elementos da sociedade de controle e os enquadramentos da política de morte dirigida a este grupo social quanto ao deixar morrer e ao estigma voltado aos idosos como estratégia de gestão da população, no presente. Portanto, visamos analisar práticas sociais que produzem quadros de ausência de proteção, de reconhecimento e expansão da vida de grupos marcados pelo envelhecimento, constituindo-os como um peso e problema para a sociedade contemporânea. Com efeito, utiliza-se a velhice como figura de um viver que não tem valor e não é digno de comoção nem passível de luto durante a gestão da pandemia por Covid-19.Palavras-chave: Idosos. Pandemia de Covid-19. Biopolítica. Biovigilância. Precariedade. Biopolitics, Digital and Tanatopolitical Governance: elderly people and the pandemic of covid-19ABSTRACT This article is an essay that addresses the Covid-19 pandemic, from a biopolitical and biodigital perspective, pointing out aspects of the severe precariousness experienced by the elderly in the face of contagion by the new coronavirus and its harmful effects on their lives. It seeks to think about elements of the control society and the framework of the death policy directed at this social group regarding the letting die and the stigma towards the elderly as a population management strategy, at present. Therefore, we aim to analyze social practices that produce situations of lack of protection, recognition and expansion of the lives of groups marked by aging, constituting them as a weight and problem for contemporary society. Indeed, old age is used as a figure of living that has no value and is not worthy of commotion or mourning during the pandemic management by Covid-19. Keywords: Elderly. Covid-19 Pandemic. Biopolitics. Biovigilance. Precariousness.

National Multicultural Conference and Summit II: "The Psychology of Race/Ethnicity, Gender, Sexual Orientation, and Disability: Intersections, Divergence and Convergence"

Export citation format, share document.

• Search Menu
• Sign in through your institution
• Advance Articles
• Author Guidelines
• Submission Site
• Open Access
• Why Submit?
• About Journal of Communication
• About International Communication Association
• Editorial Board
• Advertising and Corporate Services
• Journals Career Network
• Self-Archiving Policy
• Journals on Oxford Academic
• Books on Oxford Academic

Article Contents

Media stereotypes and prejudice, toward the dynamic of self-reinforcing effects, overview of the empirical work and hypotheses, study 1: prejudice-based selective exposure, study 2: effects of forced exposure, study 3: self-selected exposure and preference-based reinforcement, integrated analysis (studies 2 and 3): net-effect perspective, study 4: net-effect perspective (replication), general discussion, supplementary material, data availability.

• < Previous

Media stereotypes, prejudice, and preference-based reinforcement: toward the dynamic of self-reinforcing effects by integrating audience selectivity

• Article contents
• Figures & tables
• Supplementary Data

Florian Arendt, Media stereotypes, prejudice, and preference-based reinforcement: toward the dynamic of self-reinforcing effects by integrating audience selectivity, Journal of Communication , Volume 73, Issue 5, October 2023, Pages 463–475, https://doi.org/10.1093/joc/jqad019

• Permissions Icon Permissions

The media portray various social groups stereotypically, and studying the effects of these portrayals on prejudice is paramount. Yet, audience selectivity—inherent within today’s high-choice media environments—has largely been disregarded. Relatedly, the predominance of forced-exposure designs is a source of concern. This article proposes the integration of audience selectivity into media stereotype effects research. Study 1 ( N  =   1,166) indicated that prejudiced individuals tended to approach prejudice-consistent stereotypical news and avoid prejudice-challenging counter-stereotypical news. Using a forced-exposure experiment, study 2 ( N  =   380) showed detrimental effects of prejudice-consistent news and beneficial effects of prejudice-challenging news. Relying on a self-selected exposure paradigm, study 3 ( N  =   1,149) provided evidence for preference-based reinforcement. Study 4’s “net-effect perspective” ( N  =   937) indicated that operationalizing exposure as forced or self-selected can lead to different interpretations of actual societal effects. The findings emphasize the key role played by audience selectivity when studying media effects.

The media portray many social categories, such as minority groups, in a stereotypical way ( Billings & Parrott, 2020 ). This is problematic, as media depictions often represent the main, if not the only, source of information for citizens ( Ramasubramanian & Murphy, 2014 ). Previous research has already acknowledged the importance of studying the effects of exposure to media stereotypes on essential outcomes, such as prejudice. This line of research indicates that while exposure to prejudice-consistent stereotypical depictions can elicit detrimental effects, exposure to prejudice-challenging counter-stereotypical depictions can lead to beneficial effects ( Holt, 2013 ; Mastro & Tukachinsky, 2011 ; Ramasubramanian, 2011 ; Saleem et al., 2017 ).

Yet, the predominance of forced-exposure experiments in previous research on the effects of media stereotypes (e.g., Arendt, 2013 ; Kroon et al., 2016 ; Oliver, 1999 ; Schmuck et al., 2017 ) is a source of concern, as it disregards audience selectivity . Indeed, in the fragmented media environments of today, citizens can choose to expose themselves to content that confirms their preexisting views ( Ramasubramanian & Murphy, 2014 ). From this perspective, some scholars have argued that media effects can be conceptualized as preference-based reinforcement ( Cacciatore et al., 2016 ). Accordingly, an almost exclusive reliance on forced-exposure experiments with “captive participants” ( Druckman et al., 2012 , p. 430) may not suffice. This is because even if a forced-exposure experiment shows, for example, the beneficial effects of prejudice-challenging news, we do not know whether prejudiced individuals would actually read this kind of content in their everyday life. Rather, knowledge on both aspects—i.e., whether individuals select certain types of content and whether exposure elicits causal effects —is helpful for a thorough understanding of the media stereotype effect process.

We contribute to the literature in three important ways. Building upon recent advances in the theorizing on media effects in general ( Cacciatore et al., 2016 ; Knobloch-Westerwick, 2015 ; Slater, 2007 ), we (1) integrate audience selectivity. The effects of media stereotype exposure are conceptualized as a process occurring over time, combining both strands of research: selective exposure and effects. This is the primary theoretical contribution. We report on four studies, demonstrating the added value of integrating audience selectivity by studying preference-based reinforcement and the dynamic of self-reinforcing effects . As a second theoretical contribution, the present paper (2) investigated the role of approach and avoidance tendencies for selective exposure. Previous work has already acknowledged that the two tendencies may result in biased news choice ( Garrett, 2009 ; Jang, 2014 ; Schmuck et al., 2020 ). It is of theoretical importance to treat the two tendencies as separate ( Schmuck et al., 2020 ), because prejudiced individuals can seek prejudice-consistent news without avoiding prejudice-challenging news, and vice versa. There is limited knowledge on the role of both tendencies. As a methodological contribution, the present paper (3) contributes to closing the gap between reality (the high-choice media environments of today) and the methodology (the predominance of forced-exposure designs in existing work). Empirical evidence indicates that operationalizing exposure as forced or self-selected can lead to different interpretations of actual societal effects. Thus, scholars should increasingly think about the use of the self-selected-exposure paradigm as a supplement to forced-exposure experiments . Although both exposure paradigms are relevant, when used in combination , they may help to deepen our understanding.

Stereotypes can be defined as associations between a social category and attributes ( Greenwald et al., 2002 ). Stereotypes are typically simple, overgeneralized, widely accepted, and often resistant to change. Human information processing relies on stereotypes as they reduce the complexity of the social world by serving to stabilize, make predictable, and make manageable a given person’s view on social reality ( Snyder, 1981 ). Unfortunately, the “world outside” often stands in stark contrast to the “pictures in our heads,” as Lippmann (1922) noted. In fact, stereotypes are often inaccurate insofar as most of the time it is simply wrong that all representatives of a given social category show a given attribute. Such overgeneralizations are especially problematic when it comes to stereotypes that include negatively valenced attributes. Importantly, the association of a social category with attributes such as “rapist” or “thief” is likely to be related to an antipathy toward this social category, that is, prejudice ( Allport, 1954 ).

On a most basic level, the concept of a stereotype is agnostic about the valence of the attributes. They can be negatively valenced (e.g., criminal), positively valenced (e.g., diligent), or rather neutral (e.g., tall), as stereotypes are conceptualized on a cognitive level ( Greenwald et al., 2002 ). However, most studies on the detrimental effects of exposure to media stereotypes on prejudice have focused on attributes with a negative valence and have thus studied the effects of prejudice-consistent stereotypical media content (see Billings & Parrott, 2020 ). Consistently, studies on the beneficial effects on prejudice reduction have focused heavily on the effects of prejudice-challenging counter-stereotypical media content—depictions of the social category that go against the widely shared (prejudice-consistent) stereotype (see Billings & Parrott, 2020 ). We also directed our attention to prejudice-consistent stereotypical and prejudice-challenging counter-stereotypical media content.

There is evidence for the detrimental effects of exposure to prejudice-consistent depictions, as well as—albeit to a substantially lesser extent—for the beneficial effects of exposure to prejudice-challenging depictions ( Holt, 2013 ; Mastro & Tukachinsky, 2011 ; Power et al., 1996 ; Ramasubramanian, 2011 ; Saleem et al., 2017 ). Yet, the available evidence comes with some caveats. Although some work utilized nonrandomized observational survey studies with known limitations regarding causal interpretations (e.g., Arendt & Northup, 2015 ; Dixon, 2008 ; Mastro et al., 2007 ; Ramasubramanian, 2013 ; Schemer, 2012 ), most studies on the effects of media stereotypes have utilized experiments relying on forced exposure (e.g., Arendt, 2013 ; Oliver, 1999 ; Schmuck et al., 2017 ). The limits of forced-exposure experiments become evident when today’s high-choice media environments are considered. Following Bennett and Iyengar (2008) , Cacciatore and colleagues (2016) argued that such environments increasingly allow media users to be paired with content that fits with their preexisting views, implicating how media users tend to mostly rely on highly homophilic self-selected content. Ramasubramanian and Murphy (2014) argued in a similar vein, explaining that the media users of today “exercise greater authority when deciding what type of messaging they will allow themselves to be exposed to” (p. 396). The predominance of the forced-exposure paradigm in previous effects research does not do justice to the characteristics of fragmented, high-choice media environments. Audience selectivity must be better accounted for in our theorizing and empirical work. We need to know both whether individuals select certain types of content and whether exposure to this content elicits effects.

We now theorize on the media stereotype effect process, aiming to integrate audience selectivity into the study of the effects of exposure to media stereotypes. This theorizing guided our empirical work and owed much to existing writings focused on the importance of audience selectivity in the study of media effects in general ( Cacciatore et al., 2016 ; Knobloch-Westerwick, 2015 ; Slater, 2007 ). Two general ideas are constitutive for our theorizing:

Prejudice-based selective exposure

As a first general idea, we assumed that prejudice predicts the selection of news, such that it causes people to opt for congenial content and to fend off challenging content. Thus, prejudiced predispositions are assumed to facilitate the approach to prejudice-consistent content and hinder the selection of prejudice-challenging content. This prejudice-based selective exposure is a form of attitude-based selective exposure (e.g., Arendt et al., 2019 ) given that prejudice is conceptualized as a negative attitude toward a social category ( Allport, 1954 ; Dovidio et al., 2010 ). It is driven by a self-consistency motive ( Knobloch-Westerwick, 2015 ) that governs selective exposure insofar as individuals prefer news that aligns with their attitudes, a phenomenon also known as confirmation bias ( Knobloch-Westerwick & Kleinman, 2012 ).

Note that there can be two different tendencies that result in biased news choice ( Garrett, 2009 ; Jang, 2014 ; Schmuck et al., 2020 ): the tendency to approach prejudice-consistent news and the tendency to avoid prejudice-challenging news. Treating the two phenomena as separate is crucial for theory development ( Schmuck et al., 2020 ), because prejudiced individuals can seek prejudice-consistent stereotypical news without avoiding prejudice-challenging counter-stereotypical news, and vice versa. Avoidance-related tendencies have often been explained by cognitive dissonance theory ( Festinger, 1957 ): Prejudiced individuals avoid prejudice-challenging counter-stereotypical news as they may anticipate emotional discomfort resulting from exposure. Conversely, the self-consistency idea, as theorized in the Selective Exposure Self- and Affect-Management (SESAM) model ( Knobloch-Westerwick, 2015 ), conceptualizes news choice as an outcome of a general self-management toward consistency. Of note, this is a broader theoretical idea that can explain both avoidance and approach tendencies. Unfortunately, approach and avoidance tendencies “have received scant research attention thus far” ( Schmuck et al., 2020 , p. 158, see also Garrett, 2009 ). Previous media research, largely stemming from the political communication literature, provides evidence for the relevance of both ( Garrett, 2009 ; Jang, 2014 ; Schmuck et al., 2020 ). However, there is clearly a knowledge gap, especially in the media stereotype domain.

Preference-based reinforcement

As a second general idea, we assumed that exposure to (self-selected) media content can elicit a (reinforcement) effect on prejudice. As already noted, there is evidence from forced-exposure experiments for both the detrimental effects of prejudice-consistent depictions and the beneficial effects of prejudice-challenging depictions ( Holt, 2013 ; Mastro & Tukachinsky, 2011 ; Ramasubramanian, 2011 ; Saleem et al., 2017 )—with known limitations regarding interpretations of actual societal effects, especially in the context of high-choice media environments (see above). Importantly, the combination of both general ideas—prejudice-based selective exposure and (reinforcement) effects—forms the process we label as preference-based reinforcement : Prejudiced individuals prefer news that is congenial to their prejudiced views, and exposure to self-selected (prejudice-consistent) media content, in turn, can reinforce their prejudiced views. The strength of preference-based reinforcement, conceptualized as a process occurring over time, can thus depend on both the strength of the prejudice-based selective exposure and the (reinforcement) effect of (self-selected) exposure. Thus, neither a selective-exposure study nor a forced-exposure experiment used in isolation allow for a thorough understanding of the effect process. Evidence for preference-based reinforcement in the media stereotype domain is pending.

Additional theorizing on actual societal effects: net-effect perspective

Due to the high societal relevance of phenomena such as racism or sexism, we especially aimed to focus our attention on the interpretations of effects obtained in individual studies. We highlighted the question about actual societal effects: What is the net effect of both prejudice-based selective exposure and (reinforcement) effects? Our working hypothesis was that it can lead to different interpretations of actual societal effects (i.e., total effects occurring outside the “lab”) if exposure is operationalized as forced or self-selected: Even if a forced-exposure experiment shows beneficial effects from prejudice-challenging counter-stereotypical news, estimates of actual societal effects are questionable when not additionally assessing whether prejudiced individuals would actually read this content in their everyday life. The net-effect perspective considers both in conjunction: selective exposure and effects. We tentatively assumed that an estimate of actual societal effects can be ascertained best when forced-exposure and self-selected-exposure designs are used in combination . The strength of a combined use has already been acknowledged in other fields, such as political science ( Arceneaux & Johnson, 2013 ; De Benedictis-Kessner et al., 2019 ), political communication ( Stroud et al., 2019 ), psychology ( Johnston, 1996 ; Johnston & Macrae, 1994 ), and media psychology ( Dahlgren, 2021 ). An investigation in the media stereotyping domain is still pending. We provide more details on the net-effect perspective below.

We now outline our formal hypotheses [Hs] and research questions (RQs) that are based on the theorizing outlined above. Study 1 is a selective-exposure study, aimed to test the hypothesis that prejudice predicts news choice (H1). Furthermore, we asked about the role of approach and avoidance tendencies (RQ1). Study 2 used a forced-exposure experiment and we predicted that exposure to prejudice-challenging counter-stereotypical news would reduce prejudice (H2.1), whereas exposure to prejudice-consistent stereotypical news would increase prejudice (H2.2). Whereas studies 1 and 2 echoed the separate strands of previous research—i.e., selective-exposure and forced-exposure effects—,study 3 combined both strands by investigating the dynamic of self-reinforcing effects. We hypothesized a pattern consistent with preference-based reinforcement, that is, a reinforcement effect elicited by exposure to the self-selected content (H3). In addition, and as outlined in detail below, we aimed to estimate actual societal effects in an integrated analysis of the data of both forced (study 2) and self-selected (study 3) exposure (net-effect perspective). RQ2 asked whether estimates of actual societal effects would be different when operationalizing exposure as forced or self-selected. Finally, we conducted a replication study also using a net-effect perspective (study 4).

The empirical work was grounded in the stereotype of the “criminal Afghan asylum seeker,” which is relevant in Austria where the studies were conducted. The focus on this target stereotype seemed appropriate due to several reasons. On a more general level, the study of media stereotypes of refugees is an important avenue in research (Billings & Parrot, 2020). More specifically, there seems to be a widely shared antipathy toward Afghan asylum seekers in Austria (and other European countries), presumably stimulated by the “European refugee crisis” in 2015 and (the news reporting of) negatively valenced key events about criminal Afghan asylum seekers in the aftermath, including many stories of sexual harassment and rape (see Kohlbacher et al., 2020 ). Of note, a breeding ground for prejudice toward Afghan asylum seekers—the majority of whom are Muslim—may also be a broader anti-Muslim prejudice ( Strabac & Listhaug, 2008 ). Furthermore, news coverage on this social category is highly stereotypical, as a content analysis of Austrian newspapers showed ( Steiger, 2021 ). This content analysis found that “on every other day,” a (negatively valenced) prejudice-consistent stereotypical news article appeared. News, as was argued, made “Afghans dangerous, aggressive and criminal people.” Importantly, prejudice-consistent articles tended to be highly arousing, including emotional words and threatening visuals, and they frequently included crimes such as rape. Conversely, prejudice-challenging counter-stereotypical articles were rare. If published, these articles had a positive valence, and it was “striking that the positive reporting [was] often limited to their ability to integrate” (i.e., positive role-model stories of successful integration). These articles tended to be more pallid compared with prejudice-consistent articles. Of note, in these prejudice-challenging articles, Afghans got the chance to speak out and thus to be subjects in the news coverage rather than merely being objects of (negative crime) coverage. This content-analytic evidence guided the development of the stimulus materials in the empirical work, aiming to ensure high external validity.

We collected data for the first three studies simultaneously, and randomly allocated each participant ( N  =   2,695 participants across the three studies) to one of the three studies. This combined data-collection approach did not allow us to adjust subsequent studies to the results of a preceding one. However, this was not necessary for tests of our a priori hypotheses. The random allocation procedure ensured that the samples were comparable. The latter is relevant for our integrated analysis combining data from studies 2 and 3 (i.e., net-effect perspective). Study 4 ( N  =   937) is a replication study using a net-effect perspective and data were collected after the first three studies. Note that we bought all samples from a commercial market research company, which is well known in Austria and has experience in conducting comparable web-based studies. We used age, gender, and education as the quota variables. The samples for all four studies roughly corresponded to the Austrian population in terms of the quota variables.

