Perceiving a meaningful connection between two things when no such relationship actually exists.
Imagine you notice that the two times you wore your red shirt, your team won the game. You didn't notice the five times your team won when you weren't wearing it, or the three times you wore it and they lost. Your brain grabbed onto the memorable match and said 'Aha, these go together!' — even though the shirt has nothing to do with winning.
Illusory correlation occurs through two primary mechanisms: distinctiveness-based and expectancy-based processes. In the distinctiveness-based form, when two statistically infrequent events co-occur (such as a rare behavior performed by a member of a minority group), their shared novelty makes the pairing disproportionately memorable, inflating the perceived frequency of their co-occurrence. In the expectancy-based form, pre-existing beliefs or stereotypes lead people to selectively notice and recall instances that confirm the expected association while disregarding contradictory evidence. The bias is particularly insidious because it operates beneath conscious awareness, causing people to build confident causal narratives from coincidental co-occurrences, and once established, these false associations become self-reinforcing through confirmatory information processing.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Investors perceive false links between unrelated market indicators and stock performance — such as believing that a specific economic report always precedes a rally — because the memorable instances of co-occurrence overshadow the many times the pattern did not hold.
Clinicians overestimate the association between certain patient demographics and diagnoses, or between specific symptoms and rare conditions, because distinctive co-occurrences (e.g., a young patient with a serious disease) are more memorable than routine presentations, distorting diagnostic reasoning.
Teachers may form false associations between student appearance or background characteristics and academic ability, overestimating how often students from a particular group struggle, because the few memorable instances crowd out the many counterexamples.
People form false beliefs about their partner's behavior — such as thinking their partner always forgets important dates — because the emotionally charged instances of forgetting are far more salient than the many times their partner remembered.
Product teams attribute success or failure to incidental design features (like color choices or launch timing) based on a few memorable coincidences, creating internal myths that influence future product decisions without supporting data.
Managers develop false beliefs that employees from certain teams or backgrounds are less reliable, because a few memorable instances of poor performance by those individuals are more salient than the base rate of performance across the entire workforce.
Media coverage of rare crimes committed by members of minority groups makes these events highly memorable, leading the public to drastically overestimate the association between that group and criminal behavior, fueling prejudice and discriminatory policy support.
Loren J. Chapman (1967) coined the term and demonstrated it experimentally in 'Illusory correlation in observational report' (Journal of Verbal Learning and Verbal Behavior). Loren and Jean Chapman (1967, 1969) extended it to clinical psychology. David Hamilton and Robert Gifford (1976) applied it to stereotype formation in social psychology.
In ancestral environments, quickly detecting co-occurrences between events — even at the cost of many false positives — was far less costly than missing a genuine association. Noticing that a certain plant grew near a water source, or that a particular sound preceded a predator's appearance, conferred survival advantages. The asymmetry between the cost of a false alarm (wasted vigilance) versus a missed detection (death) selected for a hair-trigger pattern-detection system.
Machine learning models trained on imbalanced datasets can learn spurious correlations between minority-class features and specific outcomes, mirroring the distinctiveness-based mechanism in human cognition. For example, if a training set contains few examples of a demographic group, and those few examples happen to co-occur with negative labels, the model may encode a false association. Recommender systems and search algorithms can also amplify illusory correlations present in user behavior data, reinforcing stereotypical associations at scale.
50 cards are free to preview. Buyers unlock the rest of the deck plus the interactive training — Spot-the-Bias Quiz unlimited, Swipe Deck with spaced repetition, My Blindspots, Decision Pre-Flight, the Printable Deck + Cheat Sheets, and the Field Guide e-book. $29.50$59.
