How Good Is Human Lie Detection
Most people believe they can spot a liar by the eyes or the fidgeting. The pooled evidence puts human accuracy at 54 per cent, the cues people trust are largely absent, and the variation sits more with the person speaking than the person judging.
Answer to the question
Can people tell when someone is lying?
Barely. A 2006 meta-analysis of 206 documents found that people sort lies from truths correctly 54 per cent of the time, against 50 for guessing. They catch 47 per cent of lies and recognise 61 per cent of truths, and the cues most people trust, such as gaze, do not reliably separate the two.
Key takeaways
- Bond and DePaulo (2006) pooled 206 documents and 24,483 judges: average accuracy 54 per cent, with 47 per cent of lies and 61 per cent of truths judged correctly.
- DePaulo and colleagues (2003) combined 1,338 estimates of 158 cues and found that many behaviours had little or no measurable link to lying.
- A 2006 survey across 75 countries and 43 languages found a dominant stereotype that liars avert their gaze; Hartwig and Bond (2011) call gaze aversion not a valid indicator.
- Hartwig and Bond (2011) found people rarely rely on the wrong cues (correlations of .59 and .72) and concluded that the signal itself is weak.
- Bond and DePaulo (2008) analysed 247 samples and found differences in judges' ability to be minute, with the best judges no more accurate than chance would produce.
A survey across 75 countries and 43 languages found one belief about liars that crossed cultures: they avert their gaze. The Global Deception Research Team reported it as the dominant stereotype, and it is the working theory behind most everyday lie detection. The research on lie detection says the theory does not hold up, and that the people who rely on it do about as well as everyone else. The standard reference for that research is a pooled analysis of laboratory studies. Its headline figure sits much nearer a coin toss than a skill, and the split inside it explains why.
What sits inside the 54 per cent figure
The best estimate of lie detection accuracy, meaning how well people sort lies from truths, is a meta-analysis by Charles Bond and Bella DePaulo, drawn from studies in which people tried to separate lies from truths in real time, with no special aids or training.
Published in 2006, it pooled 206 documents and 24,483 judges. Average accuracy was 54 per cent.
Guessing scores 50.
| Message being judged | Judged correctly |
|---|---|
| Lies | 47 per cent |
| Truths | 61 per cent |
| All messages | 54 per cent |
Judges caught fewer lies than a coin would and recognised more truths than it would. The authors reported that people tend to see the person they are talking to as honest. That tendency, usually called truth-bias, accounts for the shape of the table, because a judge who leans towards believing will recognise honest people and miss liars.
Of every ten lies, the judges missed more than five.
Of every ten truths, they accepted six.
Four points above chance is a measurable result and a thin one. The authors called it nontrivial relative to differences between judges, a mean d of roughly .40, and it is still far from the confidence people place in their own judgement. Judges also did better with audible lies than with visible ones, an odd finding for anyone who believes the face gives liars away. The same analysis reported that people who were motivated to be believed came across as more deceptive.
What gaze, fidgeting and blushing really tell us
The Global Deception Research Team ran two studies, an open-ended question about how to tell when someone is lying and a questionnaire, and offered a social-control reading of the belief.
The belief about gaze has also had a direct test. In 2003 DePaulo and five colleagues combined 1,338 estimates of 158 possible cues to lying and found that many behaviours had little or no measurable connection to lying. Liars were less forthcoming and told less convincing stories. Their accounts held fewer ordinary imperfections and fewer unusual details, and they made a more negative impression and looked more tense. The 2003 analysis is not a finding that liars behave exactly like truth-tellers. It is a finding that the differences are small and mostly not the ones people look for.
The differences grew when liars were highly motivated, especially when the motive was tied to identity rather than money, and when the lie concerned a transgression.
Maria Hartwig and Charles Bond summarised that body of work in 2011 in blunt terms. Gaze aversion, they wrote, is not a valid indicator of lying, and many signs of nervousness, such as fidgeting, blushing or speech disturbances, are not systematically linked to it. A related question, what a face can reveal about feeling, is taken up in [What a Camera Can Read on a Face](https://thewhyweproject.com/articles/what-a-camera-can-read-on-a-face).
The best way to improve judgment achievement is to increase behavioral differences between liars and truth tellers.
The same 2011 paper holds the less expected result. Using four meta-analyses built on a lens model, Hartwig and Bond compared the cues people rely on with the cues that actually occur. The correlation was .59 across studies and .72 within studies, the latter drawn from 25 sender samples. People rarely rely on the wrong cues, and their intuitive judgements were more accurate than the beliefs they would state if asked. The weakness is in the signal, because behaviour only faintly separates liars from truth-tellers.
