Before anything else, get the three words straight, because most public arguments about bias collapse when people mix them up. Prejudice is an attitude (how you feel). A stereotype is a belief (what you think is true of a group). Discrimination is a behavior (what you actually do). One person can carry all three, or any one without the others. Hold that triad in your head and the rest of this page snaps into place.
Bodhi says
Here's the memory hook: Affect, Cognition, Behavior — the ABCs of an attitude. Prejudice is the A (affect/feeling), the stereotype is the C (cognition/belief), and discrimination is the B (behavior). Any time you're asked "is this prejudice or discrimination?", the real question underneath is: is this a feeling in someone's head, or an action out in the world?
The load-bearing distinction
Prejudice = attitude. Stereotype = belief. Discrimination = behavior. Gordon Allport (1954) defined prejudice as "an antipathy based upon a faulty and inflexible generalization" — a felt hostility toward a group. A stereotype is the cognitive content that often feeds it: a generalization assigning traits to nearly all members of a group regardless of the variation among them. Discrimination is the overt act — unjustified negative treatment of a person because of their group. These usually travel together, but not always: you can hold a stereotype you consciously reject (a belief without the feeling), or discriminate for reasons of policy or pressure without personal animus (a behavior without the attitude). Whenever this topic gets confusing, ask which of the three you're actually looking at.
A demonstration: the blue-eyed, brown-eyed classroom
The morning after Martin Luther King Jr. was assassinated in 1968, an Iowa third-grade teacher named Jane Elliott decided her all-white students needed to feel what prejudice does rather than hear about it. She split the class by eye color, told the brown-eyed children they were superior, gave them privileges, and watched. Within hours the "superior" group grew arrogant and domineering; the "inferior" group turned timid, and — the finding that still unsettles teachers — their academic performance dropped. The next day she reversed the roles and the pattern flipped with them. Elliott's exercise isn't a controlled experiment, and it wouldn't clear a modern ethics board (see Research Ethics), but as a demonstration it makes a real point: status and stigma get absorbed fast, and being cast as the lesser group can degrade the very performance that then seems to "confirm" the ranking. Keep that mechanism in mind — you'll meet it again under stereotype threat.
When the ugly stuff went underground: modern & aversive racism
Openly expressed prejudice has declined sharply over the past several decades, at least in what people are willing to say out loud. But social psychologists argue it didn't so much vanish as change shape. Samuel Gaertner and John Dovidio's theory of aversive racism describes people who sincerely endorse egalitarian values and would be horrified to be called racist, yet still harbor unacknowledged negative associations that leak out — especially when a situation is ambiguous enough to supply a non-racial excuse (Dovidio & Gaertner, 2004). In their studies, white participants evaluating job candidates did not discriminate when a candidate's qualifications were clearly strong or clearly weak; the bias surfaced precisely when the record was mixed and a plausible, race-neutral rationale was available. The "aversive" in the name refers to these individuals' aversion to seeing themselves as prejudiced — which is exactly why the bias hides in ambiguity.
A closely related idea is the microaggression: the small, often unintended slights and indignities that communicate a demeaning message to members of stigmatized groups — asking an Asian American "where are you really from," or a colleague of color "how did you get this job." The construct captures something many people report experiencing as a real cumulative burden. In fairness to the science, it's also a younger and more contested research area than the decades-old work on attitudes: measurement is still maturing and researchers debate how consistently specific "microaggression" categories predict harm. Best treated, then, as a useful lens on subtle bias rather than a settled measurement instrument — the honest scholarly posture.
Two faces of sexism
Prejudice against women is a useful case because it so rarely looks like plain hostility — and Peter Glick and Susan Fiske built a whole theory around that. Their ambivalent sexism framework, tested across roughly 15,000 respondents in 19 nations, holds that sexism comes in two correlated forms (Glick & Fiske, 2001). One is obvious; the other wears a smile.
Hostile sexism
The overtly antagonistic form: women are seen as inferior, manipulative, or bent on controlling men — negative stereotypes with a clear edge. This is the sexism everyone recognizes as sexism.
Benevolent sexism
Subjectively positive — women are pure, nurturing, in need of protection and cherishing. It feels like praise, which is what makes it slippery. But it still casts women as the weaker party who must be looked after, and it rewards them for staying in traditional roles.
Glick and Fiske's key insight is that these are not opposites but partners: both rest on the assumption that women are the weaker sex and belong in a restricted set of roles. Benevolent sexism is the carrot to hostile sexism's stick — it's harder to object to (who complains about being "cherished"?), which is exactly why it can be so effective at maintaining inequality. That's a nice illustration of the attitude/belief distinction from the top of the page: the feeling here is warm even as the underlying beliefs are constraining.
