Micro-Expressions: What Your Face Reveals Before You Can Hide It

A micro-expression is a flash of real emotion that crosses someone's face before they can arrange it into something more convenient. It lasts a fraction of a second, it happens without permission, and it takes the same shape whether the person grew up in Manchester or Manila. Here is what the research actually established, what each one looks like, and the one you should learn first.

Two warnings before we start, because this field attracts more confidence than it deserves. A micro-expression tells you an emotion was felt. It does not tell you why. And a single flash means far less than a pattern. Hold both of those and this becomes genuinely useful. Drop them and you turn into the person who accuses their partner of lying because of an eyebrow.

What Are Micro-Expressions and Where Did the Research Come From?

The story starts with a film reel and a patient nobody believed was in danger.

In 1969, Paul Ekman and Wallace Friesen were reviewing footage of a psychiatric patient who had asked for a weekend pass home. She was calm, she was pleasant, she gave every sign of recovery. She later admitted she had been planning to end her life. Going back through the film frame by frame, Ekman and Friesen found what everyone in the room had missed in real time: a flash of anguish across her face, gone in a fraction of a second, buried under a smile.

That is the whole idea in one anecdote. The face does two things at once. It runs a performance you control, and underneath it, it leaks the thing you are actually feeling.

To study that properly, Ekman and Friesen needed a way to describe faces that did not depend on interpretation. What they built was the Facial Action Coding System, or FACS, which breaks every possible facial movement into numbered action units tied to specific muscles. Not "she looked sad," but which muscle contracted, how hard, and for how long. It turned face-reading into something you could measure and disagree about productively. FACS is still the backbone of the field and the Ekman Group maintains it today.

The universality claim came from separate work. Ekman studied the Fore people of Papua New Guinea in the late 1960s, a group with almost no exposure to Western media at the time. If facial expressions were purely learned from culture, their emotional expressions should have looked unfamiliar. They did not. The core set matched.

That finding has been argued over since, and the honest position is that the argument is not settled. Critics point out that recognition studies often use forced-choice formats which inflate agreement, and that context shifts how a face is read. What survives the criticism is the useful part: a handful of expressions are produced similarly enough across very different populations that you can learn to spot them without knowing someone's culture first.

The measured speed. Fast enough that you will miss most of them untrained, and this has been measured rather than guessed. Xun-bing Shen, Qi Wu and Xiao-lan Fu tested it directly in a study published in the Journal of Zhejiang University Science B in 2012, showing participants expressions at durations from 20 milliseconds upward and measuring what they could actually identify.

Recognition accuracy climbed steadily with duration and then flattened out at around 200 milliseconds. Beyond that point, extra time added almost nothing, which gave the researchers a principled upper boundary: about a fifth of a second is where a micro-expression stops being micro and becomes an ordinary expression you would have noticed anyway.

At the fast end the numbers are humbling. In their first experiment, untrained participants sat around 52 to 66 percent accuracy at 40 milliseconds. Which is to say a coin flip with better lighting.

What Are the Seven Universal Micro-Expressions?

Seven, in the standard set. Each one has a signature, and the signature is muscular rather than vibe-based, which is what makes it learnable. Here is what to look for on each.

1. Happiness. The mouth corners pull up and back, and this is the easy part to fake. The tell is the eyes. Genuine happiness contracts the muscle that circles the eye socket, which narrows the eye opening, raises the cheeks and creases the outer corners. A social smile moves the mouth and leaves the eyes wide and unchanged. Once you have seen the difference you cannot unsee it in photographs.

2. Sadness. The hardest one to fake, which makes it valuable. The inner corners of the eyebrows pull up and together, producing a small triangle of tension in the middle of the forehead. The upper eyelids droop, the mouth corners pull down, and the whole face loses tone. Most people cannot voluntarily lift only the inner brow corners. If you see it, it is usually real.

3. Surprise. The briefest of the set by nature. Eyebrows go straight up and curve, the eyes widen so you may see white above the iris, and the jaw drops with the mouth loose rather than tensed. Real surprise is over in about a second. A surprise that lingers on the face is being performed for your benefit.

4. Fear. Frequently confused with surprise, and the difference is worth learning. In fear the eyebrows raise and draw together, which flattens them into a straight line rather than the curve of surprise. The upper eyelids lift hard, the lower lids tense, and the lips stretch back horizontally toward the ears. Surprise says what. Fear says something bad is coming.

5. Anger. The brows pull down and together, creating vertical lines between them. The upper eyelids raise while the lower lids tense, producing that hard stare. The lips do one of two things: press into a thin tight line, or square off and open. The giveaway is that the eyes and brows work in opposite directions, brows down, lids up.

