Facial expression glossary: action units, Duchenne smiles, and what survived testing

You meet these words in the places men go when a photo feels off: a face-analysis video, a dating-photo thread, or a confident-looking chart explaining why your eyes supposedly reveal your character. The vocabulary sounds scientific, which makes it hard to tell what is a real measurement and what is just a good sales sentence.

Here is the useful answer first: facial movement can be coded with a serious system, but movement is not a transparent window into a person's feelings or honesty. The coding system is real. The mind-reading it got sold as is not. This glossary separates the two so you can use the research without letting it use you.

Key numbers

• 1978 — the Facial Action Coding System was published; its scoring version dates from 2002. Source
• 527 pages — the FACS manual sold by the Paul Ekman Group, beside a 197-page guide. Source
• 50–100 hours — the rights holder's estimate for learning FACS alone. Source
• 44 or 46 — peer-reviewed sources disagree on how many action units exist. 44-source 46-source
• 97% — cosmetic patients who contracted the eye muscle while posing a big smile. Source
• 54% — the U.S. Government Accountability Office's figure for spotting deception. Source
• 1 record — what a PubMed search for “resting bitch face” returns. Source

Table of contents

• How to read this glossary
• The coding vocabulary: FACS and action units
• Smiles, and the Duchenne story
• Micro-expressions and the lie-detection claim
• What a face can and cannot tell you about emotion
• The words for a neutral face
• Terms that circulate without a source
• How to cite this page
• The bottom line
• Sources

How to read this glossary

Every entry carries one of three labels, and the label grades the evidence, not your face. Measured means a defined observation or test supports the claim. Contested means credible sources disagree, or the answer changes with the task. No evidence found means we looked for an authoritative source and found none worth citing.

The coding vocabulary: FACS and action units

Facial Action Coding System (FACS)

Measured — FACS is a descriptive system for coding visible facial movement: it records what moved, not what the person felt. The manual was first published in 1978 and revised in 2002, and learning it takes an estimated 50–100 hours. Official FACS overview

Real coding is slower than a webcam badge. Intensity is scored from 0 to 5, certification requires .70 agreement with an expert, and one minute of video takes 1–2 hours by hand.

Action unit (AU)

Measured — An action unit is a named pattern of muscle movement, not an emotion, and the names are deliberately dull: AU1 Inner Brow Raiser, AU4 Brow Lowerer, AU6 Cheek Raiser, AU12 Lip Corner Puller, AU43 Eyes Closed. Peer-reviewed sources report either 44 or 46 units, so the count is contested. AU names and the 44 count 46-count review

Forum language turns “AU12 detected” into “genuine happiness detected,” which is a category error: the first describes a lip movement, the second adds an emotion claim that only context can support.

Smiles, and the Duchenne story

Duchenne smile

Contested — A Duchenne smile combines AU12, the lip-corner puller, with AU6, the cheek raiser around the eye. The term comes from a 1990 paper, but its popular version, that AU6 cannot be faked, did not survive a direct test. Original definition Direct test

In that test, 97% of cosmetic patients contracted the eye muscle when asked to pose their biggest smile, and fell to 19% only after botulinum toxin paralysed it. A yearbook sample reported about 45% Duchenne smiles in a posed photo session, exactly the situation the marker was supposed to rule out.

That does not make smiles fake. It makes AU6 evidence that a muscle moved rather than a truth serum, and a felt-versus-posed experiment found similar numbers of Duchenne and non-Duchenne smiles whether people smiled spontaneously or were told to pose. Test of felt versus posed smiles

None of this argues for a neutral face. A study of 303 photographed targets judged by 987 observers found smiling faces gave slightly more accurate personality judgments, and a forced-choice study found a smile can help attractiveness judgments. Smiling and attractiveness For your own pictures, start with our Photo guide.

Micro-expressions and the lie-detection claim

Micro-expression

Contested — Micro-expressions are very brief facial movements, first described in psychotherapy films as “micromomentary facial expressions” in 1966. Their famous duration is not settled: one standard citation gives 1/25 to 1/5 of a second, later work an upper limit of 0.5 seconds. Historical and duration review

How rarely they appear is the bigger problem. Across 104,550 video frames and 697 expressions, they showed up in only 2% of expressions and 21.95% of participants, while emotionally inconsistent leakage appeared in 100% of participants and lasted longer than the definition allows. Microexpression evidence Official definition page

Micro-expression lie detector

No evidence found — The claim that a trained observer can read a lie off a fleeting facial signal is not supported by the deception literature. A GAO review of four meta-analyses covering more than 400 studies put human accuracy at 54%, and a later review of more than 200 studies found nonverbal judges no better than chance. GAO review Research review

The scale of that mistake is worth seeing. By fiscal year 2012 the TSA programme employed about 3,000 behaviour-detection officers at 176 airports, and had cost approximately $900 million since 2007. A scientific-looking checklist manufactures confidence without accuracy, so do not use it to interrogate a date.

