Micro expressions are involuntary facial expressions that flash across your face in roughly 40 to 200 milliseconds, far too fast for most people to consciously control or suppress. They reveal a person’s genuine emotional state for a fraction of a second before the face returns to a neutral or masked expression. Unlike the expressions you hold deliberately (a polite smile, a sympathetic frown), micro expressions are produced by a separate neural pathway tied to your brain’s emotional centers, which is why they’re so difficult to fake or hide.
How Long They Last
The defining feature of a micro expression is its speed. Most researchers place the duration somewhere between 1/25 of a second (40 milliseconds) and 1/2 of a second, though the exact cutoff is debated. Experimental evidence suggests the critical boundary between a micro expression and a regular (macro) expression falls at about 200 milliseconds. Anything shorter than that tends to be processed differently by the brain, perceived as a flash rather than a deliberate display. For perspective, a single eye blink takes about 300 to 400 milliseconds, so micro expressions can come and go faster than you blink.
Regular facial expressions, by contrast, typically last between half a second and four seconds. They’re the ones you notice in conversation: a friend’s grin, a coworker’s furrowed brow. Micro expressions occupy the gap between a completely invisible muscular twitch and these longer, conscious displays.
Why Your Face Betrays You
Your brain controls voluntary and emotional facial movements through two separate neural pathways. When you deliberately smile for a photo, the motor cortex sends signals through a pathway that runs along the front of the brainstem. But when you experience a genuine emotion, a different route activates. Signals originate from deeper structures, particularly the amygdala and the medial prefrontal cortex, and travel through a distinct pathway that runs along the back of the brainstem.
Because these two systems are anatomically independent, one can fire before the other catches up. When you feel a flash of anger or disgust but want to appear calm, your emotional pathway produces the expression first. Your voluntary system then suppresses it, but not before the emotion briefly surfaces on your face. That momentary leak is the micro expression. This is also why people with certain types of brain damage can lose control of one system while the other still works: some patients can smile on command but not when genuinely amused, or vice versa.
The Seven Emotions They Reveal
Psychologist Paul Ekman, whose work shaped most of what we know about micro expressions, identified seven core emotions that produce distinct facial patterns:
- Anger: lowered brows, tightened lips, flared nostrils
- Contempt: a one-sided lip raise (the only asymmetrical expression of the seven)
- Disgust: wrinkled nose, raised upper lip
- Enjoyment: raised cheeks, crow’s feet around the eyes, upturned mouth
- Fear: raised and drawn-together brows, widened eyes, slightly open mouth
- Sadness: inner corners of the brows pulled up, drooping eyelids, downturned mouth
- Surprise: raised brows, wide eyes, dropped jaw
These are cataloged using the Facial Action Coding System (FACS), a tool Ekman and his colleague Wallace Friesen developed in 1978. FACS breaks every possible facial movement into 46 independent “action units,” each corresponding to a specific muscle or muscle group. A trained coder watches video frame by frame and identifies which action units fired, how long they lasted, and when they started and stopped. This system gives researchers a standardized vocabulary for describing exactly what a face did, rather than relying on subjective impressions.
How Well People Spot Them
Most people are not naturally good at catching micro expressions. In controlled studies, untrained participants correctly identified micro expressions about 49% of the time, essentially a coin flip. That’s not because people are unobservant. It’s because these expressions are genuinely fast and subtle, often involving only part of the face for a tenth of a second.
Training makes a significant difference. After completing structured micro expression training programs, participants in the same studies improved to about 75% accuracy. These programs typically involve watching video clips of faces at various speeds, learning to associate specific muscle movements with specific emotions, and practicing until recognition becomes more automatic. The training produces large, measurable effects on performance, and some of that improvement appears to stick over time.
The most well-known training tool is the Micro Expression Training Tool (METT), developed by Ekman’s group. It walks users through each of the seven emotions, shows examples at full speed and in slow motion, and then tests recognition under time pressure. Studies evaluating METT have found large effect sizes, meaning the improvement isn’t marginal but represents a genuine shift in perceptual ability.
The Lie Detection Problem
Micro expressions are often associated with lie detection, and this is where the science gets considerably murkier. The popular idea is straightforward: if someone is lying, their true emotions will leak through micro expressions, and a skilled observer can spot the deception. Reality is more complicated.
For micro expressions to reliably indicate deception, several things would need to be true simultaneously. Lying would need to consistently produce specific negative emotions. Those emotions would need to reliably produce specific facial expressions. Those expressions would need to appear frequently enough to be caught. And catching them would need to actually distinguish lies from truth. Each link in that chain has problems.
Deception doesn’t always produce guilt or fear. Some people feel relief, pleasure, or even excitement when lying successfully, a phenomenon sometimes called “duping delight.” And even when lying does produce a negative emotion, that emotion doesn’t map neatly onto a single facial expression. Felt fear, for instance, can manifest as expressions associated with anxiety, anger, contempt, or surprise. A single emotion can produce multiple different expressions, and multiple emotions can produce the same expression. There is no one-to-one correspondence between what someone feels and what their face does.
Perhaps most importantly, the bulk of research on emotions during deception has actually studied regular, longer-lasting expressions rather than true micro expressions. The ability to detect deception from a sustained look of discomfort doesn’t translate to the ability to catch a 100-millisecond flash and correctly interpret what it means. Researchers have explicitly cautioned that micro expressions are “not the best way to catch a liar.”
The Debate Over Universality
Ekman’s original claim was that these seven facial expressions are universal, meaning people in every culture produce and recognize the same expressions for the same emotions. The foundational studies supporting this were conducted between 1969 and 1975 in small-scale, isolated societies in the Pacific. For decades, this was treated as settled science.
More recent research has challenged that conclusion. Studies conducted since 2008 across a wider range of societies have found considerably more diversity in how people interpret facial expressions. The methodological concern centers on how the original studies were designed. Participants were typically shown a face and asked to choose from a short list of emotion words. This format allows people to arrive at the “correct” answer through elimination or by matching broad qualities like whether a face looks pleasant or unpleasant, rather than truly recognizing a specific emotion.
When researchers use less constrained methods, letting participants freely describe what they see instead of picking from a list, agreement drops substantially. Himba people in Namibia, Hadza people in Tanzania, and Trobriand Islanders in Papua New Guinea rarely volunteered the emotion labels that universality theory would predict. Trobriand adolescents, for example, consistently interpreted the wide-eyed gasping face (the expected “fear” expression) as signaling an intent to attack rather than fear. This doesn’t mean facial expressions are meaningless across cultures, but it does suggest they carry more cultural variation than the universality framework allows.
AI and Automated Detection
Computer vision systems have made rapid progress in recognizing facial expressions, including micro expressions. AI models trained on facial data now achieve diagnostic accuracies ranging from about 85% to 95% when analyzing expressions alone, with some specialized systems reporting even higher numbers in narrow applications. Systems that combine facial analysis with other data sources, like voice patterns or brain imaging, consistently push above 95%.
Most of this work has focused on clinical applications rather than lie detection. Researchers are exploring whether automated expression analysis could help screen for conditions like autism or depression, where subtle differences in facial behavior may serve as early indicators. In one study comparing AI to human evaluators for identifying autism-related facial patterns, the AI system achieved 80.5% accuracy, while human experts reached 83.1% and non-experts 78.3%. The gap between humans and machines is narrowing, though context and nuance still give experienced human observers an edge in many real-world settings.