Accuracy of deception judgments.
Level 1 - systematic review of randomized trials
Meta-analysis of behavioral experimental studies (by design analogy, non-clinical CEBM).
PubMed 16859438 · doi:10.1207/s15327957pspr1003_2
What was done
The authors conducted a meta-analysis synthesizing findings from 206 documents involving 24,483 judges evaluating real-time deception without specialized aids or training. The analysis evaluated overall discrimination accuracy, modality differences (audible vs. visible), target motivation effects, and interpersonal perception biases.
What was found
Judges achieved an average of 54% correct lie-truth judgments, correctly classifying 47% of lies as deceptive and 61% of truths as nondeceptive. The mean accuracy effect size was d ≈ 0.40, corresponding to the ~60th percentile of meta-analyzed social psychology effect sizes. Detection rates showed little variance across studies. People were more accurate at judging audible lies than visible lies. Targets motivated to be believed appeared more deceptive, and judges generally perceived interaction partners as honest.
Why it matters
This benchmark meta-analysis establishes that unaided human deception detection is barely above chance, constrained by a systematic truth bias where genuine truths are recognized more reliably than actual lies.
Limits
The findings are limited to real-time, unaided judgments and cannot be generalized to specialized forensic tools, extended investigative contexts, or trained deception examiners. Abstract does not report confidence intervals, detailed stimulus characteristics, or demographic breakdowns.
Cited by
- contradicts Women are better at detecting lies and are better liars than men.