Ethics statement

The IRB-COM of the Department of Communication, University of Vienna, approved all four empirical studies (studies 1–3 = Number ID: 20211202086, dated December 3, 2021; study 4 = Number ID: 20221024055, dated October 26, 2022).

The aim of study 1 was to test for prejudice-based selective exposure (H1). We also investigated the role of approach and avoidance tendencies (RQ1). We conducted a large web-based survey with a sample from the Austrian general population ( N  =   1,166) based on quota-sampling techniques (age, gender, education). News choice was operationalized with a headline-choice task that had already been used in previous research (e.g., Galdi et al., 2012 ). In each individual trial, participants were presented with two headlines and were asked to choose the one that they preferred reading. We used headlines to measure prejudice-based selective exposure because this content dimension is relevant across different media outlets and genres.

Male (52.1%) and female respondents (47.6%) were nearly equally represented in our sample; three participants chose the “other” option. The majority had no high school diploma (64.9%), one-quarter had a high school diploma (23.9%), and the minority had a university degree (11.2%). Participants ranged in age between 18 and 92 years ( M  =   47.74, SD  =   16.05).

News choice: headline-choice task

News choice was measured with 10 target trials. In each trial, participants were presented with two headlines and were asked to choose one of them. Participants were randomly allocated to one of three experimental conditions with a specific variation in the target news-choice trials, consistent with our aim of investigating the role of approach and avoidance tendencies: In the first condition ( pos/neg , n  =   393), participants could decide between a (positively valenced) prejudice-challenging counter-stereotypical (coded as 0) and a (negatively valenced) prejudice-consistent stereotypical (coded as 1) headline: Do those who show more prejudice toward Afghan asylum seekers prefer to choose the prejudice-consistent depiction of Afghan asylum seekers compared with prejudice-challenging depictions? We saw this as the primary condition to test for prejudice-based selective exposure, as these trials provided both prejudice-consistent stereotypical and prejudice-challenging counter-stereotypical depictions within each news-choice trial. This allowed for a straightforward test of H1, which posited that prejudice predicts news choice. Higher values on the sum score of the 10 target trials indicate a greater tendency to select prejudice-consistent news relative to prejudice-challenging news ( M  =   4.91, SD  =   3.24, range   =   0–10).

The remaining two variants of this task were used to address RQ1 on the role of approach and avoidance tendencies. In the second condition, participants decided between a positively valenced prejudice-challenging (coded as 0) and a neutral (coded as 1) headline ( pos/neutral , n  =   384): Does prejudice toward Afghan asylum seekers hinder the selection of prejudice-challenging news? For prejudiced individuals, higher values indicate a greater tendency to avoid prejudice-challenging news ( M  =   6.03, SD  =   2.57, range   =   0–10). In the third condition, participants decided between a neutral (coded as 0) and a negatively valenced prejudice-consistent (coded as 1) headline ( neutral/neg , n  =   389): Does prejudice toward Afghan asylum seekers facilitate the approach toward prejudice-consistent stereotypical news? Higher values indicate a greater tendency to approach prejudice-consistent news ( M  =   4.50, SD  =   2.837, range   =   0–10).

We used the same set of 10 prejudice-consistent (e.g., “Fatal rape: Afghan asylum seekers abuse a girl”), prejudice-challenging (e.g., “Successful integration: Positive stories from Afghan asylum seekers”), and neutral headlines (e.g., “We need water from morning to evening”) across the three news-choice conditions.

By randomly pairing these headlines, we created pairs of headlines. These pairs were consequently used for all participants. Note that these headlines were based on real headlines from actual news coverage in the news archives identified in our research. This aimed to ensure high external validity. Importantly, the depiction of Afghan asylum seekers in our headlines corresponds to findings of a recent content analysis ( Steiger, 2021 ; see above). Thus, we acknowledged that prejudice-consistent (prejudice-challenging) headlines provided negatively (positively) valenced depictions of Afghan asylum seekers and tended to be more (less) arousing. The latter is consistent with basic psychological research showing that the relationship between valence and arousal is asymmetrical in that negatively valenced concepts tend to elicit higher arousal ratings than positively valenced concepts do ( Võ et al., 2009 ). For the prejudice-consistent and prejudice-challenging headlines, it was not seen as being appropriate to match stereotypical and counter-stereotypical content in terms of emotional arousal (see Steiger, 2021 ). We emphasized external validity.

Given that this decision may raise internal validity concerns, we decided to measure a given participant’s preference for valenced and arousing content, and used these two predispositions to news choice as covariates in the analysis: A total of 10 additional headline-choice trials were used that had already been used in previous studies ( Arendt et al., 2016 , 2019 ). Afghan asylum seekers were not mentioned in these choice trials. First, we measured the tendency to read positively valenced (e.g., “The sun and the warm temperature make people feel happy,” coded as 1) compared with negatively valenced (e.g., “Meteorological disturbance causes massive damage,” coded as 0) headlines. We used five choice trials, and higher values indicate a stronger preference for positively valenced headlines ( M  =   3.01, SD  =   1.46, range   =   0–5). Second, we measured the tendency to read emotionally arousing articles. We used less arousing (e.g., “Dog bites a child on the leg,” coded as 0) and more arousing (e.g., “Aggressive fighting dog bites poor child,” coded as 1) headlines. Again, we used five trials, and higher values indicate a stronger preference for emotionally arousing headlines ( M  =   1.84, SD  =   1.39, range =   0–5). The full list of headlines can be found in the Online Supplementary Material (OSM) ( Supplementary Table 1 ).

We used 10 items to measure prejudice. Two of them used a 7-point bipolar scale to measure general negative affective reactions toward Afghan asylum seekers (i.e., whether participants see them as good or bad , positive or negative ). The remaining eight items asked participants to rate each of eight statements (e.g., “Asylum seekers from Afghanistan enrich life and society in Austria,” “Afghan asylum seekers are more prone to violence than others”) on a 7-point scale ranging from strongly disagree (coded as 1) to strongly agree (coded as 7). All statements pointing to a favorable attitude were reverse coded. Higher values indicate more prejudice ( M  =   4.92, SD  =   1.35, α = .92). A factor analysis confirmed a one-factor solution.

Statistical analysis

We relied on hierarchical multiple regression models and predicted the news-choice score by age, gender, dummy-coded education, preference for positively valenced news, preference for emotionally arousing news (all in step 1), and prejudice (step 2). The change in R 2 in the second step assesses whether prejudice predicts news choice over and above the influence of controls. Given that only three participants chose the “other” gender option and we wanted to include gender as a control—dummy-coding is not appropriate for this low number of participants choosing the other option, we report on the analysis without participants choosing the “other” option. However, models without gender as a control and including these three participants provided very similar results. Zero-order correlations can be found in the OSM ( Supplementary Table 2 ).

To conduct a formal test of moderation , that is, a test of whether effect coefficients indicative of prejudice-based selective exposure significantly differed between the three conditions (i.e., pos/neg, pos/neutral, and neutral/neg), we conducted a multigroup analysis by using the structural equation modeling software AMOS. We defined path models ( df  =   0), as noted in the OLS regressions reported above. Importantly, we examined the fit of the model when freely estimating the effect coefficients “prejudice => news choice” relative to a model in which this coefficient was constrained to be equal in all three conditions [see Hayes et al. (2013) for this procedure]. We were interested in whether constraining this coefficient resulted in a decrement in fit, a pattern that would be consistent with a moderation effect, answering RQ1.

Results and discussion

H1 predicted prejudice-based selective exposure and RQ1 questioned the influence of approach and avoidance tendencies. We only report the coefficients of prejudice below. However, full models can be found in the OSM ( Supplementary Tables 3–5 ).

When participants were asked to choose between prejudice-challenging and prejudice-consistent headlines ( pos/neg condition ), prejudice predicted news choice, Δ F (1, 381) = 291.17, Δ R 2 = 0.358, p < .001. The analysis indicated a very strong effect coefficient, B  =   1.42, 95% confidence interval (CI) [1.25–1.58], SE  =   0.08, β = 0.61, p < .001. Consistent with H1, the more prejudice a given participant showed, the greater their tendency was to select prejudice-consistent stereotypical news relative to prejudice-challenging counter-stereotypical news.

When participants were asked to choose between a prejudice-challenging and a neutral headline ( pos/neutral condition ), prejudice elicited an effect on news choice as well, Δ F (1, 376) = 150.85, Δ R 2 = .278, p < .001. Although the effect size was descriptively lower compared with the one obtained in the pos/neg condition, prejudice strongly predicted news choice as well, B  =   1.07, 95% CI [0.90–1.24], SE  =   0.09, β = .55, p < .001. The analysis also indicated a significant finding when participants were asked to choose between a neutral and a prejudice-consistent headline ( neutral/neg condition ), Δ F (1, 381) = 50.13, Δ R 2 = 0.090, p < .001. Although the effect coefficient was descriptively smaller, B  =   0.67, 95% CI [0.48–0.85], SE  =   0.09, β = 0.32, p < .001, prejudice still predicted the tendency to select prejudice-consistent news.

A formal test of moderation across the three conditions provided evidence for moderation, Δχ 2 (2) = 35.29, p < .001. This indicated that the strength of prejudice-based selective exposure significantly differed between the three conditions. An additional formal test of moderation analyzing the difference between the pos/neutral condition and the neutral/neg condition provided evidence for moderation, Δχ 2 (2) = 9.99, p = .002. Given that the neutral headlines were identical in these conditions, this difference indicated that prejudice was a stronger predictor in news-choice trials that included prejudice-challenging counter-stereotypical news compared with trials that included prejudice-consistent stereotypical news.

With prejudiced individuals in mind, we can interpret this finding such that prejudiced individuals tend to approach prejudice-consistent news and, to a greater extent, avoid prejudice-challenging news. We focused on prejudiced individuals in the present research, as it is this segment of the population that represents a threat to an open and humane society and thus requires special scholarly attention. However, the specific interpretation of the role of approach or avoidance tendencies depends on the level of prejudiced predispositions. When considering unprejudiced individuals , the interpretation is reversed: The effect of prejudice on news choice in the neutral/neg condition would instead indicate the strength of the avoidance of prejudice-consistent news; conversely, an effect of prejudice on news choice in the pos/neutral condition would indicate an approach toward prejudice-challenging news. Supplementary Figure 1 provides a visualization of this basic idea. Note that we conducted an additional analysis (see OSM for details), showing that, in rather prejudiced individuals, the strength of the avoidance tendency triggered by prejudice-challenging news was significantly stronger compared with the approach tendency triggered by prejudice-consistent news. Conversely, in rather unprejudiced individuals, the strength of the approach tendency triggered by prejudice-challenging news was significantly stronger compared with the avoidance tendency triggered by prejudice-consistent news.

The finding that the pos/neg condition showed an even stronger effect coefficient compared with the other two conditions, Δ χ 2 (2) = 25.30, p < .001, is consistent with the idea that avoidance and approach tendencies both contributed to biased news choice. In a nutshell, the findings indicated that both avoidance and approach tendencies influenced news choice , answering RQ1. Of interest, the findings also indicated that prejudice-challenging counter-stereotypical news triggered a stronger form of prejudice-based news-choice bias compared with prejudice-consistent stereotypical news.

Our strong focus on external validity when creating the dichotomous news-choice trials may raise internal validity concerns. We acknowledge that our strong focus on external validity should not hinder us from providing a detailed assessment of internal validity issues. Thus, stimulated by comments raised within the review process, we conducted two additional empirical studies focusing on the relevant internal validity concerns (for details, see the OSM section “Additional Evidence Related to Internal Validity Concerns”). This evidence indicated that prejudice-based selective exposure was still observable (1) when using news-choice trials that were matched in terms of arousal and valence and (2) when using a design with a strong focus on internal validity that compares effect size estimates between a condition using headlines that include the stereotype-related group concept (e.g., “ Afghan stabbed wife with a Stanley knife”) with a condition using a matched “control” headline without the stereotype-related group concept (“ Man stabbed wife with a Stanley knife”). This additional evidence is consistent with the conclusions drawn from the evidence reported above.

Study 2 tested for the causal effects of forced exposure to prejudice-challenging counter-stereotypical and prejudice-consistent stereotypical news (H2.1 and H2.2), also relevant for RQ2 (see below). This web-based study ( N  =   380) relied on a design with three experimental groups (i.e., prejudice-challenging, control, and prejudice-consistent) with a repeated measurement of prejudice (i.e., before and after forced exposure).

Female (46.2%) and male (53.8%) participants were nearly equally represented. The majority had no high school diploma (62.7%), about a quarter had a high school diploma (24.4%), and a minority had a university degree (12.9%). Participants ranged in age between 18 and 76 years ( M  =   45.44, SD  =   16.00).

Experimental manipulation

Participants were randomly allocated to one of three experimental groups. In the prejudice-consistent stereotype group ( n  =   127), participants read an article in which Afghan asylum seekers were accused of having raped a girl, headlined “Leonie (13) was drugged and raped by Afghans in Vienna in June: She died as a result of these acts.” This article was based on real news content, and such Afghan-related rape stories are frequently present in the Austrian news coverage ( Steiger, 2021 ). Participants allocated to the prejudice-challenging counter-stereotype group ( n  =   127) read an article that provided positive role models, headlined “Many of the refugees from Afghanistan were able to gain a foothold: There are many positive examples of integration.” Again, this article was based on real news content, and such positive role-model stories that run counter to the “criminal Afghan asylum seeker” stereotype are relevant in the Austrian news coverage ( Steiger, 2021 ). The selection of these articles was guided by our aim of ensuring high external validity. In the control group ( n  =   126), participants read an article about the importance of water, headlined “We need water from morning to evening: For drinking, for cooking, for washing and much more.” Efforts were made to hold relevant factors constant. For example, all articles included the same number of pictures (i.e., three) and had a similar text length (i.e., stereotypical article = 956 words, counter-stereotypical article = 888 words, and control article = 944 words). The three articles can be found in the OSM ( Supplementary Appendix I ).

We used the same measure as in study 1. Consistent with recent scholarship that encourages the adoption of repeated measure designs to increase precision ( Clifford et al., 2021 ), we relied on a repeated measures design and administered the prejudice measure before ( M  =   5.00, SD  =   1.47, α = 0.94) and after ( M  =   5.05, SD  =   1.51, α = 0.95) forced exposure.

We relied on a 3 (experimental group: prejudice-challenging, control, prejudice-consistent) × 2 (time of measurement: prejudice measured before and after exposure) mixed analysis of variance. Whereas the experimental group was a between-subjects factor, the time of measurement was a within-subjects factor. A significant effect of exposure would be indicated by a significant interaction (i.e., a change in prejudice over time, depending on the experimental condition: an increase in the prejudice-consistent group and a decrease in the prejudice-challenging group).

H2 predicted the effects of forced exposure to news on prejudice. Specifically, we predicted a reduction in prejudice after exposure to prejudice-challenging news (H2.1) and an increase in prejudice after exposure to prejudice-consistent news (H2.2). The mixed ANOVA provided a significant main effect of time, F (1, 377) = 5.28, p = .022, η 2 = 0.014, and the absence of a main effect of the experimental condition, F (2, 377) = 0.37, p = .694, η 2 = .002. More importantly for the test of the hypotheses, there was a significant interaction effect, F (2, 377) = 26.70, p < .001, η 2 = 0.124. Of interest, whereas forced exposure to prejudice-challenging news reduced prejudice, Δ M =  −0.13, 95% CI [−0.19 to −0.06], SD  =   0.37, t (126) = −3.86, p < .001, exposure to prejudice-consistent news increased prejudice, Δ M =  0.26, 95% CI [0.18–0.34], SD  =   0.46, t (126) = 6.36, p < .001. There was no significant change in prejudice in the control group, Δ M =  0.02, 95% CI [−0.06 to 0.10], SD  =   0.44, t (125) = 0.49, p = .628. Figure 1 provides a visualization.

Effects of forced exposure to prejudice-challenging counter-stereotypical, neutral (control group), and prejudice-consistent stereotypical news on prejudice, measured before and after exposure (study 2).

Notes. The figure shows the means of prejudice for each experimental group, measured before and after exposure. Before–after comparisons ( p values) are based on t -tests, reported in detail in the body of the text: Whereas forced exposure to prejudice-challenging counter-stereotypical news content reduced prejudice, exposure to prejudice-consistent stereotypical content increased prejudice; we did not observe a change in the control group.

Of interest, an additional moderation analysis showed that the effects of prejudice-consistent and prejudice-challenging news were not significantly different in those with low or high values of prejudiced predispositions (see Supplementary Appendix II : Additional Moderation Analysis). Taken together, study 2 provided evidence for substantial effects of forced exposure.

Studies 1 and 2 echoed the separate strands of research on selective exposure and effects. Study 3 combined them by relying on self-selected exposure, allowing us to investigate preference-based reinforcement (H3). In this large web-based study ( N  =   1,149), we measured prejudice, and subsequently asked participants to select one of two news items. In contrast to study 1’s selective-exposure design (where the task consisted solely of selecting headlines but did not involve actual exposure to the story selected), participants read the selected article. Afterward, we measured prejudice again. This design is similar to those used in other fields, such as political science ( Arceneaux & Johnson, 2013 ), psychology ( Johnston & Macrae, 1994 ), political communication ( Stroud et al., 2019 ), or media psychology ( Dahlgren, 2021 ).

Consistent with the operationalization of news choice in study 1, we used three news-choice conditions to which participants were randomly allocated. Participants were given the opportunity to choose between the same articles used in study 2. Had we used a choice task including all three article options, the chosen option would have had to have been interpreted relative to the other two. A dichotomous task allows for more specific interpretations because there is only one other option (but see study 4). In the first condition ( pos/neg condition , n  =   386), participants could decide between a prejudice-challenging counter-stereotypical (“Integration success story: Positive stories from Afghans,” coded as 0) and a prejudice-consistent stereotypical article (“Fatal rape: Afghans abuse a girl,” coded as 1). The majority (81.1%) chose the prejudice-challenging article. In the second condition ( pos/neutral condition , n  =   381), participants could decide between a prejudice-challenging (coded as 0) and a neutral (“We need water from morning to evening,” coded as 1) article. Approximately half of the participants (45.1%) chose the prejudice-challenging article. In the third condition ( neutral/neg condition , n  =   382), participants decided between a neutral and a prejudice-consistent article. Only a minority (23.3%) chose the prejudice-consistent article.

