What do Laboratory False Confession Paradigms Measure? A Calibration Meta-Analysis

Abstract In the most widely used laboratory false confession paradigm, 37% of participants “confess,” typically under procedures involving little more than an accusation. As a calibration benchmark, that baseline rate would correspond to more than a quarter-million false confessions annually in the United States under conservative assumptions, far above the cumulative historical count of known false confession exonerations. We reanalyze 99 experimental conditions drawn from 41 studies using hierarchical Bayesian estimation, random effects meta-analysis, publication-bias diagnostics, calibration simulations, dose–response analysis, and statistical power calculations. The pooled false confession rate is 41%, with a 95% prediction interval spanning 3% to 93% (I 2 = 93%), a level of instability difficult to reconcile with measurement of a stable psychological construct. Calibration simulations show that the observed laboratory means are statistically compatible only with real-world prevalence rates well above the range supported by available estimates. Interrogation tactics account for less than 1% of the variation in confession rates; design features other than the coercive manipulation account for most of the observed variation, with the no-tactic baseline alone explaining 60% of the false evidence ploy rate. Power calculations help explain why high-baseline paradigms are experimentally useful: they make relative tactic comparisons feasible at ordinary laboratory sample sizes, but that same feature limits the field relevance of the absolute magnitudes. These findings raise construct and ecological validity concerns: dominant laboratory paradigms appear to capture low-stakes compliance more readily than interrogation-induced false confession. Substantial caution is therefore warranted when extrapolating experimental results to legal, policy, or forensic contexts.