Evidence locker · triangulate the cluster

Illusory Correlation

Illusory correlation is the tendency to perceive a relationship between two variables when no such relationship exists, or to overestimate the strength of a weak relationship.

Evidential strength ↑ (n / replications)

robustnoise

Hamilton & Gifford stereotype formation

— · n≈40

Chapman & Chapman Rorschach signs

— · n?

Redelmeier & Tversky arthritis and weather

— · n≈18

★ Invalid Rorschach signs for homosexua…

field incident · 1969

★ Arthritis pain and weather belief

field incident · 1996

★ Congressional claims linking immigrat…

field incident · 2006

Robust

The claim

The term was coined by Loren Chapman in 1967 to describe how observers report correlations between word pairs that do not actually exist, and was extended to social perception by Hamilton and Gifford (1976), who showed that perceiving an illusory correlation between minority groups and undesirable behaviors is sufficient to produce stereotypic judgments.

Effect size → (null · small · moderate · large)

Reading the scatter. ★ chits are real-world incidents (vivid, but n = 1, so they sit low on strength). The strongest, highest-powered evidence sits in the upper-right robust quadrant; the effect holds where the data is best.

Exhibit index

— · n≈40

Hamilton & Gifford stereotype formation. Participants attributed more desirable behaviors to Group A (M=6.7) than Group B (M=6.0) and more undesirable behaviors to Group B (M=5.6) than Group A (M=4.4), demonstrating that distinctiveness of minority-negative behavior pairings produces stereotypic judgments.

Hamilton & Gifford, 1976 · n=40 undergraduates (Experiment 1)

— · n?

Chapman & Chapman Rorschach signs. Participants recaptured the same illusory associations perceived by expert clinicians—for example, that responses referring to the anus, feminine clothing, or genitals were indicators of homosexuality—despite zero actual correlation in the stimulus materials.

Loren J. Chapman & Jean P. Chapman, 1969 · Undergraduate participants

— · n≈18

Redelmeier & Tversky arthritis and weather. No correlation was found between pain severity and any weather variable (temperature, barometric pressure, humidity). The persistence of the belief was attributed to illusory correlation: patients noticed and remembered coincidences of pain and bad weather while ignoring non-coincident instances.

Donald A. Redelmeier & Amos Tversky, 1996 · n=18 rheumatoid arthritis patients

Why the shortcut forms

The dominant account is the distinctiveness-based explanation: when two distinctive (infrequent) events co-occur, their joint distinctiveness makes them especially salient, drawing disproportionate attention and leading to overestimation of their co-occurrence frequency. In Hamilton and Gifford's paradigm, minority group membership and negative behaviors are both infrequent, making their combination maximally distinctive. This increased attention results in better memory for distinctive co-occurrences and inflated frequency estimates.

An alternative account proposes that illusory correlations arise from ad hoc inferences from base rates rather than from memory advantages for distinctive combinations. More recently, the Accentuation Theory (Sherman et al., 2009) has been proposed as an integrative framework.

Replication

Robust

The effect has been widely replicated across diverse paradigms and domains. Mullen and Johnson (1990) conducted a meta-analysis of 19 studies finding a reliable overall effect, with larger effects for negative distinctive stimuli. The effect has been replicated in children, in ERP paradigms, and with real social groups.

The effect sizes from the original Hamilton & Gifford (1976) studies were reported as F-statistics rather than standardized effect sizes, making comparison difficult. Some boundary conditions (e.g., when the SDA account vs. accentuation theory applies) remain contested.

What to log instead

Seek systematic data rather than relying on salient anecdotes. Explicitly count co-occurrences and non-occurrences rather than relying on memory, which overweights distinctive events.

Training designed to reduce illusory correlation by encouraging consideration of all four cells of a contingency table has shown some success, though the effect is resistant to simple instruction (Kurtz & Garfield, 1978).

Catch it in the act

Watch for when you associate a rare event with another rare event—such as linking a minority group with a negative behavior simply because both stand out. Be suspicious of 'patterns' that emerge from a few striking examples rather than systematic data.

Adjacent lockers

Open the same case on your own draft.