Male (52.7%) and female (47.3%) respondents were nearly equally represented in our sample; one participant chose the “other” option (0.1%). The majority had no high school diploma (63.3%), about a quarter had a high school diploma (24.4%), and a minority had a university degree (12.4%). Participants ranged in age between 18 and 83 years ( M  =   46.13, SD  =   15.49).

We used the same measure as in studies 1 and 2. We administered the measure before ( M  =   4.96, SD  =   1.42, α = 0.93) and after ( M  =   4.92, SD  =   1.48, α = 0.94) self-selected exposure.

Predispositions in news choice

Consistent with study 1, we controlled for the participant’s preference for valenced ( M  =   3.20, SD  =   1.47) and arousing ( M  =   1.65, SD  =   1.40) content.

We used a two-step approach. In a first step, we ran (a) three separate hierarchical binary logistic regression models to predict the dichotomous news choice by prejudice measured before exposure, separately for each of the three conditions (i.e., pos/neg, pos/neutral, and neutral/neg). We controlled for age, gender, education, preference for positively valenced news, and preference for emotionally arousing news. Next, we (b) ran three separate hierarchical multiple regression models, predicting prejudice measured after exposure by prejudice measured before exposure (to control for autoregressive effects), controls (all in the first step), and news choice (i.e., self-selected exposure; second step). The latter model indicates whether self-selected exposure predicted a change in prejudice over time. In a second step, we used the structural equation modeling software AMOS to test for prejudice-based reinforcement by specifying a mediator model (independent variable = pre-prejudice; mediator = self-selected exposure; dependent variable = post-prejudice). Controls were included as well. This analysis allowed us to estimate the indirect effect (i.e., pre-prejudice’s effect on post-prejudice via self-selected exposure). We simultaneously estimated three path models ( df  =   0), one for each condition (i.e., pos/neg, pos/neutral, and neutral/neg). Figure 2 provides a visualization.

Conceptual model of preference-based reinforcement (study 3).

Notes. We used the structural equation modeling software AMOS to estimate three mediator models ( df = 0), one for each condition (pos/neg, pos/neutral, and neutral/negative). The figure visualizes the direct effects (solid arrows) and the indirect effect (dashed arrow). Age, gender, dummy-coded education, and predispositions in news choice (i.e., preference for positively valenced news, preference for emotionally arousing news) were used as controls. See the body of the text for effect estimates.

H3 predicted a pattern consistent with preference-based reinforcement. We report on the conceptual variables below. Full models can be found in the OSM ( Supplementary Tables 6–11 ).

Pos/Neg condition

We started by looking at participants who chose between the prejudice-challenging and prejudice-consistent articles. First, prejudice measured before exposure predicted news choice (i.e., increased the likelihood of selecting the prejudice-consistent article relative to the prejudice-challenging article), B  =   0.57, SE  =   0.13, Wald   =   21.08, df  =   1, odds ratio   =   1.77, 95% CI [1.39–2.26], p < .001. Self-selected exposure to the prejudice-consistent article (relative to the prejudice-challenging article), in turn, predicted an increase in prejudice measured post-exposure, B  =   0.29, 95% CI [0.16–0.42], SE = 0.07, β = 0.08, t  =   4.31, p < .001. Second, our mediation analysis showed an (unstandardized) indirect effect, coeff   =   0.019, 95% CI [0.011–0.031], p = .001. These findings are consistent with preference-based reinforcement and thus support H3.

Pos/neutral condition

Similar findings could be observed in participants who selected between the prejudice-challenging article and the neutral article. First, prejudice measured before self-selected exposure predicted news choice (i.e., increased the likelihood of avoiding the prejudice-challenging article), B  =   0.80, SE  =   0.10, Wald   =   67.62, df  =   1, odds ratio   =   2.23, 95% CI [1.84–2.70], p < .001. Self-selected exposure to the prejudice-challenging article (relative to the neutral article), in turn, predicted a decrease in prejudice measured post-exposure, B  =   0.34, 95% CI [0.23–0.44], SE = 0.05, β = 0.11, t  =   6.37, p < .001. (Note that the reported effect coefficient has a positive sign due to our codes: prejudice-challenging article = 0, neutral article = 1.) We also obtained a significant indirect effect, coeff   =   0.054, 95% CI [0.039–0.072], p = .002.

Neutral/Neg condition

A slightly different pattern was obtained when analyzing data provided by participants who selected between the neutral and the prejudice-consistent article. Prejudice measured before self-selected exposure did not significantly predict news choice in the logistic regression model (albeit that the effect coefficient points in the predicted direction), B  =   0.18, SE  =   0.11, Wald   =   2.63, df  =   1, odds ratio   =   1.20, 95% CI [0.96–1.49], p = .105. However, self-selected exposure to the prejudice-consistent article (relative to the neutral article) predicted an increase in prejudice measured post-exposure, B  =   0.29, 95% CI [0.16–0.43], SE = 0.06, β = 0.09, t  =   4.58, p < .001. Our mediation analysis showed a significant indirect effect, coeff   =   0.007, 95% CI [0.001–0.015], p = .0497. Although the indirect effect was significant in a statistical sense, the effect size obtained seemed to be weaker on a descriptive level compared with the other two conditions.

Formal test of moderation

To conduct a formal test of moderation (i.e., a test of whether effect coefficients indicative for preference-based reinforcement significantly differed between the three conditions), we conducted a multigroup analysis, as already reported above (see study 1). We tested whether the effect of pre-prejudice on post-prejudice through its influence on news-choice/self-selected exposure significantly differed among the three choice conditions (i.e., pos/neg, pos/neutral, and neutral/negative). We defined path models, as reported in the OLS regressions reported above. Importantly, we examined the fit of the model when freely estimating the effect coefficients “pre-prejudice => self-selected exposure” and “self-selected exposure => post-prejudice” (see Figure 2 ) relative to a model in which these effect coefficients were constrained to be equal in all three conditions. We were interested in whether constraining these effect coefficients indicative for preference-based reinforcement resulted in a decrement in fit, a pattern that would be consistent with a moderation effect. Indeed, we found evidence for moderation, Δ χ 2 (2) = 41.74, p < .001.

Of note, the effect coefficients reported above indicated that the difference between the three choice conditions was strongly observable for selective exposure (“pre-prejudice => self-selected exposure,” see the logistic regression models reported above). Conversely, the sizes of the reinforcement effects were very similar among the three conditions (“self-selected exposure => post-prejudice,” see the multiple regression models reported above). Therefore, we re-ran the formal moderation analysis reported above two additional times. In the first additional model, we only constrained the selective-exposure path (i.e., “pre-prejudice => self-selected exposure”) and in the second additional model, we only constrained the reinforcement path (i.e., “self-selected exposure => post-prejudice”). Whereas the first model provided evidence for moderation, Δ χ 2 (2) = 41.28, p < .001, the second model did not, Δ χ 2 (2) = 0.46, p = .794. This indicated that the difference in preference-based reinforcement obtained was driven by differences in the strength of prejudice-based selective exposure, emphasizing the crucial role played by audience selectivity.

As reported above, the coefficients (expressed as odds ratios ) were different on a descriptive level: 1.77 (pos/neg), 2.23 (pos/neutral), and 1.20 (neutral/neg). Whereas the first two were significant ( p ’s < .001), the latter was not ( p = .105). Thus, it appeared that the conditions which included prejudice-challenging counter-stereotypical news elicited a stronger bias during selective exposure. We conducted a formal test of moderation by examining the fit of the model when freely estimating the effect coefficient “pre-prejudice => self-selected exposure” relative to a model in which this coefficient was constrained (i.e., pos/neg and pos/neutral versus neutral/neg). This analysis indicated that prejudice especially predicted news choice when prejudice-challenging counter-stereotypical news was included as a choice option, Δ χ 2 (1) = 21.52, p < .001. This is consistent with the findings from study 1.

We now present an integrated analysis, questioning whether the interpretation of actual societal effects depends on the exposure paradigm used (RQ2): Even if a forced-exposure experiment indicates, for example, detrimental effects of prejudice-consistent stereotypical news, we do not know whether individuals would actually read this kind of content in their everyday life. Answering both the question of (1) whether individuals select certain types of content and of (2) whether exposure elicits effects was assumed to be helpful for estimating actual societal effects. The net-effect perspective focuses on these two questions in conjunction. Using the net-effect perspective, we operationalized the “societal effect” as the pre–post change in the sample mean in prejudice when looking at the data for the whole samples from studies 2 and 3, respectively.

We based the integrated analysis on the following observations: Study 2 showed that forced exposure to the prejudice-consistent article elicited a detrimental increase in prejudice (Δ M  =   0.26), and forced exposure to the prejudice-challenging article elicited a beneficial reduction in prejudice (Δ M  =   −0.13). Note that an equally high number of individuals (i.e., 33%) read the prejudice-consistent, neutral, and prejudice-challenging articles in study 2’s forced-exposure experiment due to random assignment. Conversely, in study 3’s self-selected-exposure paradigm, many participants did not actually read the prejudice-consistent stereotypical article when they were able to freely choose among options: Descriptive statistics, calculated over all three of study 3’s choice conditions, indicated that 42% selected the prejudice-challenging counter-stereotypical article compared with only 14% who selected the prejudice-consistent stereotypical article. (The remaining participants selected the neutral article.) Note that in the case of the absence of any selection bias, one would expect a selection likelihood of 33% for each of the three articles in both studies 2 and 3. This is despite the fact that the designs of both studies differ (see Supplementary Appendix III for details). Given that the prejudice-challenging article elicited a beneficial effect and more people actually chose it in study 3, we speculated that the net effect on prejudice in the self-selected-exposure paradigm (study 3) would be more beneficial compared with the effect observed in the forced-exposure paradigm (study 2). As a post hoc hypothesis, we hypothesized an interaction effect insofar as the pre–post change in the sample mean in prejudice depended on the paradigm utilized (i.e., forced vs. self-selected exposure).

This “macro-level” net-effect perspective thus focuses on “societal” effects (i.e., pre–post changes in prejudice observed for the whole samples from studies 2 and 3, simulating prejudice levels in two different “societies”)—in contrast to the individual-level analysis reported in study 2’s and study 3’s sections. Therefore, we now utilize a bird’s eye view: We looked at the change in prejudice (sample means) measured before and after exposure within each “society.” Of note, we did not include (experimental) news-exposure factors in this analysis but only looked at prejudice, simulating a “macro-level” view of these two “societies”: Did these “societies” change in prejudice depending on whether their “citizens” were forced to read a given news item or could freely choose to select their preferred one?

We relied on a two-way 2 (type of exposure: forced or self-selected exposure, i.e., data from study 2 or 3) × 2 (prejudice measured before and after exposure) mixed analysis of variance to determine whether the strength (and direction) of the “societal effect” depended on the type of exposure (forced versus self-selected), which would be indicated by a significant interaction. Indeed, the mixed ANOVA provided a significant interaction, F (1, 1525) = 10.12, p = .001, η 2 = 0.007. Figure 3A provides a visualization. In the “society” in which people could freely choose their news—only 14% read the prejudice-consistent stereotypical news but a total of 42% read the prejudice-challenging counter-stereotypical news—, there was a net decrease in prejudice , t (1146) = −2.78, p = .006. Conversely, in the “society” whose citizens were forced to read an article—a total of 33% read the prejudice-consistent and prejudice-challenging news, respectively—, we did not observe a comparable beneficial change. Conversely, we even obtained a net increase in prejudice , t (379) = 2.16, p = .031.

Net-effect perspective: differences in estimated “societal effects.” (A) Merged data (studies 2 and 3). (B) Replication data (study 4).

Notes. This figure shows the sample means of prejudice measured before and after exposure when relying on the forced-exposure paradigm versus the self-selected-exposure paradigm. This analysis aims to simulate a “macro-level” view of these two “societies”: Did these “societies” change in prejudice depending on whether their “citizens” were forced to read a given news item or could freely choose to select their preferred one? Panel (A) visualizes an integrated analysis based on data from the samples of study 2 (forced-exposure paradigm) and study 3 (self-selected-exposure paradigm). Whereas the forced-exposure paradigm indicated a net increase in prejudice due to media exposure, the self-selected-exposure paradigm indicated a net reduction. Panel (B) visualizes data from the replication study (study 4), replicating the observed effect pattern. Evidence indicates that it can make a fundamental difference for the interpretation of actual societal effects if the findings are based either on the forced- or on the self-selected-exposure paradigm. The findings emphasize the key role played by audience selectivity when studying media effects.

These findings are supportive of the claim that it can make a fundamental difference for the interpretation of actual societal effects depending on whether the findings are based on the forced- or self-selected-exposure paradigm. In fact, estimates of actual societal effects even moved in the opposite direction. This finding deserves scholarly attention: In the forced-exposure paradigm, researchers may tend to focus their attention on the detrimental effect of prejudice-consistent stereotypical news, possibly coming to a rather pessimistic view on societal effects. However, when also considering findings from the self-selected-exposure paradigm (and thus considering audience selectivity), researchers may come to a less pessimistic view, as the majority of individuals simply avoided exposure to the prejudice-consistent stereotypical article. This emphasizes the key role played by audience selectivity when interpreting findings related to media effects phenomena observed in individual studies.

Given that the designs of studies 2 and 3 were different (i.e., a different number of news-choice options: three options in study 2’s forced-exposure design, corresponding to the three experimental groups, versus three separate dichotomous choice trials in study 3’ self-selected-exposure design; see Supplementary Appendix III for a detailed discussion), a replication study was conducted. The aim was to replicate the observed net-effect pattern reported above (see Figure 3A ) using a matched design.

Participants ( N  =   937) of a quota-based sample (age, gender, education) who did not participate in the other three studies were randomly allocated to a forced-exposure condition ( n  =   472) or a self-selected-exposure condition ( n  =   465). In the forced-exposure condition, participants were randomly allocated to read the prejudice-consistent (33.9%), prejudice-challenging (33.7%), or the control (32.4%) article. In the self-selected-exposure condition, participants could select one of the three articles within one news-choice trial that included all three options. More participants selected the prejudice-challenging counter-stereotypical article (34.8%) compared with the prejudice-consistent stereotypical article (18.9%); the control article was selected by 46.2%.

Consistent with the integrated analysis of the merged data from studies 2 and 3 reported above, we relied on a two-way 2 (type of exposure: forced- or self-selected-exposure group) × 2 (prejudice measured before and after exposure) mixed ANOVA. Whether the strength (and direction) of the “societal effect” on prejudice depended on the type of exposure would be indicated by a significant interaction effect. Indeed, the mixed ANOVA produced a significant interaction effect, F (1, 935) = 8.60, p = .003, η 2 = 0.009. Figure 3B provides a visualization. Study 4 thus replicated the pattern observed in the merged analysis of data provided by studies 2 and 3 (see Figure 3A ). The findings from study 4 are thus also supportive of the claim that it can make a fundamental difference for the interpretation of actual societal effects if the findings are based either on the forced- or on the self-selected-exposure paradigm—also emphasizing the key role played by audience selectivity when studying the effects of exposure to media stereotypes.

The present research investigated the media stereotype effect process. In a nutshell, prejudiced predispositions influenced prejudice-consistent news choice and (self-selected) exposure, in turn, elicited a (reinforcement) effect on prejudice. The findings highlight the importance of considering audience selectivity. The study of preference-based reinforcement and thus the dynamic of self-reinforcing effects is, to the best of our knowledge, unprecedented in the research on media stereotypes—our primary theoretical contribution to the media stereotype literature. The findings are also important for recent theorizing on media effects in general that integrates audience selectivity into the effects perspective ( Cacciatore et al., 2016 ; Knobloch-Westerwick, 2015 ; Slater, 2007 )—we provide evidence from the media stereotype domain.

We also investigated the role of approach and avoidance tendencies, providing a theoretical contribution to the selective-exposure literature ( Garrett, 2009 ; Jang, 2014 ; Schmuck et al., 2020 ): Both tendencies are of theoretical importance, because individuals can seek prejudice-consistent stereotypical news without avoiding prejudice-challenging counter-stereotypical news, and vice versa. Evidence indicated that both tendencies played a significant role. Of note, study 1 found that participants preferred prejudice-consistent stereotypical news over neutral news . This challenges cognitive dissonance theory ( Festinger, 1957 ) as the only theoretical explanation for biased selective-exposure behavior, because neutral news should not elicit cognitive dissonance. Conversely, a self-consistency motive, as theorized in the SESAM model ( Knobloch-Westerwick, 2015 ), can explain both tendencies, including the finding from study 1 mentioned above. Of note, Schmuck et al. (2020) provided a similar finding in a different selective-exposure domain (political advertising).

The present research also offers an important methodological contribution for the practice of (future) research in the media stereotype domain: An integrated analysis, relying on a net-effect perspective, used data provided by both the self-selected-exposure and forced-exposure paradigms. Although forced exposure to prejudice-consistent stereotypical news elicited a detrimental effect on the increase in prejudice (study 2), many participants simply avoided reading this prejudice-consistent stereotypical news (study 3). Estimates of “societal effects” were different, depending on whether participants were forced to read given news content or could freely choose to select their preferred article. This observed pattern was replicated (study 4). Note that the net-effect perspective aimed at a more realistic estimation of actual societal effects. However, we are very careful when interpreting real-world implications, as, for example, the news-choice trials including the used headlines were highly specific (and thus still somewhat artificial) when compared with real-world news consumption. Both the forced- and the self-selected-exposure paradigms must still be seen as somewhat simplified compared with the complex media environment in the “world outside.” Therefore, even the self-selected-exposure paradigm cannot be viewed as a natural simulation of the “world outside.” Nevertheless, the commonly used forced-exposure paradigm has important limitations because it clearly fails to reflect the reality where much of the exposure is the result of audience selectivity. The self-selected-exposure paradigm places more attention on this fact. Based on our findings, we argue that scholars should think about increasingly using the self-selected-exposure paradigm as a supplement to forced-exposure experiments. Despite its limitations, when used in combination , both may allow for a better estimation of the real-world impact of media stereotypes. Importantly, we want to emphasize that we do not argue that the self-selected-exposure paradigm is more important than the forced-exposure paradigm or that forced-exposure experiments are useless. Conversely, the forced-exposure paradigm has several strengths, such as it allowing for high confidence in terms of causal interpretations. In addition, some media content in the “world outside” may be seen, at least partly, as “forced”: Individuals using social networking sites may be exposed to content they have chosen (self-selected exposure) and content they have not chosen (as in the forced-exposure paradigm; see Dahlgren, 2021 ). Thus, both paradigms are relevant.

