Prejudice: How a Generalization Hardens Into an Attitude

The word has a second, unrelated meaning in law, where a case dismissed with prejudice cannot be brought again; this page is about the psychological sense, an attitude toward people based on their group. Allport's definition, which still frames the field, was an antipathy based on a faulty and inflexible generalization, felt or expressed, directed at a group or at an individual because they belong to it. Antipathy makes it an attitude rather than a belief. Generalization ties it to the ordinary act of putting people in categories. Inflexible is what separates prejudice from prejudgment: everyone forms expectations about groups from limited experience, and a prejudgment becomes a prejudice only when it stops being corrected by new information.

That link to categorization is why prejudice sits in the codex quadrant about what memory keeps: the mind stores groups and the traits that go with them, and fills in a new person from the group. Allport called this the normality of prejudgment. The stereotype is the content of the category; prejudice is the feeling attached to it; discrimination is the behavior that follows. The three usually travel together but they are not the same, and a person can hold a stereotype without hostility, or act on a rule without feeling anything at all.

Group membership alone is enough to start it. Tajfel and colleagues showed in 1971 that schoolboys sorted into two groups on a trivial pretext, supposedly by their estimates of dots or their taste in abstract painters and in fact at random, gave more of a reward worth real money to anonymous members of their own group than to the other. The boys knew one another from school; what was missing was any history between groups that had not existed an hour earlier. In-group favoritism of that kind is the seed; the fully formed attitude adds beliefs about the out-group, usually that it is more uniform, more threatening or less deserving than it is, and the emotion of contempt, fear or disgust.

Allport's scale is a claim about how the attitude is expressed and how expression escalates. Antilocution, hostile talk among the like-minded, is the mildest form and the commonest. Avoidance costs the prejudiced person something, since they rearrange their own life to keep away. Discrimination denies the group opportunities. Physical attack and extermination are the end of the ladder. Allport did not claim that everyone climbs it, or that any society must; he claimed that each rung lowers the barrier to the next, and his illustration was a case in which the whole ladder had been climbed.

Gordon Allport published The Nature of Prejudice in 1954, the year the United States Supreme Court decided Brown v. Board of Education, and the book organized a scattered literature into a field. It is a synthesis of about 500 pages rather than a single experiment. Its first chapter sets out the definition and the scale; later chapters cover in-groups, stereotypes, the family and the media, and what reduces the attitude once formed.

The scale appears in the first chapter as a way of ordering what prejudice looks like in action, and Allport noted there that many people never go beyond the mildest rung, hostile talk among the like-minded. He put no figure on how often one rung leads to the next. His illustration was the recent history of Germany, traced from hostile talk to avoidance of Jewish neighbors, to the discrimination of the Nuremberg laws, to the burning of synagogues and street attacks, and finally to the camps. The point was that the lower rungs are not harmless because they are mild; they are the conditions under which the higher rungs become thinkable.

The book's most tested idea is the contact hypothesis, set out in chapter 16: contact between groups reduces prejudice most reliably when the groups meet as equals, pursue common goals, cooperate rather than compete, and have the support of authorities and custom. Pettigrew and Tropp's 2006 meta-analysis pooled 515 studies with more than 250,000 participants and found that contact was associated with lower prejudice, with a mean correlation of about negative 0.21. The association was stronger when Allport's conditions were met, but held without them; the authors concluded that the conditions help but are not necessary. A 2019 review by Paluck, Green and Green added a caution: few of the studies are randomized, their results vary widely, and very few test adult racial or ethnic prejudice, the case the hypothesis was written about.

Most prejudice is expressed on the lower rungs, in settings where it is easy to miss.

Prejudice is reduced by changing conditions more reliably than by changing minds directly, and Allport's own prescriptions have held up better than most.

• Arrange contact that meets the conditions. Equal status, a shared goal, cooperation and institutional support are the conditions under which contact reduces prejudice most reliably; contact without them helps less and can go wrong. A mixed team with a common deadline does more than a diversity lecture.
• Individuate. Learning one real person's name, history and particulars makes the category harder to apply to them, and the habit of individuating generalizes.
• Test the generalization against the record. For any belief about a group, ask what the base rate is and whether the belief would survive the data; prejudgments that cannot be revised are the ones Allport defined as prejudice.
• Interrupt the first rung. Antilocution is where the scale starts and where it is cheapest to stop. Declining to laugh, or naming what was said, raises the cost of saying it again.
• Change the rules rather than the hearts. Anonymized applications, structured interviews and audited lending decisions reduce discrimination whatever the decision makers feel.

Prejudice is the attitude; its neighbors are the beliefs, behaviors and biases that surround it.

• It is not stereotyping, which is the belief about what members of a group are like. A stereotype supplies the content; prejudice adds the hostility. The two usually appear together but can come apart.
• It is not in-group favoritism, the preference for one's own group that Tajfel found in groups created minutes earlier on a trivial pretext. Favoring your own group is the starting point; prejudice requires a negative attitude toward the other.
• It is not out-group homogeneity bias, the perception that members of other groups are more alike than members of your own. That bias feeds prejudice by making the category feel accurate, but it is a judgment about variability rather than an attitude.
• It is not discrimination, which is the behavior. Allport's scale puts discrimination at the third rung precisely because prejudice can exist without it, in talk and avoidance.