What a face can and cannot tell you about emotion

Facial expression as an emotion read-out

Contested — A facial configuration is not a universal read-out of an emotion. People do smile when happy and scowl when angry more often than chance, but the same configuration carries different meanings across cultures, situations and people. Psychological Science in the Public Interest review

The measured link is weaker than the folklore. For the six classic basic emotions, a meta-analysis found average co-occurrence effects of .13 for full FACS-measured expressions and .23 with partial ones, with an average intensity correlation of .30. Spanish controls matched faces to emotions on 83%–100% of trials, while smile-to-happiness matches reached only 58% and 56% on two island populations.

Who is looking matters too. Emotions are recognised more accurately when expresser and perceiver share a national, ethnic or regional group, so your reading of a face is partly about the relationship between two people, not one face's geometry. Meta-analysis

The words for a neutral face

Resting bitch face

No evidence found — “Resting bitch face” is a popular phrase rather than a settled scientific construct, and a PubMed title-and-abstract search returns exactly 1 record, a narrative review in cosmetic dermatology. PubMed search

We are not dismissing the experience the phrase describes: something measurable sits underneath it. Across 597 coded faces, perceived resting negativity predicted attractiveness more strongly for women, while perceived threat predicted judgments more strongly for men. That describes how strangers read a still face; it is not a verdict on your character.

Emotion overgeneralisation

Measured — Emotion overgeneralisation is the researchable version of that idea. Neutral faces that strangers read as threatening objectively resemble anger, faces read as negative resemble disgust and fear, and positive-looking ones resemble happiness. Neutral-face mechanism

A blink is not automatically nerves either. Measured blink rates were 17 per minute at rest, 26 during conversation and 4.5 while reading. Blink-rate study If your photos keep landing closed-off, how to look more approachable is the better question.

Terms that circulate without a source

We checked the rights holder's public pages, peer-reviewed reviews, PubMed, and the reports below. These five claims did not survive the check.

• “The human face can make more than 10,000 expressions.” The traceable version is a combinatorial count built on 43 action units, not expressions anyone observed, and 43 matches neither the 44 nor the 46 in peer-reviewed sources.
• “A microexpression lasts 1/25 of a second.” The same official page also says 0.5 seconds or less, and the research gives ranges, not one defended figure.
• “About 71% of people can deliberately produce a Duchenne smile.” We found the paper and its DOI but no primary text we could read, so the 97% test is the stronger figure.
• “Wide-smiling baseball players lived about 7 years longer.” A replication across samples of 224, 527 and 13,530 players found no effect once birth year was controlled.
• “A single authoritative AU count.” Peer-reviewed sources give 44 and 46, media reports give 43, and the rights holder's page gives none.

We are not calling them false. Nobody has earned the right to repeat them so confidently.

How to cite this page

Cite this glossary as an editorial reference page from Real World Appeal, with its title and updated date. For a scientific claim, follow the link to the original paper; this page is a map, not a replacement for the evidence.

The bottom line

The coding system is real. The mind-reading it got sold as is not. FACS gives researchers a disciplined vocabulary for movement, and action-unit intensity can be measured from 0 to 5. Duchenne smiles, microexpressions and neutral-face cues are hypotheses, not permission to infer honesty, desire or character from a single frame.

To see how your own face lands in a first impression, try the free test. It is not a clinical instrument and a photo-based read has limits, but it beats treating a viral facial-expression number as mind-reading.

Sources

• Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., & Pollak, S. D. (2019). Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements. Psychological Science in the Public Interest, 20(1), 1–68. Linked source
• Ekman, P., Davidson, R. J., & Friesen, W. V. (1990). The Duchenne smile: Emotional expression and brain physiology. Journal of Personality and Social Psychology, 58(2), 330. Linked source
• Elfenbein, H. A., & Ambady, N. (2002). On the universality and cultural specificity of emotion recognition: A meta-analysis. Psychological Bulletin, 128(2), 203–235. Linked source
• Etcoff, N. L., Stock, S., Krumhuber, E. G., & Reed, C. L. (2021). The effect of Botox on the quality of facial movement. Frontiers in Psychology. Linked source
• Golle, J., Mast, F. W., & Lobmaier, J. S. (2014). Something to smile about: The interrelationship between attractiveness and emotional expression. Cognition & Emotion. Linked source
• Hester, N. (2019). Perceived negative emotion in neutral faces: Gender differences and the role of attractiveness. Emotion, 19(8), 1490–1494. Linked source
• Krumhuber, E. G., & Manstead, A. S. R. (2009). Can Duchenne smiles be feigned? New evidence on felt and false smiles. Emotion. Linked source
• Porter, S., & ten Brinke, L. (2008). Reading Between the Lies: Identifying Concealed and Falsified Emotions in Universal Facial Expressions. Psychological Science. Linked source
• Said, C. P., Sebe, N., & Todorov, A. (2009). Structural resemblance to emotional expressions predicts evaluation of neutral faces. Emotion. Linked source
• Skiendziel, T., Rösch, A. G., & Schultz, J. (2019). A comprehensive overview of facial action coding system for facial expression recognition. PLOS ONE. Linked source
• U.S. Government Accountability Office. (2013). Aviation Security: TSA Should Limit Future Funding for Behavior Detection Activities, GAO-14-159. Linked source
• Paul Ekman Group. (2026). Facial Action Coding System; Micro Expressions. Linked source Microexpressions source