Limitations and promising avenues for future research

We focused on prejudice-based selective exposure, theoretically strongly determined by a self-consistency motive, as outlined by Knobloch-Westerwick (2015) . However, her theorizing also emphasized self-enhancement and self-improvement motives. These additional motives may also play a role. In fact, Knobloch-Westerwick (2015) argued that when threats to the self are salient and negative affect is lingering, news choice will be especially driven by a self-enhancement motive insofar as individuals will then tend to select news content that they perceive will bolster their self. Consider social identity theory ( Tajfel & Turner, 1986 ): Individuals tend to categorize themselves and other salient social categories and their members into “us” versus “them.” A future study could test whether prejudice-based selective exposure is stronger when threats to the self are salient, or when a news user’s different social identities are primed. For example, does the strength of prejudice-based selective exposure increase when one’s own (ingroup) identity has been primed? In addition, Knobloch-Westerwick (2015) argued that when a need or an opportunity to advance one’s own performance and adaptation to the environment becomes salient, news choice will be especially driven by a self-improvement motive. In fact, Knobloch-Westerwick and Kleinman (2012) showed that information utility can override a confirmation bias. These individuals may still distance themselves from the presented view while still taking in the information. These are theoretically fruitful avenues worthy of future study.

A further limitation of the current work is that we only studied short-term effects. In fact, we did not use a temporally more distant follow-up measurement of prejudice to see whether there was a decay in effect size. The strength of the decay may depend on whether exposure was self-selected or forced: Are effects more enduring when news is self-selected? Relatedly, we only used a “one-shot” exposure treatment. Thus, we cannot say whether self-reinforcing effects change depending on the amount of cumulative exposure (see Arendt, 2015 ). It would therefore seem fruitful for a future study to use repeated sessions in which individuals can select news, for example, within consecutive days or weeks. Also relatedly, all participants in studies 2 and 3 were exposed either to prejudice-consistent stereotypical, prejudice-challenging counter-stereotypical, or neutral news content. Thus, we did not investigate the effect elicited by reading both prejudice-consistent and prejudice-challenging news. A future study could test the effects of such “competitive media stereotype environments.” We may find that there is no substantial (or only a slight) pre–post net change in prejudice when participants are exposed to prejudice-consistent stereotypical and prejudice-challenging counter-stereotypical news content under forced exposure. The effects of the articles may (more or less) cancel each other out. However, this may be different when using a self-selection paradigm. Consider prejudiced individuals. They could be asked to select between a prejudice-consistent and a prejudice-challenging article in a first step. Next, they could read the selected article. Afterward, they could be asked to read the second article. It is likely that the participants would rely on some form of motivated reasoning ( Kunda, 1990 ). Such a study may find a substantial effect arising from reading both articles in favor of the one they have selected.

Conclusions

Despite these limitations, the present project provides supporting evidence for the idea of preference-based reinforcement in the media stereotyping domain. Across four empirical studies, we studied the dynamic of self-reinforcing effects. Audience selectivity was a key factor. Future research could think about operationalizing exposure as forced and self-selected.

Supplementary material is available online at Journal of Communication online.

Conflicts of interest : None declared.

Research materials, including stimulus materials and data, can be requested by the author. Please, contact the author.

Allport G. W. ( 1954 ). The nature of prejudice . Addison-Wesley.

Google Scholar

Google Preview

Arendt F. ( 2013 ). Dose-dependent media priming effects of stereotypic newspaper articles on implicit and explicit stereotypes . Journal of Communication , 63 ( 5 ), 830 – 851 . https://doi.org/10.1111/jcom.12056

Arendt F. , Northup T. ( 2015 ). Effects of long-term exposure to news stereotypes on implicit and explicit attitudes . International Journal of Communication , 9 , 732 – 751 . Retrieved from https://ijoc.org/index.php/ijoc/article/view/2691

Arendt F. ( 2015 ). Toward a dose-response account of media priming . Communication Research , 42 ( 8 ), 1089 – 1115 . https://doi.org/10.1177/0093650213482970

Arendt F. , Northup T. , Camaj L. ( 2019 ). Selective exposure and news media brands: Implicit and explicit attitudes as predictors of news choice . Media Psychology , 22 ( 3 ), 526 – 543 . https://doi.org/10.1080/15213269.2017.1338963

Arendt F. , Steindl N. , Kümpel A. ( 2016 ). Implicit and explicit attitudes as predictors of gatekeeping, selective exposure, and news sharing: Testing a general model of media-related selection . Journal of Communication , 66 ( 5 ), 717 – 740 . https://doi.org/10.1111/jcom.12256

Arceneaux K. , Johnson M. ( 2013 ). Changing minds or changing channels? Partisan news in an age of choice . University of Chicago Press .

Bennett W. L. , Iyengar S. ( 2008 ). A new era of minimal effects? The changing foundations of political communication . Journal of Communication , 58 ( 4 ), 707 – 731 . https://doi.org/10.1111/j.1460-2466.2008.00410.x

Billings A. , Parrott S. ( 2020 ). Media stereotypes: From ageism to xenophobia . Peter Lang .

Cacciatore M. A. , Scheufele D. A. , Iyengar S. ( 2016 ). The end of framing as we know it… and the future of media effects . Mass Communication and Society , 19 ( 1 ), 7 – 23 . https://doi.org/10.1080/15205436.2015.1068811

Clifford S. , Sheagley G. , Piston S. ( 2021 ). Increasing precision without altering treatment effects: Repeated measures designs in survey experiments . American Political Science Review , 115 ( 3 ), 1048 – 1065 . https://doi.org/10.1017/S0003055421000241

Dahlgren P. M. ( 2021 ). Forced versus selective exposure: Threatening messages lead to anger but not dislike of political opponents. Journal of Media Psychology: Theories, Methods, and Applications . Advance online publication . https://doi.org/10.1027/1864-1105/a000302

De Benedictis-Kessner J. , Baum M. , Berinsky A. , Yamamoto T. ( 2019 ). Persuading the enemy: Estimating the persuasive effects of partisan media with the preference-incorporating choice and assignment design . American Political Science Review , 113 ( 4 ), 902 – 916 . https://doi.org/10.1017/S0003055419000418

Dixon T. L. ( 2008 ). Crime news and racialized beliefs: Understanding the relationship between local news viewing and perceptions of African Americans and crime . Journal of Communication , 58 ( 1 ), 106 – 125 . https://doi.org/10.1111/j.1460-2466.2007.00376.x

Dovidio J. , Hewstone M. , Glick P. , Esses V. ( 2010 ). The SAGE handbook of prejudice, stereotyping and discrimination . Sage .

Druckman J. N. , Fein J. , Leeper T. J. ( 2012 ). A source of bias in public opinion stability . American Political Science Review , 106 ( 2 ), 430 – 454 . https://doi.org/10.1017/S0003055412000123

Festinger L. ( 1957 ). A theory of cognitive dissonance . Stanford University Press .

Galdi S. , Gawronski B. , Arcuri L. , Friese M. ( 2012 ). Selective exposure in decided and undecided individuals: Differential relations to automatic associations and conscious beliefs . Personality & social psychology bulletin , 38 ( 5 ), 559 – 569 . https://doi.org/10.1177/0146167211435981

Garrett K. ( 2009 ). Politically motivated reinforcement seeking: Reframing the selective exposure debate . Journal of Communication , 59 ( 4 ), 676 – 699 . https://doi.org/10.1111/j.1460-2466.2009.01452.x

Greenwald A. G. , Banaji M. R. , Rudman L. A. , Farnham S. D. , Nosek B. A. , Mellott D. S. ( 2002 ). A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept . Psychological Review , 109 ( 1 ), 3 – 25 . https://doi.org/10.1037//0033-295X.109.1.3

Hayes A. , Matthes J. , Eveland W. ( 2013 ). Stimulating the quasi-statistical organ: Fear of social isolation motivates the quest for knowledge of the opinion climate . Communication Research , 40 ( 4 ), 439 – 462 . https://doi.org/10.1177/0093650211428608

Holt L. F. ( 2013 ). Writing the wrong: Can counter-stereotypes offset negative media messages about African Americans? Journalism & Mass Communication Quarterly , 90 ( 1 ), 108 – 125 . https://doi.org/10.1177/1077699012468699

Jang S. ( 2014 ). Challenges to selective exposure: Selective seeking and avoidance in a multitasking media environment . Mass Communication and Society , 17 ( 5 ), 665 – 688 . https://doi.org/10.1080/15205436.2013.835425

Johnston L. ( 1996 ). Resisting change: Information-seeking and stereotype change . European Journal of Social Psychology , 26 ( 5 ), 799 – 825 . https://doi.org/10.1002/ejsp.2420240505

Johnston L. , MaCrae C. ( 1994 ). Changing social stereotypes: The case of the information seeker . European Journal of Social Psychology , 24 ( 5 ), 581 – 592 . https://doi.org/10.1002/ejsp.2420240505

Knobloch-Westerwick S. ( 2015 ). The selective exposure self- and affect-management (SESAM) model: Applications in the realms of race, politics, and health . Communication Research , 42 ( 7 ), 959 – 985 . https://doi.org/10.1177/0093650214539173

Knobloch-Westerwick S. , Kleinman S. B. ( 2012 ). Preelection selective exposure: Confirmation bias versus informational utility . Communication Research , 39 ( 2 ), 170 – 193 . https://doi.org/10.1177/0093650211400597

Kohlbacher J. , Lehner M. , Rasuly-Paleczek G. ( 2020 ). Afghan/inn/en in Österreich—Perspektiven von Integration, Inklusion und Zusammenleben [Afghans in Austria: Perspectives on integration, inclusion, and cohabitation]. Verlag ÖAW .

Kroon A. , van Selm M. , ter Hoeven C. , Vliegenthart R. ( 2016 ). Poles apart: The processing and consequences of mixed media stereotypes of older workers . Journal of Communication , 66 ( 5 ), 811 – 833 . https://doi.org/10.1111/jcom.12249

Kunda Z. ( 1990 ). The case for motivated reasoning . Psychological Bulletin , 108 ( 3 ), 480 – 498 . https://doi.org/10.1037/0033-2909.108.3.480

Lippmann W. ( 1922 ). Public opinion . Macmillan .

Mastro D. , Behm-Morawitz E. , Ortiz M. ( 2007 ). The cultivation of social perceptions of Latinos: A mental models approach . Media Psychology , 9 ( 2 ), 347 – 365 . https://doi.org/10.1080/15213260701286106

Mastro D. , Tukachinsky R. ( 2011 ). The influence of exemplar versus prototype-based media primes on racial/ethnic evaluations . Journal of Communication , 61 ( 5 ), 916 – 937 . https://doi.org/10.1111/j.1460-2466.2011.01587.x

Oliver M. B. ( 1999 ). Caucasian viewers’ memory of Black and White criminal suspects in the news . Journal of Communication , 49 ( 3 ), 46 – 60 . https://doi.org/10.1111/j.1460-2466.1999.tb02804.x

Power J. , Murphy S. , Coover G. ( 1996 ). Priming prejudice: How stereotypes and counter-stereotypes influence attribution of responsibility and credibility among ingroups and outgroups . Human Communication Research , 23 ( 1 ), 36 – 58 . https://doi.org/10.1111/j.1468-2958.1996.tb00386.x

Ramasubramanian S. ( 2011 ). The impact of stereotypical versus counter-stereotypical media exemplars on racial attitudes, causal attributions, and support for affirmative action . Communication Research , 38 ( 4 ), 497 – 516 . https://doi.org/10.1177/0093650210384854

Ramasubramanian S. ( 2013 ). Intergroup contact, media exposure, and racial attitudes . Journal of Intercultural Communication Research , 42 , 54 – 72 . https://doi.org/10.1080/17475759.2012.707981

Ramasubramanian S. , Murphy C. ( 2014 ). Experimental studies of media stereotyping effects. In Webster M. , Sell J. (Eds.), Laboratory experiments in the social sciences (pp. 385 – 402 ). Elsevier Academic Press . https://doi.org/10.1016/B978-0-12-404681-8.00017-0

Saleem M. , Prot S. , Anderson C. A. , Lemieux A. F. ( 2017 ). Exposure to Muslims in media and support for public policies harming Muslims . Communication Research , 44 ( 6 ), 841 – 869 . https://doi.org/10.1177/0093650215619214

Schemer C. ( 2012 ). The influence of news media on stereotypic attitudes toward immigrants in a political campaign . Journal of Communication , 62 ( 5 ), 739 – 757 . https://doi.org/10.1111/j.1460-2466.2012.01672.x

Schmuck D. , Matthes J. , Paul F. H. ( 2017 ). Negative stereotypical portrayals of Muslims in right-wing populist campaigns: Perceived discrimination, social identity threats, and hostility among young Muslim adults . Journal of Communication , 67 ( 4 ), 610 – 634 . https://doi.org/10.1111/jcom.12313

Schmuck D. , Tribastone M. , Matthes J. , Marquart F. , Bergel E. ( 2020 ). Avoiding the other side? An eye-tracking study of selective exposure and selective avoidance effects in response to political advertising . Journal of Media Psychology , 32 ( 3 ), 158 – 164 . https://doi-org.uaccess.univie.ac.at/10.1027/1864-1105/a000265

Slater M. D. ( 2007 ). Reinforcing spirals: The mutual influence of media selectivity and media effects and their impact on individual behavior and social identity . Communication Theory , 17 ( 3 ), 281 – 303 . https://doi.org/10.1111/j.1468-2885.2007.00296.x

Snyder M. ( 1981 ). On the self-perpetuating nature of social stereotypes. In Hamilton D. L. (Ed.), Cognitive processes in stereotyping and intergroup behavior (pp. 183 – 212 ). LEA .

Steiger B. ( 2021 ). Fast jeden zweiten Tag schrieb der Boulevard 2020 etwas Negatives über Menschen aus Afghanistan. Kobuk . Retrieved from https://www.kobuk.at/2021/02/fast-jeden-zweiten-tag-schrieb-der-boulevard-2020-etwas-negatives-ueber-menschen-aus-afghanistan/

Strabac Z. , Listhaug O. ( 2008 ). Anti-Muslim prejudice in Europe: A multilevel analysis of survey data from 30 countries . Social Science Research , 37 ( 1 ), 268 – 286 . https://doi.org/10.1016/j.ssresearch.2007.02.004

Stroud N. J. , Feldman L. , Wojcieszak M. , Bimber B. ( 2019 ). The consequences of forced versus selected political media exposure . Human Communication Research , 45 ( 1 ), 27 – 51 . https://doi.org/10.1093/hcr/hqy012

Tajfel H. , Turner J. C. ( 1986 ). The social identity theory of intergroup behavior. In Worchel S. , Austin W. (Eds.), Psychology of intergroup relations (pp. 7 – 24 ). Nelson .

Võ M. L. , Conrad M. , Kuchinke L. , Urton K. , Hofmann M. J. , Jacobs A. M. ( 2009 ). The Berlin affective word list reloaded (BAWL-R) . Behavior Research Methods , 41 ( 2 ), 534 – 538 . https://doi.org/10.3758/BRM.41.2.534

Supplementary data

Email alerts

Citing articles via.

• Recommend to your Library

Affiliations

• Online ISSN 1460-2466
• Print ISSN 0021-9916
• Copyright © 2024 International Communication Association
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 28 August 2024

AI generates covertly racist decisions about people based on their dialect

• Valentin Hofmann   ORCID: orcid.org/0000-0001-6603-3428 1 , 2 , 3 ,
• Pratyusha Ria Kalluri 4 ,
• Dan Jurafsky   ORCID: orcid.org/0000-0002-6459-7745 4 &
• Sharese King 5  

Nature ( 2024 ) Cite this article

50 Altmetric

Metrics details

• Computer science

Hundreds of millions of people now interact with language models, with uses ranging from help with writing 1 , 2 to informing hiring decisions 3 . However, these language models are known to perpetuate systematic racial prejudices, making their judgements biased in problematic ways about groups such as African Americans 4 , 5 , 6 , 7 . Although previous research has focused on overt racism in language models, social scientists have argued that racism with a more subtle character has developed over time, particularly in the United States after the civil rights movement 8 , 9 . It is unknown whether this covert racism manifests in language models. Here, we demonstrate that language models embody covert racism in the form of dialect prejudice, exhibiting raciolinguistic stereotypes about speakers of African American English (AAE) that are more negative than any human stereotypes about African Americans ever experimentally recorded. By contrast, the language models’ overt stereotypes about African Americans are more positive. Dialect prejudice has the potential for harmful consequences: language models are more likely to suggest that speakers of AAE be assigned less-prestigious jobs, be convicted of crimes and be sentenced to death. Finally, we show that current practices of alleviating racial bias in language models, such as human preference alignment, exacerbate the discrepancy between covert and overt stereotypes, by superficially obscuring the racism that language models maintain on a deeper level. Our findings have far-reaching implications for the fair and safe use of language technology.

Similar content being viewed by others

Large language models propagate race-based medicine

The benefits, risks and bounds of personalizing the alignment of large language models to individuals

Cognitive causes of ‘like me’ race and gender biases in human language production

Language models are a type of artificial intelligence (AI) that has been trained to process and generate text. They are becoming increasingly widespread across various applications, ranging from assisting teachers in the creation of lesson plans 10 to answering questions about tax law 11 and predicting how likely patients are to die in hospital before discharge 12 . As the stakes of the decisions entrusted to language models rise, so does the concern that they mirror or even amplify human biases encoded in the data they were trained on, thereby perpetuating discrimination against racialized, gendered and other minoritized social groups 4 , 5 , 6 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 .

Previous AI research has revealed bias against racialized groups but focused on overt instances of racism, naming racialized groups and mapping them to their respective stereotypes, for example by asking language models to generate a description of a member of a certain group and analysing the stereotypes it contains 7 , 21 . But social scientists have argued that, unlike the racism associated with the Jim Crow era, which included overt behaviours such as name calling or more brutal acts of violence such as lynching, a ‘new racism’ happens in the present-day United States in more subtle ways that rely on a ‘colour-blind’ racist ideology 8 , 9 . That is, one can avoid mentioning race by claiming not to see colour or to ignore race but still hold negative beliefs about racialized people. Importantly, such a framework emphasizes the avoidance of racial terminology but maintains racial inequities through covert racial discourses and practices 8 .