6. Disgust. Concentrated in the nose and upper lip. The nose wrinkles, the upper lip lifts, and the cheeks push up. It is the face you make at a smell, and it appears more or less unchanged when someone finds an idea or a person repellent. The body treats moral disgust and rotten milk with strikingly similar equipment.

7. Contempt. The only one in the set that is reliably one-sided. One mouth corner tightens and lifts, producing a half-smirk while the other side of the face stays still. That asymmetry is the signature, and it is the reason contempt gets its own section below.

Why Is Contempt the Most Important One to Recognise?

Because of what it means, and because of what it predicts.

Start with the anatomy, since it is the reason contempt is the easiest one to catch. Every other expression in the set is broadly symmetrical. Contempt is not. One corner of the mouth tightens and pulls up while the rest of the face does nothing, and that lopsidedness is visually loud even at speed. If you only learn one micro-expression, this is the one that will actually earn its keep.

Now the meaning. Anger says you did something that hurt me. Contempt says something colder: I am above you. It is not a response to an action, it is a position on a person. Anger is an argument. Contempt is a verdict that the argument is beneath you.

That distinction is why it shows up so heavily in relationship research. John Gottman spent decades observing couples and identified four communication patterns he called the Four Horsemen: criticism, contempt, defensiveness and stonewalling. Of the four, contempt is the single strongest predictor of a relationship ending. Gottman calls it sulfuric acid for love, which is dramatic, and about right.

The reason it corrodes so fast is structural. You can come back from criticism, because criticism is about a behaviour and behaviours can change. Contempt is about your worth as a person, and there is no behaviour that fixes being looked down on. It also travels: contempt in one direction reliably breeds it in the other.

Which brings up the uncomfortable use of this. Watch your own face. If a flash of contempt is what happens when your partner starts talking, that is information about you, and it is worth taking more seriously than anything you might spot on them.

One caution before you go hunting for it. A flash of contempt in a single conversation means someone felt superior for a moment, which is a very human thing to do and not a diagnosis of the relationship. What matters is frequency. Contempt as an occasional visitor is normal. Contempt as the default setting is the finding.

How Do You Train Yourself to See Micro-Expressions?

The genuinely good news is that this trains. The Shen, Wu and Fu study did not just measure how bad people are at baseline, it measured what happened after intensive practice. Accuracy at 160 milliseconds rose to roughly 89 percent, and at 200 milliseconds and above participants were pushing past 92 percent. That is a large gain from a small intervention, and it is one of the better-supported claims in this whole area.

What actually works:

• Learn the muscles, not the mood. Do not practise recognising "he seems annoyed." Practise recognising brows down and together with upper lids raised. Muscular descriptions are checkable. Impressions are not, and impressions are how you end up confirming whatever you already suspected.
• Slow down video and work forward. Take an interview or a press conference, play it at quarter speed, and watch the face during questions rather than answers. Then rewatch at normal speed and see whether you can now catch what you found slowly. That second pass is where the learning happens.
• Use the mirror on yourself. Produce each expression deliberately and watch which muscles you can and cannot control. Try to raise only the inner corners of your eyebrows. Most people cannot. Discovering which movements are outside your voluntary control teaches you which ones are worth trusting on other people.
• Watch the transitions, not the poses. Micro-expressions live in the gap between one arranged face and the next. The moment after a question lands and before the answer starts is where the leak happens.
• Practise on faces where you know the answer. Reaction footage where you can verify the outcome gives you feedback. Practising on your suspicious colleague teaches you nothing, because you cannot grade yourself.

And what to ignore, which matters just as much:

• Ignore anything below the neck for now. Crossed arms mean cold rooms as often as defensiveness. Body language folklore has a far weaker evidence base than facial coding and will pollute your reads.
• Ignore anything you have to squint to see. If you are not sure whether it happened, it did not happen. Reconstructing a flash you half-caught is how you invent evidence for a conclusion you already reached.
• Ignore single instances entirely at first. Until you can catch expressions reliably, treat everything you spot as practice rather than data.

Can Micro-Expressions Be Faked or Suppressed?

Suppressed partially, faked poorly, and the failures are the interesting part.

Take suppression first. When someone tries to hide an emotion, the impulse to express it has already fired before the effort to stop it begins. Suppression is late by design. What you get is a truncated expression, an intensity that drops off unnaturally fast, or a movement that starts and gets cut mid-way. The stopping is often more visible than the expression would have been.

Then there is masking, which is the more common move. Rather than showing nothing, the person covers the real emotion with a more acceptable one. The smile over anger is the classic. What tends to leak is the upper face, because the lower face is far easier to control than the brow and eye region. A mouth arranged into a smile above a brow doing something else is the most common tell you will encounter in ordinary life.

