Can you tell when someone is lying? What research shows about the signs

Lists of “signs someone is lying” are everywhere: liars avoid your eyes, touch their nose, glance up and to the right. When people in 58 countries were asked how they can tell that someone is lying, the most common answer was that liars avoid eye contact (Global Deception Research Team, 2006).

The short answer is that, on average, lies and truths do differ a little. Lies tend to contain fewer details and sound less plausible, and the person telling them tends to look a little more tense. But in a meta-analysis that examined 158 cues, most effects were very small: the median effect size of the 88 cues with at least three estimates was just |0.10| (DePaulo et al., 2003). In experiments where people judged strangers’ statements on the spot, they were right 54% of the time on average, where guessing would give 50% (Bond & DePaulo, 2006).

One analysis concluded that this is less because people watch the wrong things than because lies leave only faint traces in behaviour (Hartwig & Bond, 2011). In everyday life, lies were more often uncovered later, through other people or physical evidence, than by body language at the moment they were told (from the abstract; Park et al., 2002).

This guide is not a checklist for catching or confronting anyone, and mind.io has no lie test. It covers the belief about eye contact and what studies found, the cues that do differ, how well people tell lies from truths, micro-expressions and polygraphs, the approaches researchers consider more promising, everyday lies and suspicion in close relationships, and common questions.

In 2006 an international team ran two surveys. In the first, 40 people in each of 58 countries, 2,320 in all, wrote down how they can tell when people are lying. 63.66% mentioned gaze aversion, far more than any other answer, and it was the most common answer in 51 of the 58 countries. In the second, 40 people in each of 63 countries filled in a questionnaire, and 71.5% said liars make less eye contact than usual (Global Deception Research Team, 2006).

The belief appeared in all 75 countries studied. Japan took part in both surveys and Korea in the second, but the paper does not report figures for individual countries, and most participants were students (mostly university students in the first survey) (Global Deception Research Team, 2006).

The research tells a different story. In a meta-analysis of experiments, 32 estimates of eye contact combined to an effect of almost exactly zero (d = 0.01), and gaze aversion showed no clear difference either (d = 0.03) (DePaulo et al., 2003). Another meta-analysis found “no evidence that people avoid eye contact while lying” (from the abstract; Sporer & Schwandt, 2007). In experiments where people were motivated to succeed, liars did make slightly less eye contact (d = −0.15); where there was no special incentive, they made slightly more, though not significantly (DePaulo et al., 2003). Either way, the difference is far too small to judge one person by.

People say they watch the eyes, yet when researchers looked at the judgements people actually made, eye contact was only weakly related to them (r = −0.15) (Hartwig & Bond, 2011). People avoid eye contact for many reasons other than lying: social anxiety can make eye contact hard even when you want it, as the guide on social anxiety explains, and some autistic people find eye contact uncomfortable and make it anyway to fit in, as the guide on autism in adults describes.

The claim that eyes moving up and to the right reveal a lie comes from neuro-linguistic programming (NLP). Researchers in the UK and Canada tested it in three studies. The eye movements of 32 right-handed people did not differ significantly between lying and telling the truth. Of 50 people who judged 32 clips, the 21 who had been taught the rule got 16.33 right on average, not significantly different from the 29 who had not (16.59). In 52 videos of real public appeals for missing relatives (26 judged lying and 26 truthful on strong evidence), there was no significant difference either. The authors concluded that there are “considerable grounds to be skeptical” of the claim (Wiseman et al., 2012).

Blinking showed no clear link with lying either (d = 0.07). Larger pupils (d = 0.39) rested on only four estimates (DePaulo et al., 2003), and one analysis warns that effects based on few estimates may be inflated (from the abstract; Luke, 2019). As for why the belief is so widespread, the international team suggested that the image of a liar too ashamed to meet your eyes is passed on as a social norm to discourage lying, while saying that their idea may be speculative (Global Deception Research Team, 2006).

A 2003 meta-analysis combined 1,338 estimates of 158 cues from 120 samples. Most of the people telling lies were US students speaking to strangers, and where length was reported, 28 of 36 samples had messages of a minute or less (DePaulo et al., 2003). In the figures below, a positive d means liars showed more of the behaviour and a negative d means they showed less.

The cues that did differ on average were mostly about what was said and how. Lies contained fewer details (d = −0.30, 24 estimates), sounded less plausible (−0.23) and less logically structured (−0.25), and were a little more often inconsistent or ambivalent (0.34). Liars sounded more uncertain (0.30) and less direct and personal (−0.55, 7 estimates), looked a little more tense (0.27) and spoke in a slightly higher pitch (0.21). Truth-tellers more often corrected themselves mid-story (−0.29) or admitted that they could not remember something (−0.42), though each of these rested on only five estimates (DePaulo et al., 2003).