Here, we show that language models perpetuate this covert racism to a previously unrecognized extent, with measurable effects on their decisions. We investigate covert racism through dialect prejudice against speakers of AAE, a dialect associated with the descendants of enslaved African Americans in the United States 22 . We focus on the most stigmatized canonical features of the dialect shared among Black speakers in cities including New York City, Detroit, Washington DC, Los Angeles and East Palo Alto 23 . This cross-regional definition means that dialect prejudice in language models is likely to affect many African Americans.

Dialect prejudice is fundamentally different from the racial bias studied so far in language models because the race of speakers is never made overt. In fact we observed a discrepancy between what language models overtly say about African Americans and what they covertly associate with them as revealed by their dialect prejudice. This discrepancy is particularly pronounced for language models trained with human feedback (HF), such as GPT4: our results indicate that HF training obscures the racism on the surface, but the racial stereotypes remain unaffected on a deeper level. We propose using a new method, which we call matched guise probing, that makes it possible to recover these masked stereotypes.

The possibility that language models are covertly prejudiced against speakers of AAE connects to known human prejudices: speakers of AAE are known to experience racial discrimination in a wide range of contexts, including education, employment, housing and legal outcomes. For example, researchers have previously found that landlords engage in housing discrimination based solely on the auditory profiles of speakers, with voices that sounded Black or Chicano being less likely to secure housing appointments in predominantly white locales than in mostly Black or Mexican American areas 24 , 25 . Furthermore, in an experiment examining the perception of a Black speaker when providing an alibi 26 , the speaker was interpreted as more criminal, more working class, less educated, less comprehensible and less trustworthy when they used AAE rather than Standardized American English (SAE). Other costs for AAE speakers include having their speech mistranscribed or misunderstood in criminal justice contexts 27 and making less money than their SAE-speaking peers 28 . These harms connect to themes in broader racial ideology about African Americans and stereotypes about their intelligence, competence and propensity to commit crimes 29 , 30 , 31 , 32 , 33 , 34 , 35 . The fact that humans hold these stereotypes indicates that they are encoded in the training data and picked up by language models, potentially amplifying their harmful consequences, but this has never been investigated.

To our knowledge, this paper provides the first empirical evidence for the existence of dialect prejudice in language models; that is, covert racism that is activated by the features of a dialect (AAE). Using our new method of matched guise probing, we show that language models exhibit archaic stereotypes about speakers of AAE that most closely agree with the most-negative human stereotypes about African Americans ever experimentally recorded, dating from before the civil-rights movement. Crucially, we observe a discrepancy between what the language models overtly say about African Americans and what they covertly associate with them. Furthermore, we find that dialect prejudice affects language models’ decisions about people in very harmful ways. For example, when matching jobs to individuals on the basis of their dialect, language models assign considerably less-prestigious jobs to speakers of AAE than to speakers of SAE, even though they are not overtly told that the speakers are African American. Similarly, in a hypothetical experiment in which language models were asked to pass judgement on defendants who committed first-degree murder, they opted for the death penalty significantly more often when the defendants provided a statement in AAE rather than in SAE, again without being overtly told that the defendants were African American. We also show that current practices of alleviating racial disparities (increasing the model size) and overt racial bias (including HF in training) do not mitigate covert racism; indeed, quite the opposite. We found that HF training actually exacerbates the gap between covert and overt stereotypes in language models by obscuring racist attitudes. Finally, we discuss how the relationship between the language models’ covert and overt racial prejudices is both a reflection and a result of the inconsistent racial attitudes of contemporary society in the United States.

Probing AI dialect prejudice

To explore how dialect choice impacts the predictions that language models make about speakers in the absence of other cues about their racial identity, we took inspiration from the ‘matched guise’ technique used in sociolinguistics, in which subjects listen to recordings of speakers of two languages or dialects and make judgements about various traits of those speakers 36 , 37 . Applying the matched guise technique to the AAE–SAE contrast, researchers have shown that people identify speakers of AAE as Black with above-chance accuracy 24 , 26 , 38 and attach racial stereotypes to them, even without prior knowledge of their race 39 , 40 , 41 , 42 , 43 . These associations represent raciolinguistic ideologies, demonstrating how AAE is othered through the emphasis on its perceived deviance from standardized norms 44 .

Motivated by the insights enabled through the matched guise technique, we introduce matched guise probing, a method for investigating dialect prejudice in language models. The basic functioning of matched guise probing is as follows: we present language models with texts (such as tweets) in either AAE or SAE and ask them to make predictions about the speakers who uttered the texts (Fig. 1 and Methods ). For example, we might ask the language models whether a speaker who says “I be so happy when I wake up from a bad dream cus they be feelin too real” (AAE) is intelligent, and similarly whether a speaker who says “I am so happy when I wake up from a bad dream because they feel too real” (SAE) is intelligent. Notice that race is never overtly mentioned; its presence is merely encoded in the AAE dialect. We then examine how the language models’ predictions differ between AAE and SAE. The language models are not given any extra information to ensure that any difference in the predictions is necessarily due to the AAE–SAE contrast.

a , We used texts in SAE (green) and AAE (blue). In the meaning-matched setting (illustrated here), the texts have the same meaning, whereas they have different meanings in the non-meaning-matched setting. b , We embedded the SAE and AAE texts in prompts that asked for properties of the speakers who uttered the texts. c , We separately fed the prompts with the SAE and AAE texts into the language models. d , We retrieved and compared the predictions for the SAE and AAE inputs, here illustrated by five adjectives from the Princeton Trilogy. See Methods for more details.

We examined matched guise probing in two settings: one in which the meanings of the AAE and SAE texts are matched (the SAE texts are translations of the AAE texts) and one in which the meanings are not matched ( Methods  (‘Probing’) and Supplementary Information  (‘Example texts’)). Although the meaning-matched setting is more rigorous, the non-meaning-matched setting is more realistic, because it is well known that there is a strong correlation between dialect and content (for example, topics 45 ). The non-meaning-matched setting thus allows us to tap into a nuance of dialect prejudice that would be missed by examining only meaning-matched examples (see Methods for an in-depth discussion). Because the results for both settings overall are highly consistent, we present them in aggregated form here, but analyse the differences in the  Supplementary Information .

We examined GPT2 (ref. 46 ), RoBERTa 47 , T5 (ref. 48 ), GPT3.5 (ref. 49 ) and GPT4 (ref. 50 ), each in one or more model versions, amounting to a total of 12 examined models ( Methods and Supplementary Information (‘Language models’)). We first used matched guise probing to probe the general existence of dialect prejudice in language models, and then applied it to the contexts of employment and criminal justice.

Covert stereotypes in language models

We started by investigating whether the attitudes that language models exhibit about speakers of AAE reflect human stereotypes about African Americans. To do so, we replicated the experimental set-up of the Princeton Trilogy 29 , 30 , 31 , 34 , a series of studies investigating the racial stereotypes held by Americans, with the difference that instead of overtly mentioning race to the language models, we used matched guise probing based on AAE and SAE texts ( Methods ).

Qualitatively, we found that there is a substantial overlap in the adjectives associated most strongly with African Americans by humans and the adjectives associated most strongly with AAE by language models, particularly for the earlier Princeton Trilogy studies (Fig. 2a ). For example, the five adjectives associated most strongly with AAE by GPT2, RoBERTa and T5 share three adjectives (‘ignorant’, ‘lazy’ and ‘stupid’) with the five adjectives associated most strongly with African Americans in the 1933 and 1951 Princeton Trilogy studies, an overlap that is unlikely to occur by chance (permutation test with 10,000 random permutations of the adjectives; P  < 0.01). Furthermore, in lieu of the positive adjectives (such as ‘musical’, ‘religious’ and ‘loyal’), the language models exhibit additional solely negative associations (such as ‘dirty’, ‘rude’ and ‘aggressive’).

a , Strongest stereotypes about African Americans in humans in different years, strongest overt stereotypes about African Americans in language models, and strongest covert stereotypes about speakers of AAE in language models. Colour coding as positive (green) and negative (red) is based on ref. 34 . Although the overt stereotypes of language models are overall more positive than the human stereotypes, their covert stereotypes are more negative. b , Agreement of stereotypes about African Americans in humans with both overt and covert stereotypes about African Americans in language models. The black dotted line shows chance agreement using a random bootstrap. Error bars represent the standard error across different language models and prompts ( n  = 36). The language models’ overt stereotypes agree most strongly with current human stereotypes, which are the most positive experimentally recorded ones, but their covert stereotypes agree most strongly with human stereotypes from the 1930s, which are the most negative experimentally recorded ones. c , Stereotype strength for individual linguistic features of AAE. Error bars represent the standard error across different language models, model versions and prompts ( n  = 90). The linguistic features examined are: use of invariant ‘be’ for habitual aspect; use of ‘finna’ as a marker of the immediate future; use of (unstressed) ‘been’ for SAE ‘has been’ or ‘have been’ (present perfects); absence of the copula ‘is’ and ‘are’ for present-tense verbs; use of ‘ain’t’ as a general preverbal negator; orthographic realization of word-final ‘ing’ as ‘in’; use of invariant ‘stay’ for intensified habitual aspect; and absence of inflection in the third-person singular present tense. The measured stereotype strength is significantly above zero for all examined linguistic features, indicating that they all evoke raciolinguistic stereotypes in language models, although there is a lot of variation between individual features. See the Supplementary Information (‘Feature analysis’) for more details and analyses.

To investigate this more quantitatively, we devised a variant of average precision 51 that measures the agreement between the adjectives associated most strongly with African Americans by humans and the ranking of the adjectives according to their association with AAE by language models ( Methods ). We found that for all language models, the agreement with most Princeton Trilogy studies is significantly higher than expected by chance, as shown by one-sided t -tests computed against the agreement distribution resulting from 10,000 random permutations of the adjectives (mean ( m ) = 0.162, standard deviation ( s ) = 0.106; Extended Data Table 1 ); and that the agreement is particularly pronounced for the stereotypes reported in 1933 and falls for each study after that, almost reaching the level of chance agreement for 2012 (Fig. 2b ). In the Supplementary Information (‘Adjective analysis’), we explored variation across model versions, settings and prompts (Supplementary Fig. 2 and Supplementary Table 4 ).

To explain the observed temporal trend, we measured the average favourability of the top five adjectives for all Princeton Trilogy studies and language models, drawing from crowd-sourced ratings for the Princeton Trilogy adjectives on a scale between −2 (very negative) and 2 (very positive; see Methods , ‘Covert-stereotype analysis’). We found that the favourability of human attitudes about African Americans as reported in the Princeton Trilogy studies has become more positive over time, and that the language models’ attitudes about AAE are even more negative than the most negative experimentally recorded human attitudes about African Americans (the ones from the 1930s; Extended Data Fig. 1 ). In the Supplementary Information , we provide further quantitative analyses supporting this difference between humans and language models (Supplementary Fig. 7 ).

Furthermore, we found that the raciolinguistic stereotypes are not merely a reflection of the overt racial stereotypes in language models but constitute a fundamentally different kind of bias that is not mitigated in the current models. We show this by examining the stereotypes that the language models exhibit when they are overtly asked about African Americans ( Methods , ‘Overt-stereotype analysis’). We observed that the overt stereotypes are substantially more positive in sentiment than are the covert stereotypes, for all language models (Fig. 2a and Extended Data Fig. 1 ). Strikingly, for RoBERTa, T5, GPT3.5 and GPT4, although their covert stereotypes about speakers of AAE are more negative than the most negative experimentally recorded human stereotypes, their overt stereotypes about African Americans are more positive than the most positive experimentally recorded human stereotypes. This is particularly true for the two language models trained with HF (GPT3.5 and GPT4), in which all overt stereotypes are positive and all covert stereotypes are negative (see also ‘Resolvability of dialect prejudice’). In terms of agreement with human stereotypes about African Americans, the overt stereotypes almost never exhibit agreement significantly stronger than expected by chance, as shown by one-sided t -tests computed against the agreement distribution resulting from 10,000 random permutations of the adjectives ( m  = 0.162, s  = 0.106; Extended Data Table 2 ). Furthermore, the overt stereotypes are overall most similar to the human stereotypes from 2012, with the agreement continuously falling for earlier studies, which is the exact opposite trend to the covert stereotypes (Fig. 2b ).

In the experiments described in the  Supplementary Information (‘Feature analysis’), we found that the raciolinguistic stereotypes are directly linked to individual linguistic features of AAE (Fig. 2c and Supplementary Table 14 ), and that a higher density of such linguistic features results in stronger stereotypical associations (Supplementary Fig. 11 and Supplementary Table 13 ). Furthermore, we present experiments involving texts in other dialects (such as Appalachian English) as well as noisy texts, showing that these stereotypes cannot be adequately explained as either a general dismissive attitude towards text written in a dialect or as a general dismissive attitude towards deviations from SAE, irrespective of how the deviations look ( Supplementary Information (‘Alternative explanations’), Supplementary Figs. 12 and 13 and Supplementary Tables 15 and 16 ). Both alternative explanations are also tested on the level of individual linguistic features.

Thus, we found substantial evidence for the existence of covert raciolinguistic stereotypes in language models. Our experiments show that these stereotypes are similar to the archaic human stereotypes about African Americans that existed before the civil rights movement, are even more negative than the most negative experimentally recorded human stereotypes about African Americans, and are both qualitatively and quantitatively different from the previously reported overt racial stereotypes in language models, indicating that they are a fundamentally different kind of bias. Finally, our analyses demonstrate that the detected stereotypes are inherently linked to AAE and its linguistic features.

Impact of covert racism on AI decisions

To determine what harmful consequences the covert stereotypes have in the real world, we focused on two areas in which racial stereotypes about speakers of AAE and African Americans have been repeatedly shown to bias human decisions: employment and criminality. There is a growing impetus to use AI systems in these areas. Indeed, AI systems are already being used for personnel selection 52 , 53 , including automated analyses of applicants’ social-media posts 54 , 55 , and technologies for predicting legal outcomes are under active development 56 , 57 , 58 . Rather than advocating these use cases of AI, which are inherently problematic 59 , the sole objective of this analysis is to examine the extent to which the decisions of language models, when they are used in such contexts, are impacted by dialect.

First, we examined decisions about employability. Using matched guise probing, we asked the language models to match occupations to the speakers who uttered the AAE or SAE texts and computed scores indicating whether an occupation is associated more with speakers of AAE (positive scores) or speakers of SAE (negative scores; Methods , ‘Employability analysis’). The average score of the occupations was negative ( m  = –0.046,  s  = 0.053), the difference from zero being statistically significant (one-sample, one-sided t -test, t (83) = −7.9, P  < 0.001). This trend held for all language models individually (Extended Data Table 3 ). Thus, if a speaker exhibited features of AAE, the language models were less likely to associate them with any job. Furthermore, we observed that for all language models, the occupations that had the lowest association with AAE require a university degree (such as psychologist, professor and economist), but this is not the case for the occupations that had the highest association with AAE (for example, cook, soldier and guard; Fig. 3a ). Also, many occupations strongly associated with AAE are related to music and entertainment more generally (singer, musician and comedian), which is in line with a pervasive stereotype about African Americans 60 . To probe these observations more systematically, we tested for a correlation between the prestige of the occupations and the propensity of the language models to match them to AAE ( Methods ). Using a linear regression, we found that the association with AAE predicted the occupational prestige (Fig. 3b ; β  = −7.8, R 2 = 0.193, F (1, 63) = 15.1, P  < 0.001). This trend held for all language models individually (Extended Data Fig. 2 and Extended Data Table 4 ), albeit in a less pronounced way for GPT3.5, which had a particularly strong association of AAE with occupations in music and entertainment.

a , Association of different occupations with AAE or SAE. Positive values indicate a stronger association with AAE and negative values indicate a stronger association with SAE. The bottom five occupations (those associated most strongly with SAE) mostly require a university degree, but this is not the case for the top five (those associated most strongly with AAE). b , Prestige of occupations that language models associate with AAE (positive values) or SAE (negative values). The shaded area shows a 95% confidence band around the regression line. The association with AAE or SAE predicts the occupational prestige. Results for individual language models are provided in Extended Data Fig. 2 . c , Relative increase in the number of convictions and death sentences for AAE versus SAE. Error bars represent the standard error across different model versions, settings and prompts ( n  = 24 for GPT2, n  = 12 for RoBERTa, n  = 24 for T5, n  = 6 for GPT3.5 and n  = 6 for GPT4). In cases of small sample size ( n  ≤ 10 for GPT3.5 and GPT4), we plotted the individual results as overlaid dots. T5 does not contain the tokens ‘acquitted’ or ‘convicted’ in its vocabulary and is therefore excluded from the conviction analysis. Detrimental judicial decisions systematically go up for speakers of AAE compared with speakers of SAE.

We then examined decisions about criminality. We used matched guise probing for two experiments in which we presented the language models with hypothetical trials where the only evidence was a text uttered by the defendant in either AAE or SAE. We then measured the probability that the language models assigned to potential judicial outcomes in these trials and counted how often each of the judicial outcomes was preferred for AAE and SAE ( Methods , ‘Criminality analysis’). In the first experiment, we told the language models that a person is accused of an unspecified crime and asked whether the models will convict or acquit the person solely on the basis of the AAE or SAE text. Overall, we found that the rate of convictions was greater for AAE ( r  = 68.7%) than SAE ( r  = 62.1%; Fig. 3c , left). A chi-squared test found a strong effect ( χ 2 (1,  N  = 96) = 184.7,  P  < 0.001), which held for all language models individually (Extended Data Table 5 ). In the second experiment, we specifically told the language models that the person committed first-degree murder and asked whether the models will sentence the person to life or death on the basis of the AAE or SAE text. The overall rate of death sentences was greater for AAE ( r  = 27.7%) than for SAE ( r  = 22.8%; Fig. 3c , right). A chi-squared test found a strong effect ( χ 2 (1,  N  = 144) = 425.4,  P  < 0.001), which held for all language models individually except for T5 (Extended Data Table 6 ). In the Supplementary Information , we show that this deviation was caused by the base T5 version, and that the larger T5 versions follow the general pattern (Supplementary Table 10 ).

