A micro-expression is an involuntary facial expression that flashes across a person's face in a fraction of a second — typically between 1/25th and 1/5th of a second — before they regain control of their features. Where a normal expression is something you show, a micro-expression is something that leaks. That's what makes them so compelling: they can reveal an emotion someone is actively trying to conceal.
A short history of a very short expression
Micro-expressions were first documented in 1966, when researchers Ernest Haggard and Kenneth Isaacs scanned films of therapy sessions in slow motion and noticed "micromomentary" expressions that were invisible at normal speed. The idea was then developed and popularized by psychologist Paul Ekman, whose cross-cultural research suggested that a set of basic emotions — happiness, sadness, anger, fear, surprise, disgust, and contempt — produce recognizably similar facial expressions across cultures.
In 1978, Ekman and Wallace Friesen published the Facial Action Coding System (FACS), which breaks every visible facial movement into numbered "action units" — the raising of an inner eyebrow, the tightening of a lip corner, the flaring of a nostril. FACS gave researchers a common language for describing expressions objectively, and it remains the backbone of facial-expression research (and of modern facial-analysis AI) today.
Why micro-expressions matter
The theory goes like this: when you feel an emotion, your facial muscles begin expressing it automatically. If the emotion is one you want to hide — fear during a confident denial, contempt during warm words — you suppress the expression. But suppression takes a beat, and in that beat, the true expression escapes. Micro-expressions have accordingly been studied in high-stakes settings: security screening, criminal interviews, negotiation, and clinical psychology.
Ekman's group even built training programs designed to teach people to catch them. Training helps, but the honest reality is that most people miss most micro-expressions. At a fifth of a second or less, they sit at the edge of human perception — noticing one in real time, while also listening to what the person is saying, is genuinely hard.
Key nuance: a micro-expression reveals a concealed emotion — not a lie. A flash of fear during "I didn't do it" might mean guilt, or it might mean fear of not being believed. Research by Porter and ten Brinke has shown emotional leakage appears in both genuine and falsified expressions. The signal is real; the interpretation is where humility is required.
Can you learn to read micro-expressions?
To a degree. Training studies show measurable improvement in recognizing the seven universal expressions at speed. If you want to practice the old-fashioned way, the core skills are:
- Watch the whole face, not just the eyes — contempt lives in a unilateral lip corner, disgust in the nose and upper lip.
- Establish a baseline. Read how someone's face behaves when nothing is at stake before you read it under pressure.
- Look for mismatches between words and face — a "yes" accompanied by a fleeting head shake or a smile that doesn't reach the eyes.
- Time is the tell. Genuine expressions tend to be symmetrical and smoothly faded; suppressed ones are fast, asymmetric, and abruptly cut off.
How AI detects micro-expressions
This is one place machines hold a real, structural advantage over people. A camera capturing 30 or 60 frames per second doesn't blink, doesn't get distracted, and doesn't have to listen at the same time. Modern computer-vision models trained on FACS action units can:
- Track dozens of facial landmarks frame-by-frame,
- Detect action-unit activations lasting only a few frames — squarely in micro-expression territory,
- Compare every moment against the person's own baseline from earlier in the same recording,
- Do it identically on the thousandth face as on the first.
In other words: the detection problem — catching the flash — is largely solved by machines. The interpretation problem — what the flash means — remains as open for AI as it is for humans, which is why no serious tool should claim micro-expressions prove deception. We dig into that broader question in Do Lie Detector Apps Actually Work?
Micro-expressions in SusAI
Micro-expression analysis is one of the five behavioral channels SusAI scores on every scan, alongside eye movement, vocal stress, speech patterns, and body language. Record someone answering a question and the AI flags the fleeting expressions it caught, scores the channel, and folds it into an overall truth score with a verdict — Truthful, Suspicious, or Deceptive — plus a breakdown of exactly what it saw.
It's the fastest way to experience what FACS-style analysis actually looks like on a real face — framed, as always, for entertainment and curiosity rather than judgment. Signals, not verdicts on anyone's character.