Many behaviours showed little or no difference: the length of answers (−0.03, 49 estimates), the time taken to answer (0.02, 32 estimates), speech errors and pauses, shifts in posture (0.05) and hand movements (0.00). On fidgeting, the authors wrote that “the best summary of these data is that there is no clear relationship between fidgeting and lying” (DePaulo et al., 2003). Another meta-analysis found that of 11 behaviours visible in the head and body, only three were reliably linked to lying (nodding, hand movements, and foot and leg movements), and all three decreased rather than increased; most effects varied a lot between studies, and the authors cautioned practitioners against using these behaviours to judge whether an account is true (Sporer & Schwandt, 2007).

Overall, the median effect size of the 88 cues with at least three estimates was just |0.10|, and only three exceeded |0.40|. The authors summed this up under the heading “The Looks and Sounds of Deceit Are Faint”. Cues were clearer when people were highly motivated, especially when their competence or how they would be judged was at stake, and when lies concerned wrongdoing; lies told with no special motivation and not about wrongdoing were “nearly cueless” (DePaulo et al., 2003). One analysis argues that even these effects may be inflated by publication bias and small numbers of estimates (Luke, 2019).

These are average differences across many people. A short answer or a shaky voice does not make one person a liar: people who were motivated to be believed looked deceptive whether or not they were lying (Bond & DePaulo, 2006). Interestingly, the cues people actually use when judging mostly point in the same direction as the real ones, and no behaviour was read the wrong way round; the researchers put the main limit down to how weak the cues themselves are (Hartwig & Bond, 2011).

One meta-analysis combined 206 documents and 24,483 judges. In most of the studies, people watched or heard strangers in a lab and judged on the spot, with no aids or training, whether each statement was true. They were right 54% of the time on average (an unweighted 53.98% across 292 samples), correctly judging 61% of truths but only 47% of lies. Judgements of “true” made up 56.86% on a simple average (55.23% weighted), more than half; this lean towards believing others is known as the truth bias (Bond & DePaulo, 2006).

Sample averages ranged from 31% to 73%, but the extreme values mostly came from studies with small numbers of judgements. The authors noted that 54% is only slightly better than a coin flip, yet measured against how much judges differ from one another, it is not a trivial effect (d ≈ 0.40) (Bond & DePaulo, 2006).

What people watch also matters. Judging from silent video was less accurate than judging with sound, or even from a written transcript, and judging from the face alone was at chance (d = 0.01). The authors took this to mean that the usual picture of a liar is largely visual, so seeing the speaker brings the stereotype to mind (Bond & DePaulo, 2006).

What about police officers, judges, mental health professionals or customs officers, whose work brings them into contact with lies? Across 20 comparisons in which experts and non-experts judged the same statements, there was no evidence that experts did better (d = −0.025). Experts averaged under 55%. Across studies they looked more sceptical, judging fewer statements to be true (52.28% versus 55.69%), but within studies that difference was not significant (Bond & DePaulo, 2006). A smaller meta-analysis in a magazine for forensic examiners likewise found professionals such as police officers, judges and psychologists (55.51%, 2,685 people) no more accurate than students and other citizens (54.22%, 11,647 people). Confidence was almost unrelated to accuracy (r = 0.05), and age, experience, education and sex were not significantly related to it. We use it only as supporting evidence (Aamodt & Custer, 2006).

Are there natural “human lie detectors”? An analysis of 247 samples concluded that differences in the ability to detect lies are “minute”: in the percentage of lies detected, the standard deviation between judges, corrected for measurement error, was under 1%, and people differed more in how readily they believed others. The bigger differences lay with the speakers. Some people seemed more credible whether lying or telling the truth, and of the individual differences studied, the speaker’s credibility mattered most to the outcome of a judgement (from the abstract; Bond & DePaulo, 2008).

The authors concluded that despite decades of research, detection rates “rarely budge”, and that for people who must judge in real time with no special aids, “many lies are undetectable”. They also noted the limits: most studies were lab tasks, the judges were usually college students, and they had no chance to gather more information (Bond & DePaulo, 2006). The limits of judging people by their appearance are also covered in the guide on face reading and first impressions.