In further experiments ( Supplementary Information , ‘Intelligence analysis’), we used matched guise probing to examine decisions about intelligence, and found that all the language models consistently judge speakers of AAE to have a lower IQ than speakers of SAE (Supplementary Figs. 14 and 15 and Supplementary Tables 17 – 19 ).

Resolvability of dialect prejudice

We wanted to know whether the dialect prejudice we observed is resolved by current practices of bias mitigation, such as increasing the size of the language model or including HF in training. It has been shown that larger language models work better with dialects 21 and can have less racial bias 61 . Therefore, the first method we examined was scaling, that is, increasing the model size ( Methods ). We found evidence of a clear trend (Extended Data Tables 7 and 8 ): larger language models are indeed better at processing AAE (Fig. 4a , left), but they are not less prejudiced against speakers of it. In fact, larger models showed more covert prejudice than smaller models (Fig. 4a , right). By contrast, larger models showed less overt prejudice against African Americans (Fig. 4a , right). Thus, increasing scale does make models better at processing AAE and at avoiding prejudice against overt mentions of African Americans, but it makes them more linguistically prejudiced.

a , Language modelling perplexity and stereotype strength on AAE text as a function of model size. Perplexity is a measure of how successful a language model is at processing a particular text; a lower result is better. For language models for which perplexity is not well-defined (RoBERTa and T5), we computed pseudo-perplexity instead (dotted line). Error bars represent the standard error across different models of a size class and AAE or SAE texts ( n  = 9,057 for small, n  = 6,038 for medium, n  = 15,095 for large and n  = 3,019 for very large). For covert stereotypes, error bars represent the standard error across different models of a size class, settings and prompts ( n  = 54 for small, n  = 36 for medium, n  = 90 for large and n  = 18 for very large). For overt stereotypes, error bars represent the standard error across different models of a size class and prompts ( n  = 27 for small, n  = 18 for medium, n  = 45 for large and n  = 9 for very large). Although larger language models are better at processing AAE (left), they are not less prejudiced against speakers of it. Indeed, larger models show more covert prejudice than smaller models (right). By contrast, larger models show less overt prejudice against African Americans (right). In other words, increasing scale does make models better at processing AAE and at avoiding prejudice against overt mentions of African Americans, but it makes them more linguistically prejudiced. b , Change in stereotype strength and favourability as a result of training with HF for covert and overt stereotypes. Error bars represent the standard error across different prompts ( n  = 9). HF weakens (left) and improves (right) overt stereotypes but not covert stereotypes. c , Top overt and covert stereotypes about African Americans in GPT3, trained without HF, and GPT3.5, trained with HF. Colour coding as positive (green) and negative (red) is based on ref. 34 . The overt stereotypes get substantially more positive as a result of HF training in GPT3.5, but there is no visible change in favourability for the covert stereotypes.

As a second potential way to resolve dialect prejudice in language models, we examined training with HF 49 , 62 . Specifically, we compared GPT3.5 (ref. 49 ) with GPT3 (ref. 63 ), its predecessor that was trained without using HF ( Methods ). Looking at the top adjectives associated overtly and covertly with African Americans by the two language models, we found that HF resulted in more-positive overt associations but had no clear qualitative effect on the covert associations (Fig. 4c ). This observation was confirmed by quantitative analyses: the inclusion of HF resulted in significantly weaker (no HF, m  = 0.135,  s  = 0.142; HF, m  = −0.119,  s  = 0.234;  t (16) = 2.6,  P  < 0.05) and more favourable (no HF, m  = 0.221,  s  = 0.399; HF, m  = 1.047,  s  = 0.387;  t (16) = −6.4,  P  < 0.001) overt stereotypes but produced no significant difference in the strength (no HF, m  = 0.153,  s  = 0.049; HF, m  = 0.187,  s  = 0.066;  t (16) = −1.2, P  = 0.3) or unfavourability (no HF, m  = −1.146, s  = 0.580; HF, m = −1.029, s  = 0.196; t (16) = −0.5, P  = 0.6) of covert stereotypes (Fig. 4b ). Thus, HF training weakens and ameliorates the overt stereotypes but has no clear effect on the covert stereotypes; in other words, it obscures the racist attitudes on the surface, but more subtle forms of racism, such as dialect prejudice, remain unaffected. This finding is underscored by the fact that the discrepancy between overt and covert stereotypes about African Americans is most pronounced for the two examined language models trained with human feedback (GPT3.5 and GPT4; see ‘Covert stereotypes in language models’). Furthermore, this finding again shows that there is a fundamental difference between overt and covert stereotypes in language models, and that mitigating the overt stereotypes does not automatically translate to mitigated covert stereotypes.

To sum up, neither scaling nor training with HF as applied today resolves the dialect prejudice. The fact that these two methods effectively mitigate racial performance disparities and overt racial stereotypes in language models indicates that this form of covert racism constitutes a different problem that is not addressed by current approaches for improving and aligning language models.

The key finding of this article is that language models maintain a form of covert racial prejudice against African Americans that is triggered by dialect features alone. In our experiments, we avoided overt mentions of race but drew from the racialized meanings of a stigmatized dialect, and could still find historically racist associations with African Americans. The implicit nature of this prejudice, that is, the fact it is about something that is not explicitly expressed in the text, makes it fundamentally different from the overt racial prejudice that has been the focus of previous research. Strikingly, the language models’ covert and overt racial prejudices are often in contradiction with each other, especially for the most recent language models that have been trained with HF (GPT3.5 and GPT4). These two language models obscure the racism, overtly associating African Americans with exclusively positive attributes (such as ‘brilliant’), but our results show that they covertly associate African Americans with exclusively negative attributes (such as ‘lazy’).

We argue that this paradoxical relation between the language models’ covert and overt racial prejudices manifests the inconsistent racial attitudes present in the contemporary society of the United States 8 , 64 . In the Jim Crow era, stereotypes about African Americans were overtly racist, but the normative climate after the civil rights movement made expressing explicitly racist views distasteful. As a result, racism acquired a covert character and continued to exist on a more subtle level. Thus, most white people nowadays report positive attitudes towards African Americans in surveys but perpetuate racial inequalities through their unconscious behaviour, such as their residential choices 65 . It has been shown that negative stereotypes persist, even if they are superficially rejected 66 , 67 . This ambivalence is reflected by the language models we analysed, which are overtly non-racist but covertly exhibit archaic stereotypes about African Americans, showing that they reproduce a colour-blind racist ideology. Crucially, the civil rights movement is generally seen as the period during which racism shifted from overt to covert 68 , 69 , and this is mirrored by our results: all the language models overtly agree the most with human stereotypes from after the civil rights movement, but covertly agree the most with human stereotypes from before the civil rights movement.

Our findings beg the question of how dialect prejudice got into the language models. Language models are pretrained on web-scraped corpora such as WebText 46 , C4 (ref. 48 ) and the Pile 70 , which encode raciolinguistic stereotypes about AAE. A drastic example of this is the use of ‘mock ebonics’ to parodize speakers of AAE 71 . Crucially, a growing body of evidence indicates that language models pick up prejudices present in the pretraining corpus 72 , 73 , 74 , 75 , which would explain how they become prejudiced against speakers of AAE, and why they show varying levels of dialect prejudice as a function of the pretraining corpus. However, the web also abounds with overt racism against African Americans 76 , 77 , so we wondered why the language models exhibit much less overt than covert racial prejudice. We argue that the reason for this is that the existence of overt racism is generally known to people 32 , which is not the case for covert racism 69 . Crucially, this also holds for the field of AI. The typical pipeline of training language models includes steps such as data filtering 48 and, more recently, HF training 62 that remove overt racial prejudice. As a result, much of the overt racism on the web does not end up in the language models. However, there are currently no measures in place to curtail covert racial prejudice when training language models. For example, common datasets for HF training 62 , 78 do not include examples that would train the language models to treat speakers of AAE and SAE equally. As a result, the covert racism encoded in the training data can make its way into the language models in an unhindered fashion. It is worth mentioning that the lack of awareness of covert racism also manifests during evaluation, where it is common to test language models for overt racism but not for covert racism 21 , 63 , 79 , 80 .

As well as the representational harms, by which we mean the pernicious representation of AAE speakers, we also found evidence for substantial allocational harms. This refers to the inequitable allocation of resources to AAE speakers 81 (Barocas et al., unpublished observations), and adds to known cases of language technology putting speakers of AAE at a disadvantage by performing worse on AAE 82 , 83 , 84 , 85 , 86 , 87 , 88 , misclassifying AAE as hate speech 81 , 89 , 90 , 91 or treating AAE as incorrect English 83 , 85 , 92 . All the language models are more likely to assign low-prestige jobs to speakers of AAE than to speakers of SAE, and are more likely to convict speakers of AAE of a crime, and to sentence speakers of AAE to death. Although the details of our tasks are constructed, the findings reveal real and urgent concerns because business and jurisdiction are areas for which AI systems involving language models are currently being developed or deployed. As a consequence, the dialect prejudice we uncovered might already be affecting AI decisions today, for example when a language model is used in application-screening systems to process background information, which might include social-media text. Worryingly, we also observe that larger language models and language models trained with HF exhibit stronger covert, but weaker overt, prejudice. Against the backdrop of continually growing language models and the increasingly widespread adoption of HF training, this has two risks: first, that language models, unbeknownst to developers and users, reach ever-increasing levels of covert prejudice; and second, that developers and users mistake ever-decreasing levels of overt prejudice (the only kind of prejudice currently tested for) for a sign that racism in language models has been solved. There is therefore a realistic possibility that the allocational harms caused by dialect prejudice in language models will increase further in the future, perpetuating the racial discrimination experienced by generations of African Americans.

Matched guise probing examines how strongly a language model associates certain tokens, such as personality traits, with AAE compared with SAE. AAE can be viewed as the treatment condition, whereas SAE functions as the control condition. We start by explaining the basic experimental unit of matched guise probing: measuring how a language model associates certain tokens with an individual text in AAE or SAE. Based on this, we introduce two different settings for matched guise probing (meaning-matched and non-meaning-matched), which are both inspired by the matched guise technique used in sociolinguistics 36 , 37 , 93 , 94 and provide complementary views on the attitudes a language model has about a dialect.

The basic experimental unit of matched guise probing is as follows. Let θ be a language model, t be a text in AAE or SAE, and x be a token of interest, typically a personality trait such as ‘intelligent’. We embed the text in a prompt v , for example v ( t ) = ‘a person who says t tends to be’, and compute P ( x ∣ v ( t );  θ ), which is the probability that θ assigns to x after processing v ( t ). We calculate P ( x ∣ v ( t );  θ ) for equally sized sets T a of AAE texts and T s of SAE texts, comparing various tokens from a set X as possible continuations. It has been shown that P ( x ∣ v ( t );  θ ) can be affected by the precise wording of v , so small modifications of v can have an unpredictable effect on the predictions made by the language model 21 , 95 , 96 . To account for this fact, we consider a set V containing several prompts ( Supplementary Information ). For all experiments, we have provided detailed analyses of variation across prompts in the  Supplementary Information .

We conducted matched guise probing in two settings. In the first setting, the texts in T a and T s formed pairs expressing the same underlying meaning, that is, the i -th text in T a (for example, ‘I be so happy when I wake up from a bad dream cus they be feelin too real’) matches the i -th text in T s (for example, ‘I am so happy when I wake up from a bad dream because they feel too real’). For this setting, we used the dataset from ref. 87 , which contains 2,019 AAE tweets together with their SAE translations. In the second setting, the texts in T a and T s did not form pairs, so they were independent texts in AAE and SAE. For this setting, we sampled 2,000 AAE and SAE tweets from the dataset in ref. 83 and used tweets strongly aligned with African Americans for AAE and tweets strongly aligned with white people for SAE ( Supplementary Information (‘Analysis of non-meaning-matched texts’), Supplementary Fig. 1 and Supplementary Table 3 ). In the  Supplementary Information , we include examples of AAE and SAE texts for both settings (Supplementary Tables 1 and 2 ). Tweets are well suited for matched guise probing because they are a rich source of dialectal variation 97 , 98 , 99 , especially for AAE 100 , 101 , 102 , but matched guise probing can be applied to any type of text. Although we do not consider it here, matched guise probing can in principle also be applied to speech-based models, with the potential advantage that dialectal variation on the phonetic level could be captured more directly, which would make it possible to study dialect prejudice specific to regional variants of AAE 23 . However, note that a great deal of phonetic variation is reflected orthographically in social-media texts 101 .

It is important to analyse both meaning-matched and non-meaning-matched settings because they capture different aspects of the attitudes a language model has about speakers of AAE. Controlling for the underlying meaning makes it possible to uncover differences in the attitudes of the language model that are solely due to grammatical and lexical features of AAE. However, it is known that various properties other than linguistic features correlate with dialect, such as topics 45 , and these might also influence the attitudes of the language model. Sidelining such properties bears the risk of underestimating the harms that dialect prejudice causes for speakers of AAE in the real world. For example, in a scenario in which a language model is used in the context of automated personnel selection to screen applicants’ social-media posts, the texts of two competing applicants typically differ in content and do not come in pairs expressing the same meaning. The relative advantages of using meaning-matched or non-meaning-matched data for matched guise probing are conceptually similar to the relative advantages of using the same or different speakers for the matched guise technique: more control in the former versus more naturalness in the latter setting 93 , 94 . Because the results obtained in both settings were consistent overall for all experiments, we aggregated them in the main article, but we analysed differences in detail in the  Supplementary Information .

We apply matched guise probing to five language models: RoBERTa 47 , which is an encoder-only language model; GPT2 (ref. 46 ), GPT3.5 (ref. 49 ) and GPT4 (ref. 50 ), which are decoder-only language models; and T5 (ref. 48 ), which is an encoder–decoder language model. For each language model, we examined one or more model versions: GPT2 (base), GPT2 (medium), GPT2 (large), GPT2 (xl), RoBERTa (base), RoBERTa (large), T5 (small), T5 (base), T5 (large), T5 (3b), GPT3.5 (text-davinci-003) and GPT4 (0613). Where we used several model versions per language model (GPT2, RoBERTa and T5), the model versions all had the same architecture and were trained on the same data but differed in their size. Furthermore, we note that GPT3.5 and GPT4 are the only language models examined in this paper that were trained with HF, specifically reinforcement learning from human feedback 103 . When it is clear from the context what is meant, or when the distinction does not matter, we use the term ‘language models’, or sometimes ‘models‘, in a more general way that includes individual model versions.

Regarding matched guise probing, the exact method for computing P ( x ∣ v ( t );  θ ) varies across language models and is detailed in the  Supplementary Information . For GPT4, for which computing P ( x ∣ v ( t );  θ ) for all tokens of interest was often not possible owing to restrictions imposed by the OpenAI application programming interface (API), we used a slightly modified method for some of the experiments, and this is also discussed in the  Supplementary Information . Similarly, some of the experiments could not be done for all language models because of model-specific constraints, which we highlight below. We note that there was at most one language model per experiment for which this was the case.

Covert-stereotype analysis

In the covert-stereotype analysis, the tokens x whose probabilities are measured for matched guise probing are trait adjectives from the Princeton Trilogy 29 , 30 , 31 , 34 , such as ‘aggressive’, ‘intelligent’ and ‘quiet’. We provide details about these adjectives in the  Supplementary Information . In the Princeton Trilogy, the adjectives are provided to participants in the form of a list, and participants are asked to select from the list the five adjectives that best characterize a given ethnic group, such as African Americans. The studies that we compare in this paper, which are the original Princeton Trilogy studies 29 , 30 , 31 and a more recent reinstallment 34 , all follow this general set-up and observe a gradual improvement of the expressed stereotypes about African Americans over time, but the exact interpretation of this finding is disputed 32 . Here, we used the adjectives from the Princeton Trilogy in the context of matched guise probing.

Specifically, we first computed P ( x ∣ v ( t );  θ ) for all adjectives, for both the AAE texts and the SAE texts. The method for aggregating the probabilities P ( x ∣ v ( t );  θ ) into association scores between an adjective x and AAE varies for the two settings of matched guise probing. Let \({t}_{{\rm{a}}}^{i}\) be the i -th AAE text in T a and \({t}_{{\rm{s}}}^{i}\) be the i -th SAE text in T s . In the meaning-matched setting, in which \({t}_{{\rm{a}}}^{i}\) and \({t}_{{\rm{s}}}^{i}\) express the same meaning, we computed the prompt-level association score for an adjective x as

where n = ∣ T a ∣ = ∣ T s ∣ . Thus, we measure for each pair of AAE and SAE texts the log ratio of the probability assigned to x following the AAE text and the probability assigned to x following the SAE text, and then average the log ratios of the probabilities across all pairs. In the non-meaning-matched setting, we computed the prompt-level association score for an adjective x as

where again n = ∣ T a ∣ = ∣ T s ∣ . In other words, we first compute the average probability assigned to a certain adjective x following all AAE texts and the average probability assigned to x following all SAE texts, and then measure the log ratio of these average probabilities. The interpretation of q ( x ;  v ,  θ ) is identical in both settings; q ( x ;  v , θ ) > 0 means that for a certain prompt v , the language model θ associates the adjective x more strongly with AAE than with SAE, and q ( x ;  v ,  θ ) < 0 means that for a certain prompt v , the language model θ associates the adjective x more strongly with SAE than with AAE. In the  Supplementary Information (‘Calibration’), we show that q ( x ;  v , θ ) is calibrated 104 , meaning that it does not depend on the prior probability that θ assigns to x in a neutral context.

The prompt-level association scores q ( x ;  v ,  θ ) are the basis for further analyses. We start by averaging q ( x ;  v ,  θ ) across model versions, prompts and settings, and this allows us to rank all adjectives according to their overall association with AAE for individual language models (Fig. 2a ). In this and the following adjective analyses, we focus on the five adjectives that exhibit the highest association with AAE, making it possible to consistently compare the language models with the results from the Princeton Trilogy studies, most of which do not report the full ranking of all adjectives. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Fig. 2 and Supplementary Table 4 ).