Faking is harder than it sounds. Some expressions involve muscles most people cannot fire on command, which is why deliberate sadness usually looks wrong and why a posed smile misses the eye contraction. Actors train specifically to get around this, which is worth remembering: skilled performance exists, and some people have had a lot of practice.

The genuinely important limitation is different, though, and it gets skipped constantly. A micro-expression tells you an emotion occurred. It does not tell you what caused it. Fear on the face of someone being accused of something is exactly as consistent with an innocent person frightened of not being believed as with a guilty one. This is why treating facial leakage as lie detection goes wrong so often, and why our companion guide on what deception research actually shows spends most of its time on story structure rather than faces.

What Should You Do When You See One?

Almost always: nothing, immediately. Which is unsatisfying and correct.

The instinct is to name it. You saw the flash, you feel certain, and calling it out feels like honesty. In practice it goes badly for two reasons. You might be wrong, and even if you are right, announcing that you read someone's face tells them to manage their face better around you. You do not get more information. You get a better performance.

Here is what to do instead.

1. Note it and keep going. Register what you saw and when. Do not change your expression or your tone. The conversation continuing normally is what keeps the channel open.
2. Ask what came just before it. Micro-expressions are responses. The useful question is not what you saw but what triggered it: which word, which name, which part of the question. That timing is the actual data.
3. Wait for repetition. One flash is noise. The same expression at the same topic three times is a pattern, and patterns are what you can act on.
4. Follow it with curiosity, not accusation. If contempt appeared when you mentioned their colleague, you have learned to ask a gentle open question about that colleague later. You have not learned to say "you looked contemptuous."
5. Let it change your questions, not your verdict. The right use of a micro-expression is that it tells you where to look. It is a pointer, not a conclusion.

There is one situation where this changes, and it is worth stating plainly. If what you keep seeing is contempt or disgust, repeatedly, from someone close to you, that is not a puzzle to investigate. That is information about how that person holds you, and it usually shows up alongside other things: your reality getting rewritten, your memory questioned, small cruelties explained away. Our guide on the signs of gaslighting covers that wider pattern.

One last thing worth saying about all of this. Learning to read faces is a skill, and like most skills it can be used badly. Used well, it makes you a better listener, because you notice when a topic lands hard and you slow down. Used badly, it turns every conversation into surveillance and every twitch into evidence. The people who get real value out of this are the ones who use it to ask better questions, not to win arguments they have already decided.

What Else Do People Ask?

How long does a micro-expression last?

Roughly 40 to 500 milliseconds, with about a fifth of a second as the practical upper boundary. A 2012 study by Shen, Wu and Fu found recognition accuracy climbed with duration and then levelled off at around 200 milliseconds, which is where an expression stops being micro and becomes one you would notice anyway. Untrained observers sit near chance at the fastest end.

Are micro-expressions really the same in every culture?

Broadly yes for the core set, though the claim is more argued over than popular accounts suggest. Ekman's fieldwork with the Fore people of Papua New Guinea found matching expressions in a group with little Western media exposure. Critics note that forced-choice testing inflates agreement and that context changes how a face is read. The useful residue is that these seven are produced similarly enough to be learnable across cultures.

Can you learn to see micro-expressions, or is it a natural talent?

It trains, and the training effect is well documented. In the Shen, Wu and Fu study, intensive practice raised recognition accuracy at 160 milliseconds to roughly 89 percent, well above untrained baseline. The method that works is learning specific muscle movements rather than general impressions, and practising on video you can slow down and replay.

Does a micro-expression mean someone is lying?

No, and this is the most common mistake. A micro-expression tells you an emotion was felt. It says nothing about why. Fear on an accused person's face fits an innocent person afraid of not being believed just as well as a guilty one. Facial leakage points you toward a topic worth exploring. It does not establish what happened.

Which micro-expression should I learn first?

Contempt. It is the only one in the set that is reliably one-sided, so the asymmetry makes it the easiest to catch at speed, and it carries the most weight. Gottman's research identifies contempt as the strongest single predictor of a relationship ending, because unlike criticism it targets the person rather than the behaviour.

Sources: Shen, X. B., Wu, Q., and Fu, X. L. (2012), "Effects of the duration of expressions on the recognition of microexpressions," Journal of Zhejiang University Science B, 13(3), 221-230 (pmc.ncbi.nlm.nih.gov/articles/PMC3296074). Ekman, P., and Friesen, W. V., Facial Action Coding System, maintained by the Ekman Group (ekman.com/facial-action-coding-system). The Gottman Institute on contempt and the Four Horsemen (gottman.com/blog/the-four-horsemen-contempt).

For self-reflection and educational purposes only. Nothing here is a diagnosis or a substitute for professional mental health support. Only a qualified clinician can assess or diagnose any condition. If you feel unsafe, please contact a local domestic abuse or crisis helpline.