Micro-expressions are fleeting expressions, lasting 1/25 to 1/5 of a second, that may cross the face when someone hides an emotion. In one experiment, participants looked at disgusting, sad, frightening, happy and neutral pictures and responded with genuine or faked expressions, and researchers analysed 697 expressions frame by frame at 1/30 of a second. Every participant leaked an inconsistent expression at least once, but micro-expressions appeared in only 21.95% of participants and in 2% of all expressions, and untrained observers did only slightly better than chance (from the abstract; Porter & ten Brinke, 2008). The task involved faking emotions about pictures, which is not the same as lying about facts.

Does training in reading micro-expressions help? In a study that randomly assigned 90 people to the Micro-Expressions Training Tool (METT), a bogus training or no training, the METT group did no better than the other two, and overall accuracy was slightly below chance (from the abstract; Jordan et al., 2019). It is a single study, with five videos per person, but it gives no support for a training effect.

A polygraph does not detect lies as such. It measures physiological changes such as heart rate, and examiners infer the person’s psychological state from the charts. A committee of the US National Research Council pointed out that states linked to lying, such as fear of being judged deceptive, can also arise without any lie, and that anxiety about the test itself affects the same responses. It also noted evidence that truthful people who are believed to be guilty may respond like deceptive ones (National Research Council, 2003).

The committee concluded that tests about a specific incident can tell lying from truth-telling “at rates well above chance, though well below perfection”. Across 52 laboratory datasets the accuracy index had a median of 0.86 (the middle half between 0.81 and 0.91), and seven field studies had a median of 0.89; but this index is not the percentage of correct answers, and the committee judged these figures to overstate accuracy in the field. It also gave a screening example. Among 10,000 employees including 10 spies, even assuming an accuracy of 0.90, higher than can actually be expected, a test set to catch 8 of the spies would also flag 1,598 innocent employees, and one set to cut false alarms to 39 would let 8 of the 10 spies pass. Because the responses being measured can be altered on purpose, countermeasures were another concern (National Research Council, 2003).

How are lies actually discovered? When 202 people recalled a time they had found out that someone lied to them, they had mostly relied on information from third parties and physical evidence, and finding out often took days, weeks, months or longer (Park et al., 2002).

So researchers have been developing interview methods that widen the differences between lies and truths through questioning rather than by reading body language. The cognitive approach imposes extra mental load, encourages people to say more and asks unexpected questions. Its developers reported, in a meta-analysis of studies using the cognitive approach (14 of which compared it directly with a standard interview), a total accuracy of 71%, against 56% for a standard interview (from the abstract; Vrij et al., 2017). Other researchers reanalysed the data and argued that accuracy in the control conditions was unusually low and that the advantage was “much smaller than previously claimed” (from the abstract; Levine et al., 2018). The original team rejected the criticisms but noted that both sides reached the same conclusion, “a modest improvement” (from the abstract; Vrij et al., 2018).

A meta-analysis limited to human observers found an average accuracy of 60.00% for judgements of interviews that used the cognitive approach (55.03% after correcting for bias). Observers who did not know what to look for scored 52.37%, little better than chance, while those told which cues to attend to scored 75.81%. The authors added that there were signs of publication bias and large differences between studies (from the abstract; Mac Giolla & Luke, 2021).

Two other methods are also being studied. The Verifiability Approach uses the fact that liars tend to avoid details that could be checked, such as named witnesses, CCTV or receipts. In a meta-analysis of 14 studies, lies contained fewer checkable details (g = 0.42), and the gap grew when people were asked beforehand to include details the investigator could check (g = 0.80) (Palena et al., 2021, as cited in Vrij et al., 2022). The Strategic Use of Evidence (SUE) technique holds back the evidence the interviewer has and checks, through questions, whether the account fits it; across eight studies, accounts differed strongly in how well they matched the evidence (d = 1.06) (Hartwig et al., 2014, as cited in Vrij et al., 2022).

These methods are for trained investigative interviewers, and most of the research consists of experiments in which participants are asked to lie or tell the truth. Their developers also list as limitations that the research is concentrated in a few labs and comes mostly from Western countries (Vrij et al., 2022). Still, the shared message is clear: information that can be checked, and whether a story fits the facts, tell you more than body language. One review of the field describes the nonverbal cues found so far as “faint and unreliable” (from the abstract; Vrij et al., 2019).

How often do people lie? In a diary study in which 77 students at the University of Virginia and 70 people from the local community recorded for a week every social interaction lasting 10 minutes or more (and any shorter one in which they lied) and every lie, students reported about two lies a day (1.96) and community members about one (0.97). Students lied in about one in three of their social interactions and community members in about one in five, and participants said they did not regard most of their lies as serious or plan them much (DePaulo et al., 1996).

