Stereotypes are cognitive schemas associating social groups with particular traits; they function as efficient cognitive heuristics but produce discriminatory outcomes when applied inappropriately. Implicit biases are automatic, often unconscious evaluative associations that influence behavior independently of explicit attitudes. The Implicit Association Test (IAT) measures the strength of associations between concept categories and evaluative responses via reaction time. Devine's model distinguishes the automatic activation of stereotypes (universal, due to cultural exposure) from the controlled inhibition of stereotype use (variable, depending on motivation and capacity). Stereotype threat — the fear of confirming negative stereotypes — impairs performance on relevant tasks.
Take an IAT (available online) to experience implicit bias measurement, then read Devine's model to understand the dissociation between automatic activation and controlled application. Stereotype threat experiments are best understood through Steele & Aronson's original paradigm.
You already know from social cognition that the mind uses schemas — cognitive shortcuts that organize information about the world. Stereotypes are a specific kind of schema applied to social groups: mental templates that associate categories ("elderly people," "engineers," "athletes") with clusters of traits. These schemas form because the mind constantly searches for patterns, and group membership is one of the most salient features humans track. The problem is not that we have schemas — it would be cognitively impossible to function without them — but that group schemas are overgeneralized, applied to individuals where they do not fit, and often absorbed from a biased cultural environment rather than personal experience.
Implicit bias extends this picture using the dual-process framework you know. Explicit attitudes are consciously held beliefs — what you say and believe you think about a group. Implicit biases are automatic evaluative associations that operate beneath conscious awareness, in the fast, associative System 1 process. These two can sharply diverge: a person can sincerely endorse egalitarian values while harboring implicit associations that link, for example, certain racial groups with danger or certain genders with incompetence. The Implicit Association Test (IAT) measures this gap by timing how quickly you pair concepts with evaluative categories. Slower pairing times reveal weaker or conflicting associations; faster times reveal stronger ones. The logic is that mentally incompatible pairings create interference, which shows up in milliseconds.
Patricia Devine's influential model explains how someone can hold biased implicit associations and still not act in discriminatory ways. According to Devine, all people who grow up in a culture with racial stereotypes automatically activate those stereotypes when they encounter a group member — this activation is culturally conditioned and essentially universal. What differs across people is the controlled inhibition step: whether someone has the motivation and cognitive resources to catch and suppress the stereotypic response. This maps directly onto your understanding of the fundamental attribution error — we tend to attribute behavior to character rather than situation, but the relevant situation here is internal: depleted cognitive resources, time pressure, or divided attention all reduce controlled suppression, allowing implicit biases to shape behavior even in people who would explicitly disavow them.
Stereotype threat adds another layer: group members themselves are affected. When a negative stereotype about one's group is salient in a testing context, awareness of the risk of confirming that stereotype creates performance anxiety that consumes working memory and impairs the very performance in question. Claude Steele and Joshua Aronson's original experiments showed that Black college students underperformed white peers on a verbal task when it was framed as a test of intellectual ability, but not when the same task was framed neutrally — the only difference being whether stereotype threat was activated. The implication is that measured group differences in performance often reflect situational threat rather than ability, which is a direct challenge to essentialist interpretations of test gaps.
The practical takeaway from Devine's model is important and often missed: having implicit biases does not define character. What matters is whether people treat those biases as something to monitor and override. Prejudice reduction interventions work best when they increase motivation to be egalitarian and when they provide concrete implementation strategies — specific if-then plans ("If I catch myself making an assumption about a person based on their group, I will pause and check the evidence") — rather than simply informing people of their biases. Information alone rarely changes behavior; structured practice at interception does.