Blog

Can you really read lies on a face, like in Lie to Me?

The show had a real scientist behind it. Since then the research has answered the question — and the answer is not the one the series sells.

How often people actually catch a lie

The largest review of the field pooled 206 studies and 24,483 people asked to judge whether someone was lying. Average accuracy: 54 percent. A coin gives you 50.

The detail behind that number is more interesting than the number. People correctly spotted 47 percent of lies but 61 percent of truths. We are not lie detectors — we are truth-believers. Our default is to assume the other person is honest, and that default is what liars actually exploit.

What the show does to its viewers

In 2010 researchers put 108 people in front of three things: an episode of Lie to Me, an episode of another crime drama, and nothing at all. Then everyone judged the same set of honest and deceptive interviews. The Lie to Me group was no better at telling truth from lies — and was more likely to call honest people liars. The show does not teach detection. It teaches suspicion.

Why the face lets you down

There is no Pinocchio's nose. Reviews of the behavioural cues to deception keep finding the same thing: the signals exist, but they are faint, inconsistent, and swamped by ordinary differences between people. A nervous honest person looks exactly like a calm liar — often worse.

Micro-expressions are the part of the show most people remember. In the laboratory they are real. As a practical tool for spotting a liar in a conversation they have never lived up to the promise: too rare, too ambiguous, too easy to read into.

What does work

The field moved away from watching faces and towards asking better questions. Lying takes more mental effort than remembering, so the trick is to make that effort visible. A 2017 meta-analysis found this approach reached 71 percent overall accuracy against 56 percent for ordinary questioning — the difference between guessing and actually knowing something.

  1. Ask for more detail than a prepared story contains. Invented accounts are thin; real memories have texture nobody bothered to rehearse.
  2. Ask the unexpected. A liar prepares answers to the obvious questions, not to the one about who sat where.
  3. Listen for contradiction, not for nerves. What was said, compared with what was said five minutes ago — that is the signal.

Can it be trained?

Somewhat. A meta-analysis of 30 training studies found a small-to-medium improvement — clear, but no superpower. The useful part is which training worked: teaching people to attend to what was said produced the largest gains, while training built on facial and other non-verbal cues produced only small ones. In other words, the thing the show sells is the thing that trains worst.

Where you can practise

Everyday life is a terrible training ground for this, and for one specific reason: you almost never find out the answer. You suspect a colleague, and the story simply ends. Without knowing whether you were right, ten years of experience teach you nothing.

A game of Mafia is one of the rare places where the answer arrives. You question people, you listen to how the story holds together, you commit to a decision — and then the roles are revealed and you learn what you got wrong. It reproduces both conditions the research points to: attention to what people say rather than how they look, and an answer at the end.

Sources