Abstract
This meta-analysis examines the impact of misleading or incorrect information on subsequent reports of witnessed events, with the aim of identifying key moderators of the misinformation effect – defined as reduced accuracy following exposure to misinformation. We synthesized experimental studies using the three-phase misinformation paradigm published between 1975 and 2025. A three-level mixed-effects meta-regression model was used to account for effect size non-independence and to examine moderators, including retention interval, warnings, overall test performance, and study characteristics. Across 480 studies and 1,998 effect sizes (N = 59,079), the average misinformation effect was substantial (g = 0.694, 95% CI [0.259, 1.128]) and highly heterogeneous (σ2 = 0.697, I2 = 92.21%). Control condition accuracy was positively associated with the magnitude of the effect, and post-misinformation warnings significantly reduced it. In contrast to prior findings, post-event testing did not significantly moderate the effect. These findings support the robustness of the misinformation effect and suggest that it reflects a combination of memory-related and strategic processes. Given the substantial heterogeneity observed, the misinformation effect may be better conceptualized as a constellation of related phenomena rather than a single unified effect, with important implications for theory and applied settings such as eyewitness testimony.