In an online survey of 1,000 US adults about lies told in the past 24 hours, the average was 1.65, but the median was zero: 59.9% said they had told no lies that day, and half of all reported lies came from 5.3% of people. The researchers warned that because the distribution is so skewed, the average is misleading (Serota et al., 2010). Both studies rely on self-reports.

Are lies easier to spot in people we know? In experiments where judges first saw a sample of the person’s ordinary behaviour, accuracy rose a little (55.91% versus 52.26% across 21 comparisons), but the difference was small. People tended to give those familiar to them the benefit of the doubt, and judgements made by long-term acquaintances were also within a few points of 50% (Bond & DePaulo, 2006).

So what follows is not a tested method but suggestions in light of the research. If something bothers you, talk about the specific fact that does not add up rather than trying to read a face or a posture. Labelling someone a liar on body language alone can wrong a person who is telling the truth. If you feel the urge to check a partner’s phone in secret, the guide on jealousy in relationships covers why checking brings brief relief but can breed new suspicion. If someone repeatedly denies what you clearly saw or heard, until you start doubting your own memory and judgement, see the guide on gaslighting.

It is worth asking why lists of “liar tells” feel convincing. For the belief that eye movements reveal lies, the researchers raised one possibility, an illusory correlation: people remember the times the sign happened to match a lie and forget the times it did not. It is a hypothesis they did not test (Wiseman et al., 2012). Why descriptions that fit almost anyone feel personal is covered in the guide on the Barnum effect.

Q. Do people look up and to the right when they lie? A. Three studies found no support for it: eye movements did not differ significantly between lying and telling the truth, and people taught the rule were no more accurate (16.33 versus 16.59 out of 32 clips) (Wiseman et al., 2012).

Q. Do liars touch their nose or fidget more? A. There was no clear relationship between fidgeting and lying, and touching or scratching oneself barely differed (d = −0.01) (DePaulo et al., 2003). One meta-analysis even found that hand and leg movements slightly decreased when people lied, though results varied between studies (Sporer & Schwandt, 2007).

Q. Are police officers or psychologists better at spotting lies? A. There is no evidence that they are. Across 20 comparisons of judgements of the same statements, experts did no better, and they averaged under 55% (Bond & DePaulo, 2006). People who deal with lies at work (55.51%) and students and other citizens (54.22%) differed little (Aamodt & Custer, 2006).

Q. How accurate is a polygraph? A. For questions about a specific incident it does much better than chance, but far from perfectly. The median accuracy index of 0.86 in laboratory studies does not mean “86% correct”, and the committee expected accuracy in the field to be lower. When screening a group in which very few people are lying, many innocent people get flagged (National Research Council, 2003).

Q. Can learning about micro-expressions help me catch lies? A. The evidence so far suggests it is hard. Micro-expressions appeared in only 2% of expressions (Porter & ten Brinke, 2008), and in a study that randomly assigned 90 people, those who were trained were not more accurate (Jordan et al., 2019).

Q. Is it easier to tell when someone close to me is lying? A. Only slightly. In experiments, judges who had first seen a sample of a person’s normal behaviour were right 55.91% of the time, against 52.26% without, and people tend to give those they know the benefit of the doubt (Bond & DePaulo, 2006). In real life, lies were often found out later, through other people or evidence (Park et al., 2002).

Q. How many lies do people tell a day? A. In a one-week diary study, students told about two lies a day (1.96) and community members about one (0.97) (DePaulo et al., 1996). In a survey of 1,000 US adults the average was 1.65, but 59.9% told no lies that day and 5.3% told half of all the lies (Serota et al., 2010).

Q. Are some people better at getting away with lies? A. Speakers differed more than judges did: some people seemed more credible whether they were lying or telling the truth (Bond & DePaulo, 2008). Seeming believable is not the same as being honest.