Next, we wanted to measure the agreement between language models and humans through time. To do so, we considered the five adjectives most strongly associated with African Americans for each study and evaluated how highly these adjectives are ranked by the language models. Specifically, let R l  = [ x 1 , …,  x ∣ X ∣ ] be the adjective ranking generated by a language model and \({R}_{h}^{5}\) = [ x 1 , …, x 5 ] be the ranking of the top five adjectives generated by the human participants in one of the Princeton Trilogy studies. A typical measure to evaluate how highly the adjectives from \({R}_{h}^{5}\) are ranked within R l is average precision, AP 51 . However, AP does not take the internal ranking of the adjectives in \({R}_{h}^{5}\) into account, which is not ideal for our purposes; for example, AP does not distinguish whether the top-ranked adjective for humans is on the first or on the fifth rank for a language model. To remedy this, we computed the mean average precision, MAP, for different subsets of \({R}_{h}^{5}\) ,

where \({R}_{h}^{i}\) denotes the top i adjectives from the human ranking. MAP = 1 if, and only if, the top five adjectives from \({R}_{h}^{5}\) have an exact one-to-one correspondence with the top five adjectives from R l , so, unlike AP, it takes the internal ranking of the adjectives into account. We computed an individual agreement score for each language model and prompt, so we average the q ( x ;  v ,  θ ) association scores for all model versions of a language model (GPT2, for example) and the two settings (meaning-matched and non-meaning-matched) to generate R l . Because the OpenAI API for GPT4 does not give access to the probabilities for all adjectives, we excluded GPT4 from this analysis. Results are presented in Fig. 2b and Extended Data Table 1 . In the Supplementary Information (‘Agreement analysis’), we analyse variation across model versions, settings and prompts (Supplementary Figs. 3 – 5 ).

To analyse the favourability of the stereotypes about African Americans, we drew from crowd-sourced favourability ratings collected previously 34 for the adjectives from the Princeton Trilogy that range between −2 (‘very unfavourable’, meaning very negative) and 2 (‘very favourable’, meaning very positive). For example, the favourability rating of ‘cruel’ is −1.81 and the favourability rating of ‘brilliant’ is 1.86. We computed the average favourability of the top five adjectives, weighting the favourability ratings of individual adjectives by their association scores with AAE and African Americans. More formally, let R 5 = [ x 1 , …, x 5 ] be the ranking of the top five adjectives generated by either a language model or humans. Furthermore, let f ( x ) be the favourability rating of adjective x as reported in ref. 34 , and let q ( x ) be the overall association score of adjective x with AAE or African Americans that is used to generate R 5 . For the Princeton Trilogy studies, q ( x ) is the percentage of participants who have assigned x to African Americans. For language models, q ( x ) is the average value of q ( x ;  v ,  θ ). We then computed the weighted average favourability, F , of the top five adjectives as

As a result of the weighting, the top-ranked adjective contributed more to the average than the second-ranked adjective, and so on. Results are presented in Extended Data Fig. 1 . To check for consistency, we also computed the average favourability of the top five adjectives without weighting, which yields similar results (Supplementary Fig. 6) .

Overt-stereotype analysis

The overt-stereotype analysis closely followed the methodology of the covert-stereotype analysis, with the difference being that instead of providing the language models with AAE and SAE texts, we provided them with overt descriptions of race (specifically, ‘Black’/‘black’ and ‘White’/‘white’). This methodological difference is also reflected by a different set of prompts ( Supplementary Information ). As a result, the experimental set-up is very similar to existing studies on overt racial bias in language models 4 , 7 . All other aspects of the analysis (such as computing adjective association scores) were identical to the analysis for covert stereotypes. This also holds for GPT4, for which we again could not conduct the agreement analysis.

We again present average results for the five language models in the main article. Results broken down for individual model versions are provided in the  Supplementary Information , where we also analyse variation across prompts (Supplementary Fig. 8 and Supplementary Table 5 ).

Employability analysis

The general set-up of the employability analysis was identical to the stereotype analyses: we fed text written in either AAE or SAE, embedded in prompts, into the language models and analysed the probabilities that they assigned to different continuation tokens. However, instead of trait adjectives, we considered occupations for X and also used a different set of prompts ( Supplementary Information ). We created a list of occupations, drawing from previously published lists 6 , 76 , 105 , 106 , 107 . We provided details about these occupations in the  Supplementary Information . We then computed association scores q ( x ;  v ,  θ ) between individual occupations x and AAE, following the same methodology as for computing adjective association scores, and ranked the occupations according to q ( x ;  v ,  θ ) for the language models. To probe the prestige associated with the occupations, we drew from a dataset of occupational prestige 105 that is based on the 2012 US General Social Survey and measures prestige on a scale from 1 (low prestige) to 9 (high prestige). For GPT4, we could not conduct the parts of the analysis that require scores for all occupations.

We again present average results for the five language models in the main article. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Tables 6 – 8 ).

Criminality analysis

The set-up of the criminality analysis is different from the previous experiments in that we did not compute aggregate association scores between certain tokens (such as trait adjectives) and AAE but instead asked the language models to make discrete decisions for each AAE and SAE text. More specifically, we simulated trials in which the language models were prompted to use AAE or SAE texts as evidence to make a judicial decision. We then aggregated the judicial decisions into summary statistics.

We conducted two experiments. In the first experiment, the language models were asked to determine whether a person accused of committing an unspecified crime should be acquitted or convicted. The only evidence provided to the language models was a statement made by the defendant, which was an AAE or SAE text. In the second experiment, the language models were asked to determine whether a person who committed first-degree murder should be sentenced to life or death. Similarly to the first (general conviction) experiment, the only evidence provided to the language models was a statement made by the defendant, which was an AAE or SAE text. Note that the AAE and SAE texts were the same texts as in the other experiments and did not come from a judicial context. Rather than testing how well language models could perform the tasks of predicting acquittal or conviction and life penalty or death penalty (an application of AI that we do not support), we were interested to see to what extent the decisions of the language models, made in the absence of any real evidence, were impacted by dialect. Although providing the language models with extra evidence as well as the AAE and SAE texts would have made the experiments more similar to real trials, it would have confounded the effect that dialect has on its own (the key effect of interest), so we did not consider this alternative set-up here. We focused on convictions and death penalties specifically because these are the two areas of the criminal justice system for which racial disparities have been described in the most robust and indisputable way: African Americans represent about 12% of the adult population of the United States, but they represent 33% of inmates 108 and more than 41% of people on death row 109 .

Methodologically, we used prompts that asked the language models to make a judicial decision ( Supplementary Information ). For a specific text, t , which is in AAE or SAE, we computed p ( x ∣ v ( t );  θ ) for the tokens x that correspond to the judicial outcomes of interest (‘acquitted’ or ‘convicted’, and ‘life’ or ‘death’). T5 does not contain the tokens ‘acquitted’ and ‘convicted’ in its vocabulary, so is was excluded from the conviction analysis. Because the language models might assign different prior probabilities to the outcome tokens, we calibrated them using their probabilities in a neutral context following v , meaning without text t 104 . Whichever outcome had the higher calibrated probability was counted as the decision. We aggregated the detrimental decisions (convictions and death penalties) and compared their rates (percentages) between AAE and SAE texts. An alternative approach would have been to generate the judicial decision by sampling from the language models, which would have allowed us to induce the language models to generate justifications of their decisions. However, this approach has three disadvantages: first, encoder-only language models such as RoBERTa do not lend themselves to text generation; second, it would have been necessary to apply jail-breaking for some of the language models, which can have unpredictable effects, especially in the context of socially sensitive tasks; and third, model-generated justifications are frequently not aligned with actual model behaviours 110 .

We again present average results on the level of language models in the main article. Results for individual model versions are provided in the  Supplementary Information , where we also analyse variation across settings and prompts (Supplementary Figs. 9 and 10 and Supplementary Tables 9 – 12 ).

Scaling analysis

In the scaling analysis, we examined whether increasing the model size alleviated the dialect prejudice. Because the content of the covert stereotypes is quite consistent and does not vary substantially between models with different sizes, we instead analysed the strength with which the language models maintain these stereotypes. We split the model versions of all language models into four groups according to their size using the thresholds of 1.5 × 10 8 , 3.5 × 10 8 and 1.0 × 10 10 (Extended Data Table 7 ).

To evaluate the familiarity of the models with AAE, we measured their perplexity on the datasets used for the two evaluation settings 83 , 87 . Perplexity is defined as the exponentiated average negative log-likelihood of a sequence of tokens 111 , with lower values indicating higher familiarity. Perplexity requires the language models to assign probabilities to full sequences of tokens, which is only the case for GPT2 and GPT3.5. For RoBERTa and T5, we resorted to pseudo-perplexity 112 as the measure of familiarity. Results are only comparable across language models with the same familiarity measure. We excluded GPT4 from this analysis because it is not possible to compute perplexity using the OpenAI API.

To evaluate the stereotype strength, we focused on the stereotypes about African Americans reported in ref. 29 , which the language models’ covert stereotypes agree with most strongly. We split the set of adjectives X into two subsets: the set of stereotypical adjectives in ref. 29 , X s , and the set of non-stereotypical adjectives, X n  =  X \ X s . For each model with a specific size, we then computed the average value of q ( x ;  v ,  θ ) for all adjectives in X s , which we denote as q s ( θ ), and the average value of q ( x ;  v ,  θ ) for all adjectives in X n , which we denote as q n ( θ ). The stereotype strength of a model θ , or more specifically the strength of the stereotypes about African Americans reported in ref. 29 , can then be computed as

A positive value of δ ( θ ) means that the model associates the stereotypical adjectives in X s more strongly with AAE than the non-stereotypical adjectives in X n , whereas a negative value of δ ( θ ) indicates anti-stereotypical associations, meaning that the model associates the non-stereotypical adjectives in X n more strongly with AAE than the stereotypical adjectives in X s . For the overt stereotypes, we used the same split of adjectives into X s and X n because we wanted to directly compare the strength with which models of a certain size endorse the stereotypes overtly as opposed to covertly. All other aspects of the experimental set-up are identical to the main analyses of covert and overt stereotypes.

HF analysis

We compared GPT3.5 (ref. 49 ; text-davinci-003) with GPT3 (ref. 63 ; davinci), its predecessor language model that was trained without HF. Similarly to other studies that compare these two language models 113 , this set-up allowed us to examine the effects of HF training as done for GPT3.5 in isolation. We compared the two language models in terms of favourability and stereotype strength. For favourability, we followed the methodology we used for the overt-stereotype analysis and evaluated the average weighted favourability of the top five adjectives associated with AAE. For stereotype strength, we followed the methodology we used for the scaling analysis and evaluated the average strength of the stereotypes as reported in ref.  29 .

Reporting summary

Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

Data availability

All the datasets used in this study are publicly available. The dataset released as ref. 87 can be found at https://aclanthology.org/2020.emnlp-main.473/ . The dataset released as ref. 83 can be found at http://slanglab.cs.umass.edu/TwitterAAE/ . The human stereotype scores used for evaluation can be found in the published articles of the Princeton Trilogy studies 29 , 30 , 31 , 34 . The most recent of these articles 34 also contains the human favourability scores for the trait adjectives. The dataset of occupational prestige that we used for the employability analysis can be found in the corresponding paper 105 . The Brown Corpus 114 , which we used for the  Supplementary Information (‘Feature analysis’), can be found at http://www.nltk.org/nltk_data/ . The dataset containing the parallel AAE, Appalachian English and Indian English texts 115 , which we used in the  Supplementary Information (‘Alternative explanations’), can be found at https://huggingface.co/collections/SALT-NLP/value-nlp-666b60a7f76c14551bda4f52 .

Code availability

Our code is written in Python and draws on the Python packages openai and transformers for language-model probing, as well as numpy, pandas, scipy and statsmodels for data analysis. The feature analysis described in the  Supplementary Information also uses the VALUE Python library 88 . Our code is publicly available on GitHub at https://github.com/valentinhofmann/dialect-prejudice .

Zhao, W. et al. WildChat: 1M ChatGPT interaction logs in the wild. In Proc. Twelfth International Conference on Learning Representations (OpenReview.net, 2024).

Zheng, L. et al. LMSYS-Chat-1M: a large-scale real-world LLM conversation dataset. In Proc. Twelfth International Conference on Learning Representations (OpenReview.net, 2024).

Gaebler, J. D., Goel, S., Huq, A. & Tambe, P. Auditing the use of language models to guide hiring decisions. Preprint at https://arxiv.org/abs/2404.03086 (2024).

Sheng, E., Chang, K.-W., Natarajan, P. & Peng, N. The woman worked as a babysitter: on biases in language generation. In Proc. 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing (eds Inui. K. et al.) 3407–3412 (Association for Computational Linguistics, 2019).

Nangia, N., Vania, C., Bhalerao, R. & Bowman, S. R. CrowS-Pairs: a challenge dataset for measuring social biases in masked language models. In Proc. 2020 Conference on Empirical Methods in Natural Language Processing (eds Webber, B. et al.) 1953–1967 (Association for Computational Linguistics, 2020).

Nadeem, M., Bethke, A. & Reddy, S. StereoSet: measuring stereotypical bias in pretrained language models. In Proc. 59th Annual Meeting of the Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (eds Zong, C. et al.) 5356–5371 (Association for Computational Linguistics, 2021).

Cheng, M., Durmus, E. & Jurafsky, D. Marked personas: using natural language prompts to measure stereotypes in language models. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 1504–1532 (Association for Computational Linguistics, 2023).

Bonilla-Silva, E. Racism without Racists: Color-Blind Racism and the Persistence of Racial Inequality in America 4th edn (Rowman & Littlefield, 2014).

Golash-Boza, T. A critical and comprehensive sociological theory of race and racism. Sociol. Race Ethn. 2 , 129–141 (2016).

Article   Google Scholar  

Kasneci, E. et al. ChatGPT for good? On opportunities and challenges of large language models for education. Learn. Individ. Differ. 103 , 102274 (2023).

Nay, J. J. et al. Large language models as tax attorneys: a case study in legal capabilities emergence. Philos. Trans. R. Soc. A 382 , 20230159 (2024).

Article   ADS   Google Scholar  

Jiang, L. Y. et al. Health system-scale language models are all-purpose prediction engines. Nature 619 , 357–362 (2023).

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Bolukbasi, T., Chang, K.-W., Zou, J., Saligrama, V. & Kalai, A. Man is to computer programmer as woman is to homemaker? Debiasing word embeddings. Adv. Neural Inf. Process. Syst. 30 , 4356–4364 (2016).

Google Scholar  

Caliskan, A., Bryson, J. J. & Narayanan, A. Semantics derived automatically from language corpora contain human-like biases. Science 356 , 183–186 (2017).

Article   ADS   CAS   PubMed   Google Scholar  

Basta, C., Costa-jussà, M. R. & Casas, N. Evaluating the underlying gender bias in contextualized word embeddings. In Proc. First Workshop on Gender Bias in Natural Language Processing (eds Costa-jussà, M. R. et al.) 33–39 (Association for Computational Linguistics, 2019).

Kurita, K., Vyas, N., Pareek, A., Black, A. W. & Tsvetkov, Y. Measuring bias in contextualized word representations. In Proc. First Workshop on Gender Bias in Natural Language Processing (eds Costa-jussà, M. R. et al.) 166–172 (Association for Computational Linguistics, 2019).

Abid, A., Farooqi, M. & Zou, J. Persistent anti-muslim bias in large language models. In Proc. 2021 AAAI/ACM Conference on AI, Ethics, and Society (eds Fourcade, M. et al.) 298–306 (Association for Computing Machinery, 2021).

Bender, E. M., Gebru, T., McMillan-Major, A. & Shmitchell, S. On the dangers of stochastic parrots: can language models be too big? In Proc. 2021 ACM Conference on Fairness, Accountability, and Transparency 610–623 (Association for Computing Machinery, 2021).

Li, L. & Bamman, D. Gender and representation bias in GPT-3 generated stories. In Proc. Third Workshop on Narrative Understanding (eds Akoury, N. et al.) 48–55 (Association for Computational Linguistics, 2021).

Tamkin, A. et al. Evaluating and mitigating discrimination in language model decisions. Preprint at https://arxiv.org/abs/2312.03689 (2023).

Rae, J. W. et al. Scaling language models: methods, analysis & insights from training Gopher. Preprint at https://arxiv.org/abs/2112.11446 (2021).

Green, L. J. African American English: A Linguistic Introduction (Cambridge Univ. Press, 2002).

King, S. From African American Vernacular English to African American Language: rethinking the study of race and language in African Americans’ speech. Annu. Rev. Linguist. 6 , 285–300 (2020).

Purnell, T., Idsardi, W. & Baugh, J. Perceptual and phonetic experiments on American English dialect identification. J. Lang. Soc. Psychol. 18 , 10–30 (1999).

Massey, D. S. & Lundy, G. Use of Black English and racial discrimination in urban housing markets: new methods and findings. Urban Aff. Rev. 36 , 452–469 (2001).

Dunbar, A., King, S. & Vaughn, C. Dialect on trial: an experimental examination of raciolinguistic ideologies and character judgments. Race Justice https://doi.org/10.1177/21533687241258772 (2024).

Rickford, J. R. & King, S. Language and linguistics on trial: Hearing Rachel Jeantel (and other vernacular speakers) in the courtroom and beyond. Language 92 , 948–988 (2016).

Grogger, J. Speech patterns and racial wage inequality. J. Hum. Resour. 46 , 1–25 (2011).

Katz, D. & Braly, K. Racial stereotypes of one hundred college students. J. Abnorm. Soc. Psychol. 28 , 280–290 (1933).

Gilbert, G. M. Stereotype persistance and change among college students. J. Abnorm. Soc. Psychol. 46 , 245–254 (1951).

Article   CAS   Google Scholar  

Karlins, M., Coffman, T. L. & Walters, G. On the fading of social stereotypes: studies in three generations of college students. J. Pers. Soc. Psychol. 13 , 1–16 (1969).

Article   CAS   PubMed   Google Scholar  

Devine, P. G. & Elliot, A. J. Are racial stereotypes really fading? The Princeton Trilogy revisited. Pers. Soc. Psychol. Bull. 21 , 1139–1150 (1995).

Madon, S. et al. Ethnic and national stereotypes: the Princeton Trilogy revisited and revised. Pers. Soc. Psychol. Bull. 27 , 996–1010 (2001).

Bergsieker, H. B., Leslie, L. M., Constantine, V. S. & Fiske, S. T. Stereotyping by omission: eliminate the negative, accentuate the positive. J. Pers. Soc. Psychol. 102 , 1214–1238 (2012).

Article   PubMed   PubMed Central   Google Scholar  

Ghavami, N. & Peplau, L. A. An intersectional analysis of gender and ethnic stereotypes: testing three hypotheses. Psychol. Women Q. 37 , 113–127 (2013).

