Gordon Pennycook
David RandContributions
- Demonstrates the illusory truth effect using both trivia statements and political news headlines across seven high-powered studies
- Tests three hypothesized moderators (cognitive ability, need for cognitive closure, cognitive style) with multiple measures per construct
- Conducts meta-analytic integration showing no reliable association between any individual-difference measure and the effect size
- Provides evidence that processing fluency (repetition familiarity) and coherent reference formation underlie the effect, independent of deliberative reasoning
- Contributes to understanding why misinformation persists despite individual differences in intelligence or thinking styles
Method
Across seven independent studies, participants read trivia statements (Studies 1, 3–6) or political news headlines (Studies 2, 7) presented as either repeated (already seen), new, or mixed sets. Truth judgments were made on Likert scales. Individual differences were measured via:
- Cognitive ability: Verbal intelligence (Quick Test) or fluid intelligence via cognitive reflection and reasoning tasks
- Need for Cognitive Closure (NFC): Self-report scale measuring preference for certainty and structure vs. ambiguity
- Cognitive Style: Rational-Experiential Inventory (preference for analytic vs. intuitive reasoning) and Preference for Intuition and Deliberation scales
The illusion-of-truth effect was operationalized as the difference in truth judgments between repeated and new statements. Meta-analytic integration (frequentist and Bayesian) examined correlations between individual-difference measures and effect sizes.
Results
All seven studies showed a significant illusory truth effect: participants judged repeated information as significantly more true than new information (effect sizes d = 0.17–1.15). However:
- Cognitive ability: No significant association with the illusory truth effect in any study; meta-analysis (bivariate correlations) showed r = −0.01 (95% CI: [−0.09, 0.07])
- NFC: No significant association; r = −0.05 (95% CI: [−0.18, 0.09])
- Cognitive style (intuition/deliberation measures): Weak and inconsistent associations; meta-analytic r ≈ −0.03 to −0.05, all nonsignificant
- Analytic processing (CRT): No reliable relationship with the effect
The findings held for both trivia and news headlines, and the effect persisted even when participants had conscious knowledge that statements were false.
Notes
Strengths: High-powered multi-study design (2,196 total N), careful operationalization of individual differences with multiple measures per construct, meta-analytic approach to integration, and use of both abstract and policy-relevant materials (news headlines) strengthen confidence in robustness. The preregistration of some analyses enhances credibility.
Limitations: Samples primarily from Western online platforms (Mturk); generalizability to non-Western or offline populations unclear. The null findings for individual differences, while informative, are somewhat surprising given prior theoretical work and raise questions about mechanisms—why doesn't cognitive ability protect against the effect? The paper suggests fluency and reference coherence operate relatively automatically, but explicit process measures (e.g., eye-tracking, response times) might clarify.
Impact: This work is important for misinformation researchers because it suggests the illusory truth effect is not easily mitigated by targeting cognitively sophisticated or analytically-minded audiences. Interventions aimed at reducing false-belief spreading via repeated exposure may need to target the effect's mechanism (fluency attribution) rather than assuming individual differences will naturally protect some people.