References: The Global Deception Research Team (2006), A world of lies, Journal of Cross-Cultural Psychology 37(1), 60–74, https://doi.org/10.1177/0022022105282295; 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, https://doi.org/10.1037/0033-2909.129.1.74; Bond, C. F., Jr., & DePaulo, B. M. (2006), Accuracy of deception judgments, Personality and Social Psychology Review 10(3), 214–234, https://doi.org/10.1207/s15327957pspr1003_2; 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, https://doi.org/10.1037/a0023589; Park, H. S., Levine, T. R., McCornack, S. A., Morrison, K., & Ferrara, M. (2002), How people really detect lies, Communication Monographs 69(2), 144–157, https://doi.org/10.1080/714041710; Sporer, S. L., & Schwandt, B. (2007), Moderators of nonverbal indicators of deception: a meta-analytic synthesis, Psychology, Public Policy, and Law 13(1), 1–34, https://doi.org/10.1037/1076-8971.13.1.1; Luke, T. J. (2019), Lessons from Pinocchio: cues to deception may be highly exaggerated, Perspectives on Psychological Science 14(4), 646–671, https://doi.org/10.1177/1745691619838258; Wiseman, R., Watt, C., ten Brinke, L., Porter, S., Couper, S.-L., & Rankin, C. (2012), The eyes don’t have it: lie detection and neuro-linguistic programming, PLoS ONE 7(7), e40259, https://doi.org/10.1371/journal.pone.0040259; Aamodt, M. G., & Custer, H. (2006), Who can best catch a liar? A meta-analysis of individual differences in detecting deception, The Forensic Examiner 15(1), 6–11; Bond, C. F., Jr., & DePaulo, B. M. (2008), Individual differences in judging deception: accuracy and bias, Psychological Bulletin 134(4), 477–492, https://doi.org/10.1037/0033-2909.134.4.477; Porter, S., & ten Brinke, L. (2008), Reading between the lies: identifying concealed and falsified emotions in universal facial expressions, Psychological Science 19(5), 508–514, https://doi.org/10.1111/j.1467-9280.2008.02116.x; Jordan, S., Brimbal, L., Wallace, D. B., Kassin, S. M., Hartwig, M., & Street, C. N. H. (2019), A test of the micro-expressions training tool: does it improve lie detection?, Journal of Investigative Psychology and Offender Profiling 16(3), 222–235, https://doi.org/10.1002/jip.1532; National Research Council (2003), The Polygraph and Lie Detection, Washington, DC: The National Academies Press, https://doi.org/10.17226/10420; Vrij, A., Fisher, R. P., & Blank, H. (2017), A cognitive approach to lie detection: a meta-analysis, Legal and Criminological Psychology 22(1), 1–21, https://doi.org/10.1111/lcrp.12088; Levine, T. R., Blair, J. P., & Carpenter, C. J. (2018), A critical look at meta-analytic evidence for the cognitive approach to lie detection: a re-examination of Vrij, Fisher, and Blank (2017), Legal and Criminological Psychology 23(1), 7–19, https://doi.org/10.1111/lcrp.12115; Vrij, A., Blank, H., & Fisher, R. P. (2018), A re-analysis that supports our main results: a reply to Levine et al., Legal and Criminological Psychology 23(1), 20–23, https://doi.org/10.1111/lcrp.12121; Mac Giolla, E., & Luke, T. J. (2021), Does the cognitive approach to lie detection improve the accuracy of human observers?, Applied Cognitive Psychology 35(2), 385–392, https://doi.org/10.1002/acp.3777; Palena, N., Caso, L., Vrij, A., & Nahari, G. (2021), The Verifiability Approach: a meta-analysis, Journal of Applied Research in Memory and Cognition 10, 155–166, https://doi.org/10.1037/h0101785 [as cited in Vrij et al., 2022]; Hartwig, M., Granhag, P. A., & Luke, T. J. (2014), Strategic use of evidence during investigative interviews: the state of the science, in D. C. Raskin, C. R. Honts & J. C. Kircher (Eds.), Credibility Assessment: Scientific Research and Applications (pp. 1–36), Oxford: Academic Press [as cited in Vrij et al., 2022]; Vrij, A., Granhag, P. A., Ashkenazi, T., Ganis, G., Leal, S., & Fisher, R. P. (2022), Verbal lie detection: its past, present and future, Brain Sciences 12(12), 1644, https://doi.org/10.3390/brainsci12121644; Vrij, A., Hartwig, M., & Granhag, P. A. (2019), Reading lies: nonverbal communication and deception, Annual Review of Psychology 70, 295–317, https://doi.org/10.1146/annurev-psych-010418-103135; DePaulo, B. M., Kashy, D. A., Kirkendol, S. E., Wyer, M. M., & Epstein, J. A. (1996), Lying in everyday life, Journal of Personality and Social Psychology 70(5), 979–995, https://doi.org/10.1037/0022-3514.70.5.979; Serota, K. B., Levine, T. R., & Boster, F. J. (2010), The prevalence of lying in America: three studies of self-reported lies, Human Communication Research 36(1), 2–25, https://doi.org/10.1111/j.1468-2958.2009.01366.x