Why no judge stands out
If lie detection were a skill, some people would have it, and the best judges would stand out in the data. Bond and DePaulo tested that in 2008, with psychometric analyses of 247 samples in which people judged strangers' veracity in real time. Who laboratory studies sample is a question in its own right, taken up in [Who the Studies Were Run On](https://thewhyweproject.com/articles/who-the-studies-were-run-on).
After correcting for measurement error, the ability differences were minute: the standard deviation in the percentage of lies detected was under 1 per cent, and the best judges were no more accurate than a chance process would produce. The case for a gifted minority was not borne out in that pool. Judges differed more in how willing they were to regard statements as truthful than in how well they could tell lies from truths.
Hartwig and Bond reported in 2011 that presumed lie experts, people who routinely assess credibility in their professional lives, do not perform better than lay judges do.
Aja Raden's The Truth About Lies, the seed for this article, sorts lies into nine named cons, among them the Big Lie, Three-Card Monte and the Bait and Switch. Each is told from the con artist's side. The meta-analyses describe the other side of the table, where the average judge is the mark and the mark's default is to believe. Raden's book is built around the person telling the lie, and in this literature too the variation sits more with the sender than with the judge. [Why Experts Fall for Scams](https://thewhyweproject.com/articles/why-experts-fall-for-scams) draws on the same book to ask who is taken in.
Change the interview, because observers are not the problem
If people already weigh the available cues sensibly and the cues are weak, teaching observers to watch harder gives lie detection little to work with. Hartwig and Bond reached that conclusion themselves: "the best way to improve judgment achievement is to increase behavioral differences between liars and truth tellers," rather than to train perceivers in the cues that are valid.
The behaviours that do separate the two groups are of a different kind from the ones people watch. Liars give less forthcoming, less detailed and less convincing accounts. Those are properties of what is said, which an interviewer can probe with further questions, and which a glance across a table cannot reach.
Because liars were less forthcoming and gave fewer unusual details, an interview that asks for more detail gives that difference room to show, which is the direction Hartwig and Bond's conclusion points, and a different use of the conversation from the face-watching that passes for lie detection.
Common questions
Can people tell when someone is lying?
Barely. A 2006 meta-analysis of 206 documents found average accuracy of 54 per cent against 50 for guessing, with 47 per cent of lies and 61 per cent of truths judged correctly. A 2008 analysis of 247 samples found minute differences in judges' ability, so no group of people is reliably good at lie detection.
Is eye contact a sign that someone is lying?
Not reliably. Hartwig and Bond concluded in 2011 that gaze aversion is not a valid indicator of lying, although a 2006 survey across 75 countries and 43 languages found it to be the dominant stereotype of how liars behave. Belief in the sign is widespread and lie detection built on it fares no better than chance.
What do liars actually do differently?
Across 158 cues, DePaulo and colleagues found in 2003 that liars were less forthcoming, told less convincing stories with fewer ordinary imperfections and unusual details, made a more negative impression and seemed more tense. The differences were small, and larger when liars were highly motivated.
Why do people believe liars so often?
Bond and DePaulo found in 2006 that judges tend to see the people they talk to as honest. Judges recognised 61 per cent of truths and 47 per cent of lies, so a lean towards belief helps with honest speakers and costs accuracy with liars. That is the lie detection problem in miniature: the judge is right when the speaker is honest and wrong when the speaker is not.
Sources
Raden, A. (2021). The Truth About Lies. Atlantic Books (UK); St. Martin's Press (US). Used for its structure of nine named cons; no detection finding is taken from it.
Bond, C. F., Jr., & DePaulo, B. M. (2006). 'Accuracy of deception judgments.' Personality and Social Psychology Review, 10(3), 214-234. doi:10.1207/s15327957pspr1003_2
Bond, C. F., Jr., & DePaulo, B. M. (2008). 'Individual differences in judging deception: Accuracy and bias.' Psychological Bulletin, 134(4), 477-492. doi:10.1037/0033-2909.134.4.477
DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). 'Cues to deception.' Psychological Bulletin, 129(1), 74-118. doi:10.1037/0033-2909.129.1.74
Hartwig, M., & Bond, C. F., Jr. (2011). 'Why do lie-catchers fail? A lens model meta-analysis of human lie judgments.' Psychological Bulletin, 137(4), 643-659. doi:10.1037/a0023589
The Global Deception Research Team (2006). 'A world of lies.' Journal of Cross-Cultural Psychology, 37(1), 60-74. doi:10.1177/0022022105282295