What are micro-expressions, and can they reveal lies?

§ 1What a micro-expression is

Faces usually hold an expression for somewhere between half a second and a few seconds. A micro-expression is much shorter. The most cited working definition, from an analysis of fast expressions by Yan and colleagues (2013), counts an expression as micro if it lasts less than 500 milliseconds in total. Older conventions used a range of 1/25 to 1/5 of a second, that is 40 to 200 ms.

The idea goes back to studies of filmed therapy sessions in the 1960s, where slowed-down film showed expressions that flashed by too fast to notice at normal speed. Researchers suggested that such brief “leaks” might show a feeling a person was trying to hide. That suggestion made micro-expressions famous, and it is the part that later research has tested most closely.

§ 2How fast is fast? What studies found

Shen, Wu and Fu (2012) flashed expressions for between 20 and 300 milliseconds. Recognition rose with duration and levelled off at about 200 ms, and practice improved performance. Very short flashes are hard, but they are not invisible.

In real behaviour, micro-expressions are uncommon. Porter and ten Brinke (2008) coded 697 expressions from people showing genuine or falsified emotions. Only about 2% were micro-expressions, they appeared in about a fifth of the participants, and each involved only the upper or the lower half of the face, not the whole face.

§ 3Can micro-expressions reveal lies?

This is the claim made famous by television, and the evidence does not support it. A meta-analysis by Bond and DePaulo (2006), covering 206 documents and 24,483 judges, found that people sort statements into lies and truths correctly 54% of the time: only slightly better than the 50% expected by guessing.

Micro-expressions do not close that gap. In Porter and ten Brinke’s study they were not more frequent when people were deceiving. In a controlled test by Jordan and colleagues (2019), people trained with a well-known micro-expression training tool were no better at detecting lies than people given bogus training or none, and overall accuracy was slightly below chance. A review by Vrij, Hartwig and Granhag (2019) describes nonverbal cues to deceit as faint and unreliable.

There is also a chain problem, laid out by Burgoon (2018). For a micro-expression to expose a lie, the lie must cause an emotion, the emotion must show on the face, the expression must be impossible to control, frequent enough to see and different between liars and truth-tellers. The evidence does not support every link in that chain.

§ 4What recognising them can still teach you

Recognising brief expressions is a trainable perceptual skill. In randomised studies, practice with feedback improved the recognition of flashed expressions, and the gain was still there two to three weeks later (Matsumoto and Hwang, 2011). What improves is the skill itself: noticing that a face moved and which movement it was. Döllinger and colleagues (2021) found large gains on the trained task but no transfer to other skills, such as recognising emotional cues in patient conversations.

• A brief expression shows that something happened in the face, not why. Surprise at a question, effort, discomfort, a memory or nothing in particular can all cause one.
• An expression that does not match the words is an inconsistency to ask about, not evidence of deception.
• Facial movements do not map one-to-one onto feelings: context, person and culture change their meaning (Barrett and colleagues, 2019).

Questions and answers

Are micro-expressions real?

Yes. Very brief expressions do occur, but they are rare and usually partial. In one study only about 2% of coded expressions were micro-expressions.

Can I learn to spot lies from someone’s face?

No method has been shown to do this reliably. People average about 54% accuracy, close to chance, and micro-expression training did not improve lie detection in a controlled test.

How long does a micro-expression last?

Most researchers today use less than half a second (500 ms). Classic definitions used 1/25 to 1/5 of a second.

Is micro-expression training useless, then?

It improves what it trains: noticing brief facial movements on practice faces. That can make you more attentive to faces, as long as you treat what you notice as a question, not an answer.