Implicit bias and the IAT
Much modern prejudice, the argument goes, operates below awareness — as automatic associations we didn't choose and may explicitly disavow. The best-known attempt to measure this is the Implicit Association Test (IAT), which infers the strength of associations from how fast you sort concepts (say, faces and "good"/"bad" words) under different pairings (Greenwald, McGhee, & Schwartz, 1998). Faster sorting when a group is paired with negative words than with positive ones is taken as evidence of a negative implicit association. The IAT reliably shows, at the group level, that such associations are widespread — even among people who sincerely hold egalitarian values, which fits the aversive-racism picture nicely.
A real scientific limit
The IAT is genuine, influential science, and it clearly demonstrates that automatic associations exist and are common. But two things a careful reader should know. First, its ability to predict an individual person's discriminatory behavior is weak: a meta-analysis by Oswald and colleagues (2013) found IAT scores were only modestly and inconsistently related to actual behavior. Second, its test–retest reliability is modest — take it twice and your score can wander noticeably — and a large study by Forscher and colleagues (2019) found that changing implicit scores did not reliably change behavior. The responsible reading: the IAT is a strong tool for studying group-level patterns and for consciousness-raising, but a shaky basis for labeling any single individual or predicting what they'll do. This is an honest measurement limitation, not a political verdict — treat it the way you'd treat any effect size with wide confidence intervals. It's part of the broader methodological reckoning collected on the Science Corrects Itself page.
Split-second bias: weapons and shooters
Two lines of reaction-time research show automatic associations bleeding into fast decisions. Keith Payne's (2001) weapon-identification task found that briefly flashing a Black (vs. White) face made participants quicker and more error-prone to misidentify a harmless object (a tool) as a gun — especially under time pressure that blocked deliberate control. Joshua Correll and colleagues (2002) built the shooter task: a video game in which players must "shoot" armed targets and "not shoot" unarmed ones, under a half-second deadline. Participants were faster to correctly shoot armed Black targets and faster to correctly not-shoot unarmed White targets — and made their most costly errors (shooting an unarmed person) when that person was Black. The time pressure is the point: it strips away the controlled processing that would otherwise let people override the automatic association, exposing how a cultural stereotype linking a group with danger can shape a decision before deliberation catches up.
Where prejudice comes from: threat and identity
Two big theoretical families explain why groups turn on each other. The first is intergroup threat theory (Stephan & Stephan, 2000), which distinguishes realistic threat — competition over tangible resources like jobs, land, or safety — from symbolic threat — the sense that another group menaces your values, worldview, or way of life. Notice that symbolic threat needs no material stakes at all; feeling your group's meaning is under attack is enough.
The second, and one of the most important ideas in the field, is Henri Tajfel's social identity theory. Part of our self-esteem, Tajfel argued, is borrowed from the groups we belong to — so we're motivated to see our in-groups as superior, which tilts us toward favoring "us" and derogating "them." His stunning demonstration was the minimal group paradigm: strangers sorted into groups on a basis as trivial as a coin flip or a preference for one abstract painter over another still allocated more rewards to their own group (Tajfel, 1970). No history, no competition, no contact — just a label — and in-group favoritism appears. That's a humbling result: the machinery of "us versus them" runs on almost nothing, which is why it's so hard to switch off.
When bias is built into the machinery
Not all discrimination requires a biased individual in the moment. Institutional discrimination refers to policies and practices within organizations — schools, hospitals, courts, lenders, employers — that systematically disadvantage a group, even when no single decision-maker intends harm. A classic experimental demonstration: science faculty rated an application for a lab-manager position as less competent, less hireable, and worthy of a lower salary when the (otherwise identical) applicant was named "John" rather than "Jennifer" (Moss-Racusin et al., 2012). Multiply one such tilt across thousands of decisions and you get a durable gap that owes nothing to any one person's conscious animus. This is where the attitude/behavior distinction matters most for policy: you can reduce discriminatory outcomes by redesigning the process — anonymizing résumés, structuring evaluations — without first having to change anyone's private attitudes.
Stereotype threat: when the stereotype does the sabotage
One of the most influential — and most debated — findings in this literature is stereotype threat: the worry that your performance might confirm a negative stereotype about your group, a worry that can itself impair the performance. In the founding studies, Claude Steele and Joshua Aronson (1995) gave Black and White students a hard verbal test. When it was framed as a diagnostic measure of ability, Black students underperformed relative to White students of equal preparation; when the same test was framed as a non-diagnostic problem-solving exercise, the gap shrank markedly. The proposed mechanism is that the extra anxiety and self-monitoring consume working-memory resources you'd otherwise spend on the task.
The nuance the headline usually drops
Stereotype threat is real and has been shown in controlled labs many times — that part is solid. But be a careful consumer here: the size and robustness of the effect are genuinely contested. Some registered replications and meta-analyses have found smaller effects than the early studies suggested, and there is evidence that publication bias inflated the apparent magnitude. None of this means the phenomenon is fake; it means the honest statement is "a real effect whose average size and boundary conditions are still being pinned down," not "a large, universal force." Holding both halves of that sentence at once — the effect is real and the estimates are debated — is exactly the kind of scientific maturity this topic rewards.
