Body Language and Lying: What the Research Actually Shows

The most confident thing people believe about lying is also the most wrong

Ask a hundred people in any country how you spot a liar and you will get the same answer: they won’t look you in the eye. The Global Deception Research Team asked exactly that question across 58 countries in 2006, and gaze aversion was the single most common belief almost everywhere — in 51 of the 58. Lie Detector UK uses instruments rather than intuition, and the research below is the reason why.

It is also, as far as the evidence goes, false. Liars do not reliably look away. Some look away more. Some look away less, precisely because they know the stereotype and are managing it. The behaviour has no stable relationship with deception at all, which is the problem: a cue that goes in both directions is not a cue.

Five myths, and what the research found

1. “Looking up and to the left means they are making it up”

This is the most specific-sounding claim in the whole field, which is part of why it spread. It comes from neuro-linguistic programming, which proposed that eye movements map onto whether someone is recalling a memory or constructing one.

It has been tested directly. Richard Wiseman and colleagues published a study in 2012 that recorded people telling truths and lies, coded their eye movements, and looked for the pattern. There was none. A second experiment trained observers in the eye-movement rule before asking them to judge — training them made no difference to their accuracy. A third analysed real televised press conferences from people later proven to be lying about a missing relative. The pattern did not appear there either.

2. “They wouldn’t look me in the eye”

Covered above as a belief; worth adding what happens when you test it. Gaze aversion appears in the meta-analytic literature with an effect size indistinguishable from zero. Worse, it is one of the cues people can deliberately control — and a liar who has heard the stereotype will often overcorrect into unnaturally steady eye contact.

The practical effect is that this cue does not just fail. It can point the wrong way, because the people most motivated to manage it are the ones with something to hide.

3. “They touched their nose — the Pinocchio effect”

Self-touching, face-touching and nose-touching are genuinely common when people are uncomfortable. The trouble is the second half of the inference. Discomfort is not deception. Being accused of something you did not do is uncomfortable. Being asked about your marriage by a stranger is uncomfortable. Sitting in a room being watched is uncomfortable.

In DePaulo and colleagues’ 2003 review of 158 behavioural cues across decades of studies, face-touching behaviours did not emerge as a reliable indicator. Most cues in that review had effect sizes so small that they could not support a judgement about an individual person on a single occasion.

4. “Crossed arms means they are closing off”

There is no evidence base for this one at all. Arms get folded because chairs have no armrests, because rooms are cold, because it is comfortable, and because people fold their arms. It carries no information about honesty.

5. “They were sweating — guilty”

This is the myth that comes closest to something real, and it is worth being precise about why it still fails as a cue you can read across a table.

Sweating is one output of the autonomic nervous system, and autonomic arousal is measurable and is part of what a polygraph records. But arousal has many causes: fear of not being believed, the stakes of the conversation, the room, caffeine, embarrassment, anger at being asked. Arousal on its own tells you the person is aroused. It does not tell you why.

The difference between that and an examination is not that the instrument sees sweat better than you do. It is that the examination compares your response to a relevant question against your own responses to other questions, minutes apart, in the same room, in the same state. It is a within-person comparison. Watching someone sweat is a between-person guess.

Why we are all so bad at this

The headline figure comes from Charles Bond and Bella DePaulo’s 2006 meta-analysis, which pooled 206 studies covering around 24,500 individual judgements. Average accuracy was 54%. Chance is 50%.

Three findings inside that number matter more than the number itself:

1. Confidence does not track accuracy. People who are certain they can tell are not measurably better than people who admit they cannot.
2. Professionals are barely better than anyone else. Police officers, judges and customs officers score close to the general population in most studies. Experience in interviewing people is not the same as feedback on whether you were right — and without feedback, you cannot improve.
3. People are better at spotting truths than lies. Which sounds encouraging until you notice the reason: most of us assume most people are telling the truth most of the time, so we are right by default whenever they are.

What the research does support

Not nothing — but far less than the body-language industry sells, and almost none of it is visual.

The cues with the least-weak support are verbal and content-based rather than physical: the amount of detail in an account, whether it can be told backwards as coherently as forwards, whether it survives an unexpected question about a detail the person had no reason to rehearse. Aldert Vrij and colleagues have spent two decades on approaches that increase what they call cognitive load — making the act of lying harder, then watching whether the account holds together — rather than watching for tells.

Even there, the honest framing is that these methods shift accuracy upward from a poor baseline. They do not produce certainty, and they are interview techniques rather than something you can apply to your partner over dinner.

What an instrument does differently

A polygraph does not read faces, and it is not a lie detector in the sense the name implies. It is a recorder. Three channels run continuously through the examination:

• Respiration — two tubes across the chest and abdomen recording the depth and rhythm of each breath.
• Cardiovascular activity — a cuff tracking pulse rate and relative blood pressure throughout, not only at the moment a question is asked.
• Electrodermal activity — fingerplates measuring the sweat glands’ response, which shifts within seconds of a relevant question.

Four things follow from that design, and each one addresses a specific failure of watching behaviour:

1. It is continuous. A gesture is one moment. A chart is a recording of the whole examination, so a reaction is read against what came before and after it.
2. It is a within-person comparison. Your responses to relevant questions are scored against your own responses to comparison questions, in the same room, minutes apart. Nervousness raises the whole trace, not one part of it — which is why being nervous does not produce a failed test.
3. The questions are agreed in advance. Every relevant question is reviewed with you before the instrument is attached. Surprise is not part of the method.
4. It can return no answer. Inconclusive is a real outcome, reported as such. Someone reading body language never returns inconclusive — they always come away with an impression.

None of that makes an examination infallible. A validated technique run by an accredited examiner sits at 95–98% — the honest figure, and not 100%, because nothing is 100%. But it is a different order of thing from 54%.

If you are watching someone for tells right now

The most useful thing in this article is probably this: stop collecting evidence you cannot interpret. Every day spent cataloguing glances makes you more certain and no better informed, and certainty built that way is what people act on and regret.

Ask the question directly, and if the answer will not settle it, get it tested properly — by us or by anyone accredited. What you should not do is keep watching.

Sources: Bond, C. F. & DePaulo, B. M. (2006), “Accuracy of Deception Judgments”, Personality and Social Psychology Review. DePaulo, B. M. et al. (2003), “Cues to Deception”, Psychological Bulletin. Wiseman, R. et al. (2012), “The Eyes Don’t Have It: Lie Detection and Neuro-Linguistic Programming”, PLoS ONE. The Global Deception Research Team (2006), “A World of Lies”, Journal of Cross-Cultural Psychology. Vrij, A., Fisher, R. P. & Blank, H. (2017), “A cognitive approach to lie detection: A meta-analysis”, Legal and Criminological Psychology.