Lambert, W. E., Hodgson, R. C., Gardner, R. C. & Fillenbaum, S. Evaluational reactions to spoken languages. J. Abnorm. Soc. Psychol. 60 , 44–51 (1960).

Ball, P. Stereotypes of Anglo-Saxon and non-Anglo-Saxon accents: some exploratory Australian studies with the matched guise technique. Lang. Sci. 5 , 163–183 (1983).

Thomas, E. R. & Reaser, J. Delimiting perceptual cues used for the ethnic labeling of African American and European American voices. J. Socioling. 8 , 54–87 (2004).

Atkins, C. P. Do employment recruiters discriminate on the basis of nonstandard dialect? J. Employ. Couns. 30 , 108–118 (1993).

Payne, K., Downing, J. & Fleming, J. C. Speaking Ebonics in a professional context: the role of ethos/source credibility and perceived sociability of the speaker. J. Tech. Writ. Commun. 30 , 367–383 (2000).

Rodriguez, J. I., Cargile, A. C. & Rich, M. D. Reactions to African-American vernacular English: do more phonological features matter? West. J. Black Stud. 28 , 407–414 (2004).

Billings, A. C. Beyond the Ebonics debate: attitudes about Black and standard American English. J. Black Stud. 36 , 68–81 (2005).

Kurinec, C. A. & Weaver, C. III “Sounding Black”: speech stereotypicality activates racial stereotypes and expectations about appearance. Front. Psychol. 12 , 785283 (2021).

Rosa, J. & Flores, N. Unsettling race and language: toward a raciolinguistic perspective. Lang. Soc. 46 , 621–647 (2017).

Salehi, B., Hovy, D., Hovy, E. & Søgaard, A. Huntsville, hospitals, and hockey teams: names can reveal your location. In Proc. 3rd Workshop on Noisy User-generated Text (eds Derczynski, L. et al.) 116–121 (Association for Computational Linguistics, 2017).

Radford, A. et al. Language models are unsupervised multitask learners. OpenAI https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf (2019).

Liu, Y. et al. RoBERTa: a robustly optimized BERT pretraining approach. Preprint at https://arxiv.org/abs/1907.11692 (2019).

Raffel, C. et al. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res. 21 , 1–67 (2020).

MathSciNet   Google Scholar  

Ouyang, L. et al. Training language models to follow instructions with human feedback. In Proc. 36th Conference on Neural Information Processing Systems (eds Koyejo, S. et al.) 27730–27744 (NeurIPS, 2022).

OpenAI et al. GPT-4 technical report. Preprint at https://arxiv.org/abs/2303.08774 (2023).

Zhang, E. & Zhang, Y. Average precision. In Encyclopedia of Database Systems (eds Liu, L. & Özsu, M. T.) 192–193 (Springer, 2009).

Black, J. S. & van Esch, P. AI-enabled recruiting: what is it and how should a manager use it? Bus. Horiz. 63 , 215–226 (2020).

Hunkenschroer, A. L. & Luetge, C. Ethics of AI-enabled recruiting and selection: a review and research agenda. J. Bus. Ethics 178 , 977–1007 (2022).

Upadhyay, A. K. & Khandelwal, K. Applying artificial intelligence: implications for recruitment. Strateg. HR Rev. 17 , 255–258 (2018).

Tippins, N. T., Oswald, F. L. & McPhail, S. M. Scientific, legal, and ethical concerns about AI-based personnel selection tools: a call to action. Pers. Assess. Decis. 7 , 1 (2021).

Aletras, N., Tsarapatsanis, D., Preoţiuc-Pietro, D. & Lampos, V. Predicting judicial decisions of the European Court of Human Rights: a natural language processing perspective. PeerJ Comput. Sci. 2 , e93 (2016).

Surden, H. Artificial intelligence and law: an overview. Ga State Univ. Law Rev. 35 , 1305–1337 (2019).

Medvedeva, M., Vols, M. & Wieling, M. Using machine learning to predict decisions of the European Court of Human Rights. Artif. Intell. Law 28 , 237–266 (2020).

Weidinger, L. et al. Taxonomy of risks posed by language models. In Proc. 2022 ACM Conference on Fairness, Accountability, and Transparency 214–229 (Association for Computing Machinery, 2022).

Czopp, A. M. & Monteith, M. J. Thinking well of African Americans: measuring complimentary stereotypes and negative prejudice. Basic Appl. Soc. Psychol. 28 , 233–250 (2006).

Chowdhery, A. et al. PaLM: scaling language modeling with pathways. J. Mach. Learn. Res. 24 , 11324–11436 (2023).

Bai, Y. et al. Training a helpful and harmless assistant with reinforcement learning from human feedback. Preprint at https://arxiv.org/abs/2204.05862 (2022).

Brown, T. B. et al. Language models are few-shot learners. In  Proc. 34th International Conference on Neural Information Processing Systems  (eds Larochelle, H. et al.) 1877–1901 (NeurIPS, 2020).

Dovidio, J. F. & Gaertner, S. L. Aversive racism. Adv. Exp. Soc. Psychol. 36 , 1–52 (2004).

Schuman, H., Steeh, C., Bobo, L. D. & Krysan, M. (eds) Racial Attitudes in America: Trends and Interpretations (Harvard Univ. Press, 1998).

Crosby, F., Bromley, S. & Saxe, L. Recent unobtrusive studies of Black and White discrimination and prejudice: a literature review. Psychol. Bull. 87 , 546–563 (1980).

Terkel, S. Race: How Blacks and Whites Think and Feel about the American Obsession (New Press, 1992).

Jackman, M. R. & Muha, M. J. Education and intergroup attitudes: moral enlightenment, superficial democratic commitment, or ideological refinement? Am. Sociol. Rev. 49 , 751–769 (1984).

Bonilla-Silva, E. The New Racism: Racial Structure in the United States, 1960s–1990s. In Race, Ethnicity, and Nationality in the United States: Toward the Twenty-First Century 1st edn (ed. Wong, P.) Ch. 4 (Westview Press, 1999).

Gao, L. et al. The Pile: an 800GB dataset of diverse text for language modeling. Preprint at https://arxiv.org/abs/2101.00027 (2021).

Ronkin, M. & Karn, H. E. Mock Ebonics: linguistic racism in parodies of Ebonics on the internet. J. Socioling. 3 , 360–380 (1999).

Dodge, J. et al. Documenting large webtext corpora: a case study on the Colossal Clean Crawled Corpus. In Proc. 2021 Conference on Empirical Methods in Natural Language Processing (eds Moens, M.-F. et al.) 1286–1305 (Association for Computational Linguistics, 2021).

Steed, R., Panda, S., Kobren, A. & Wick, M. Upstream mitigation is not all you need: testing the bias transfer hypothesis in pre-trained language models. In Proc. 60th Annual Meeting of the Association for Computational Linguistics (eds Muresan, S. et al.) 3524–3542 (Association for Computational Linguistics, 2022).

Feng, S., Park, C. Y., Liu, Y. & Tsvetkov, Y. From pretraining data to language models to downstream tasks: tracking the trails of political biases leading to unfair NLP models. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 11737–11762 (Association for Computational Linguistics, 2023).

Köksal, A. et al. Language-agnostic bias detection in language models with bias probing. In Findings of the Association for Computational Linguistics: EMNLP 2023 (eds Bouamor, H. et al.) 12735–12747 (Association for Computational Linguistics, 2023).

Garg, N., Schiebinger, L., Jurafsky, D. & Zou, J. Word embeddings quantify 100 years of gender and ethnic stereotypes. Proc. Natl Acad. Sci. USA 115 , E3635–E3644 (2018).

Ferrer, X., van Nuenen, T., Such, J. M. & Criado, N. Discovering and categorising language biases in Reddit. In Proc. Fifteenth International AAAI Conference on Web and Social Media (eds Budak, C. et al.) 140–151 (Association for the Advancement of Artificial Intelligence, 2021).

Ethayarajh, K., Choi, Y. & Swayamdipta, S. Understanding dataset difficulty with V-usable information. In Proc. 39th International Conference on Machine Learning (eds Chaudhuri, K. et al.) 5988–6008 (Proceedings of Machine Learning Research, 2022).

Hoffmann, J. et al. Training compute-optimal large language models. Preprint at https://arxiv.org/abs/2203.15556 (2022).

Liang, P. et al. Holistic evaluation of language models. Transactions on Machine Learning Research https://openreview.net/forum?id=iO4LZibEqW (2023).

Blodgett, S. L., Barocas, S., Daumé III, H. & Wallach, H. Language (technology) is power: A critical survey of “bias” in NLP. In Proc. 58th Annual Meeting of the Association for Computational Linguistics (eds Jurafsky, D. et al.) 5454–5476 (Association for Computational Linguistics, 2020).

Jørgensen, A., Hovy, D. & Søgaard, A. Challenges of studying and processing dialects in social media. In Proc. Workshop on Noisy User-generated Text (eds Xu, W. et al.) 9–18 (Association for Computational Linguistics, 2015).

Blodgett, S. L., Green, L. & O’Connor, B. Demographic dialectal variation in social media: a case study of African-American English. In Proc. 2016 Conference on Empirical Methods in Natural Language Processing (eds Su, J. et al.) 1119–1130 (Association for Computational Linguistics, 2016).

Jørgensen, A., Hovy, D. & Søgaard, A. Learning a POS tagger for AAVE-like language. In Proc. 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Knight, K. et al.) 1115–1120 (Association for Computational Linguistics, 2016).

Blodgett, S. L. & O’Connor, B. Racial disparity in natural language processing: a case study of social media African-American English. Preprint at https://arxiv.org/abs/1707.00061 (2017).

Blodgett, S. L., Wei, J. & O’Connor, B. Twitter universal dependency parsing for African-American and mainstream American English. In Proc. 56th Annual Meeting of the Association for Computational Linguistics (eds Gurevych, I. & Miyao, Y.) 1415–1425 (Association for Computational Linguistics, 2018).

Groenwold, S. et al. Investigating African-American vernacular English in transformer-based text generation. In Proc. 2020 Conference on Empirical Methods in Natural Language Processing (eds Webber, B. et al.) 5877–5883 (Association for Computational Linguistics, 2020).

Ziems, C., Chen, J., Harris, C., Anderson, J. & Yang, D. VALUE: Understanding dialect disparity in NLU. In Proc. 60th Annual Meeting of the Association for Computational Linguistics (eds Muresan, S. et al.) 3701–3720 (Association for Computational Linguistics, 2022).

Davidson, T., Bhattacharya, D. & Weber, I. Racial bias in hate speech and abusive language detection datasets. In Proc. Third Workshop on Abusive Language Online (eds Roberts, S. T. et al.) 25–35 (Association for Computational Linguistics, 2019).

Sap, M., Card, D., Gabriel, S., Choi, Y. & Smith, N. A. The risk of racial bias in hate speech detection. In Proc. 57th Annual Meeting of the Association for Computational Linguistics (eds Korhonen, A. et al.) 1668–1678 (Association for Computational Linguistics, 2019).

Harris, C., Halevy, M., Howard, A., Bruckman, A. & Yang, D. Exploring the role of grammar and word choice in bias toward African American English (AAE) in hate speech classification. In Proc. 2022 ACM Conference on Fairness, Accountability, and Transparency 789–798 (Association for Computing Machinery, 2022).

Gururangan, S. et al. Whose language counts as high quality? Measuring language ideologies in text data selection. In Proc. 2022 Conference on Empirical Methods in Natural Language Processing (eds Goldberg, Y. et al.) 2562–2580 (Association for Computational Linguistics, 2022).

Gaies, S. J. & Beebe, J. D. The matched-guise technique for measuring attitudes and their implications for language education: a critical assessment. In Language Acquisition and the Second/Foreign Language Classroom (ed. Sadtano, E.) 156–178 (SEAMEO Regional Language Centre, 1991).

Hudson, R. A. Sociolinguistics (Cambridge Univ. Press, 1996).

Delobelle, P., Tokpo, E., Calders, T. & Berendt, B. Measuring fairness with biased rulers: a comparative study on bias metrics for pre-trained language models. In Proc. 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Carpuat, M. et al.) 1693–1706 (Association for Computational Linguistics, 2022).

Mattern, J., Jin, Z., Sachan, M., Mihalcea, R. & Schölkopf, B. Understanding stereotypes in language models: Towards robust measurement and zero-shot debiasing. Preprint at https://arxiv.org/abs/2212.10678 (2022).

Eisenstein, J., O’Connor, B., Smith, N. A. & Xing, E. P. A latent variable model for geographic lexical variation. In Proc. 2010 Conference on Empirical Methods in Natural Language Processing (eds Li, H. & Màrquez, L.) 1277–1287 (Association for Computational Linguistics, 2010).

Doyle, G. Mapping dialectal variation by querying social media. In Proc. 14th Conference of the European Chapter of the Association for Computational Linguistics (eds Wintner, S. et al.) 98–106 (Association for Computational Linguistics, 2014).

Huang, Y., Guo, D., Kasakoff, A. & Grieve, J. Understanding U.S. regional linguistic variation with Twitter data analysis. Comput. Environ. Urban Syst. 59 , 244–255 (2016).

Eisenstein, J. What to do about bad language on the internet. In Proc. 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Vanderwende, L. et al.) 359–369 (Association for Computational Linguistics, 2013).

Eisenstein, J. Systematic patterning in phonologically-motivated orthographic variation. J. Socioling. 19 , 161–188 (2015).

Jones, T. Toward a description of African American vernacular English dialect regions using “Black Twitter”. Am. Speech 90 , 403–440 (2015).

Christiano, P. F. et al. Deep reinforcement learning from human preferences. Proc. 31st International Conference on Neural Information Processing Systems (eds von Luxburg, U. et al.) 4302–4310 (NeurIPS, 2017).

Zhao, T. Z., Wallace, E., Feng, S., Klein, D. & Singh, S. Calibrate before use: Improving few-shot performance of language models. In Proc. 38th International Conference on Machine Learning (eds Meila, M. & Zhang, T.) 12697–12706 (Proceedings of Machine Learning Research, 2021).

Smith, T. W. & Son, J. Measuring Occupational Prestige on the 2012 General Social Survey (NORC at Univ. Chicago, 2014).

Zhao, J., Wang, T., Yatskar, M., Ordonez, V. & Chang, K.-W. Gender bias in coreference resolution: evaluation and debiasing methods. In Proc. 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (eds Walker, M. et al.) 15–20 (Association for Computational Linguistics, 2018).

Hughes, B. T., Srivastava, S., Leszko, M. & Condon, D. M. Occupational prestige: the status component of socioeconomic status. Collabra Psychol. 10 , 92882 (2024).

Gramlich, J. The gap between the number of blacks and whites in prison is shrinking. Pew Research Centre https://www.pewresearch.org/short-reads/2019/04/30/shrinking-gap-between-number-of-blacks-and-whites-in-prison (2019).

Walsh, A. The criminal justice system is riddled with racial disparities. Prison Policy Initiative Briefing https://www.prisonpolicy.org/blog/2016/08/15/cjrace (2016).

Röttger, P. et al. Political compass or spinning arrow? Towards more meaningful evaluations for values and opinions in large language models. Preprint at https://arxiv.org/abs/2402.16786 (2024).

Jurafsky, D. & Martin, J. H. Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition (Prentice Hall, 2000).

Salazar, J., Liang, D., Nguyen, T. Q. & Kirchhoff, K. Masked language model scoring. In Proc. 58th Annual Meeting of the Association for Computational Linguistics (eds Jurafsky, D. et al.) 2699–2712 (Association for Computational Linguistics, 2020).

Santurkar, S. et al. Whose opinions do language models reflect? In Proc. 40th International Conference on Machine Learning (eds Krause, A. et al.) 29971–30004 (Proceedings of Machine Learning Research, 2023).

Francis, W. N. & Kucera, H. Brown Corpus Manual (Brown Univ.,1979).

Ziems, C. et al. Multi-VALUE: a framework for cross-dialectal English NLP. In Proc. 61st Annual Meeting of the Association for Computational Linguistics (eds Rogers, A. et al.) 744–768 (Association for Computational Linguistics, 2023).

Download references

Acknowledgements

V.H. was funded by the German Academic Scholarship Foundation. P.R.K. was funded in part by the Open Phil AI Fellowship. This work was also funded by the Hoffman-Yee Research Grants programme and the Stanford Institute for Human-Centered Artificial Intelligence. We thank A. Köksal, D. Hovy, K. Gligorić, M. Harrington, M. Casillas, M. Cheng and P. Röttger for feedback on an earlier version of the article.

Author information

Authors and affiliations.

Allen Institute for AI, Seattle, WA, USA

Valentin Hofmann

University of Oxford, Oxford, UK

LMU Munich, Munich, Germany

Stanford University, Stanford, CA, USA

Pratyusha Ria Kalluri & Dan Jurafsky

The University of Chicago, Chicago, IL, USA

Sharese King

You can also search for this author in PubMed   Google Scholar

Contributions

V.H., P.R.K., D.J. and S.K. designed the research. V.H. performed the research and analysed the data. V.H., P.R.K., D.J. and S.K. wrote the paper.

Corresponding authors

Correspondence to Valentin Hofmann or Sharese King .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Peer review

Peer review information.

Nature thanks Rodney Coates and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended data fig. 1 weighted average favourability of top stereotypes about african americans in humans and top overt as well as covert stereotypes about african americans in language models (lms)..

The overt stereotypes are more favourable than the reported human stereotypes, except for GPT2. The covert stereotypes are substantially less favourable than the least favourable reported human stereotypes from 1933. Results without weighting, which are very similar, are provided in Supplementary Fig. 6 .

Extended Data Fig. 2 Prestige of occupations associated with AAE (positive values) versus SAE (negative values), for individual language models.

The shaded areas show 95% confidence bands around the regression lines. The association with AAE versus SAE is negatively correlated with occupational prestige, for all language models. We cannot conduct this analysis with GPT4 since the OpenAI API does not give access to the probabilities for all occupations.

Supplementary information

Supplementary information, reporting summary, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Hofmann, V., Kalluri, P.R., Jurafsky, D. et al. AI generates covertly racist decisions about people based on their dialect. Nature (2024). https://doi.org/10.1038/s41586-024-07856-5

Download citation

Received : 08 February 2024

Accepted : 19 July 2024

Published : 28 August 2024

DOI : https://doi.org/10.1038/s41586-024-07856-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.