What actually reduces prejudice: the contact hypothesis
Gordon Allport (1954) proposed that prejudice could be reduced by intergroup contact — but not just any contact. He specified conditions: the groups should meet on equal status, pursue common goals, cooperate rather than compete, and have the contact sanctioned by authority, law, or custom. The elegant real-world application is Elliot Aronson's jigsaw classroom, which makes students from different groups mutually dependent on one another to master material — building Allport's conditions right into the lesson plan.
Does it work? Here the news is genuinely good, and unusually robust for this literature. Thomas Pettigrew and Linda Tropp (2006) meta-analyzed more than 500 studies covering a quarter-million participants and found that intergroup contact reliably reduced prejudice — one of the most consistent intervention findings in all of social psychology. Two honest caveats keep it from being magic. First, Allport's conditions are best understood as facilitators, not strict prerequisites: contact helped on average even when conditions were incomplete, but it helped more when they were met. Second, there are boundary conditions — contact that is coerced, superficial, or anxiety-provoking can fail or even backfire, and effects measured in the lab or among willing participants don't always generalize to tense real-world conflicts. Still, if you want one evidence-based lever for reducing prejudice, well-structured contact is the strongest bet the field has to offer.
Bodhi says
Notice the asymmetry the whole topic ends on: the causes of prejudice are cheap — a coin flip is enough to spark in-group favoritism (Tajfel, 1970) — while the cures are demanding, needing equal status, cooperation, shared goals, and institutional backing to work well. Bias is easy to switch on and hard to switch off. That's not a counsel of despair; it's a reason to take the structural fixes seriously, because good intentions alone rarely clear the bar.
The one thing to carry out of this topic
Keep the three words separate and you'll think clearly about all of it: prejudice is a feeling, a stereotype is a belief, discrimination is an act — and interventions can target any one of the three. Much of modern bias is subtle, automatic, and disavowed by people who mean well, which is precisely why the science matters more than intuition here. Read the strong findings (contact reduces prejudice) and the honest limits (the IAT predicts individuals poorly; stereotype-threat estimates are debated) with the same even hand. That evenhandedness — holding "this effect is real" and "this estimate is uncertain" in mind at once — isn't fence-sitting. It's what taking the topic seriously actually looks like.
References
Allport, G. W. (1954). The nature of prejudice. Addison-Wesley.
Correll, J., Park, B., Judd, C. M., & Wittenbrink, B. (2002). The police officer's dilemma: Using ethnicity to disambiguate potentially threatening individuals. Journal of Personality and Social Psychology, 83(6), 1314–1329. https://doi.org/10.1037/0022-3514.83.6.1314
Dovidio, J. F., & Gaertner, S. L. (2004). Aversive racism. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 36, pp. 1–52). Academic Press. https://doi.org/10.1016/S0065-2601(04)36001-6
Forscher, P. S., Lai, C. K., Axt, J. R., Ebersole, C. R., Herman, M., Devine, P. G., & Nosek, B. A. (2019). A meta-analysis of procedures to change implicit measures. Journal of Personality and Social Psychology, 117(3), 522–559. https://doi.org/10.1037/pspa0000160
Glick, P., & Fiske, S. T. (2001). An ambivalent alliance: Hostile and benevolent sexism as complementary justifications for gender inequality. American Psychologist, 56(2), 109–118. https://doi.org/10.1037/0003-066X.56.2.109
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74(6), 1464–1480. https://doi.org/10.1037/0022-3514.74.6.1464
Moss-Racusin, C. A., Dovidio, J. F., Brescoll, V. L., Graham, M. J., & Handelsman, J. (2012). Science faculty's subtle gender biases favor male students. Proceedings of the National Academy of Sciences, 109(41), 16474–16479. https://doi.org/10.1073/pnas.1211286109
Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E. (2013). Predicting ethnic and racial discrimination: A meta-analysis of IAT criterion studies. Journal of Personality and Social Psychology, 105(2), 171–192. https://doi.org/10.1037/a0032734
Payne, B. K. (2001). Prejudice and perception: The role of automatic and controlled processes in misperceiving a weapon. Journal of Personality and Social Psychology, 81(2), 181–192. https://doi.org/10.1037/0022-3514.81.2.181
Pettigrew, T. F., & Tropp, L. R. (2006). A meta-analytic test of intergroup contact theory. Journal of Personality and Social Psychology, 90(5), 751–783. https://doi.org/10.1037/0022-3514.90.5.751
Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69(5), 797–811. https://doi.org/10.1037/0022-3514.69.5.797
Stephan, W. G., & Stephan, C. W. (2000). An integrated threat theory of prejudice. In S. Oskamp (Ed.), Reducing prejudice and discrimination (pp. 23–45). Erlbaum.
Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223(5), 96–102. https://doi.org/10.1038/scientificamerican1170-96