Home / Insights / AU12 and the Duchenne smile
AU12 and the Duchenne smile: what a smile tells you about an ad
AU12 is the lip corner puller in the Facial Action Coding System: the movement of the zygomaticus major muscle that pulls the corners of the mouth up and back into a smile. On its own it marks a smile of any kind, including a polite or social one. When it appears together with AU6, the cheek raiser that lifts the cheeks and narrows the eyes, the combination is called a Duchenne smile, after the 19th-century French neurologist Guillaume Duchenne. A 1990 study by Paul Ekman, Richard Davidson and Wallace Friesen found Duchenne smiles were more common during pleasant films and tracked reported enjoyment better than other smiles. Later work is more cautious: a 2021 study led by Jeffrey Girard found that eye constriction adds little beyond how strong the smile is. In ad testing, AU12 with AU6 is a useful marker of enjoyment, best read across many viewers and against the ad's timeline.
Updated 17 October 2026 by Jonathan Prescott, Founder, Cavefish.
What AU12 is
The Facial Action Coding System (FACS), published by Paul Ekman and Wallace Friesen in 1978, gives every visible facial movement a number. AU12, the lip corner puller, is produced by the zygomaticus major muscle and pulls the corners of the mouth up and outwards. It is the core movement of every smile.
Automated facial coding detects AU12 more reliably than most other Action Units because it is large, common and easy to see on a webcam. Its intensity is usually scored from 0 (absent) to 1 (maximum). For a wider introduction, see what is FACS?
AU6 and the Duchenne smile
AU6, the cheek raiser, comes from the orbicularis oculi muscle around the eye. It lifts the cheeks, narrows the eyes and creates the crinkles often called crow's feet. A smile with AU6 and AU12 together is called a Duchenne smile; a smile with AU12 alone is sometimes called a social or polite smile.
In 1990 Ekman, Davidson and Friesen recorded facial expressions, brain activity and self-reports while people watched pleasant and unpleasant films. Duchenne smiles occurred more often during the pleasant films and tracked reported positive feelings, while other smiles did not. Read the abstract.
What the research says now
The Duchenne smile is not the clean marker of genuine enjoyment it was once taken to be. A 2021 study by Jeffrey Girard, Jeffrey Cohn and colleagues analysed 751 smiles from 136 people in situations designed to cause amusement, embarrassment, fear and pain. They found that most of the popular hypotheses about the Duchenne smile were contradicted or only weakly supported. Eye constriction did carry some information about positive feeling, but much of it was already given by how strong the smile was, and it depended heavily on context. Read the paper.
The practical lesson for ad testing: stronger smiles with cheek raise are a useful sign of enjoyment, but don't treat AU6 as a lie detector for a single viewer. Read it across a sample and in the context of the scene.
What a smile peak means on an ad curve
On a second-by-second response curve, a rise in AU12 and AU6 across many viewers shows where an ad is enjoyed: the punchline, the reveal, the moment of recognition. Where it falls flat is just as useful. A joke the creative team loves that produces no smiles is worth knowing about before the media buy.
Timing matters too. In a large webcam study of 170 ads, McDuff and colleagues found that strong positive responses immediately after the brand appeared were more likely to go with an increase in purchase intent. Read the paper. So the question is not only whether people smiled, but whether they smiled when the brand was on screen.
A worked example
The demonstration on the EchoDepth homepage shows a viewer who starts neutral, raises her brows in surprise and then breaks into a smile. The viewer is AI-generated, not a real panel member, and the Action Unit values were measured on that clip with an open-source face model (MediaPipe), so the numbers are real measurements of the video.
About three seconds in, AU12 measures 0.94 and AU6 0.29. A strong smile with a moderate cheek raise is the Duchenne pattern. That combination raises both valence and arousal, so this single viewer's Emotional Response Curve climbs sharply at that moment. In a real study the curve is averaged across every viewer in the study. See how the response curve is calculated.
Other Action Units that matter in ads
- AU1 + AU2Inner and outer brow raiser
- Surprise or a jump in attention. Common at reveals and twists.
- AU4Brow lowerer
- Concentration, confusion or displeasure. Often rises during dense product demos.
- AU9, AU10Nose wrinkler, upper lip raiser
- Disgust. Watch for it around food, hygiene and shock tactics.
- AU14Dimpler, often on one side
- Contempt or scepticism. A one-sided AU14 during a claim can mean viewers don't believe it.
- AU15Lip corner depressor
- Sadness. Expected in emotional charity or brand films; a problem elsewhere.
How each of these feeds valence, arousal and dominance is set out in the methodology and explained in valence, arousal and dominance.
Questions
What is AU12?
AU12 is Action Unit 12 in the Facial Action Coding System, the lip corner puller. It is the movement of the zygomaticus major muscle that pulls the corners of the mouth up and back to form a smile.
What do AU6 and AU12 together mean?
AU6 is the cheek raiser, which narrows the eyes and lifts the cheeks. AU6 with AU12 is called a Duchenne smile and is often associated with felt enjoyment, though newer research shows the link is weaker and more context-dependent than once thought.
Is a Duchenne smile always genuine?
No. Research from 2021 found that eye constriction adds little information beyond how strong a smile is, and that its meaning depends on context. It is a useful signal across a sample, not proof of how one person feels.
Does smiling at an ad mean the ad works?
Not on its own. Smiles show enjoyment, which research links to ad liking. Whether that helps the brand depends on timing: positive peaks right after the brand appears have been linked to higher purchase intent.
Why might viewers not smile at a funny ad?
People watching alone show far less expression than people watching together, so even a funny ad may produce small movements. That is why facial coding results are averaged across many viewers rather than read one face at a time.
References
- Ekman, P., & Friesen, W. V. (1978). Facial Action Coding System. Consulting Psychologists Press.
- Ekman, P., Davidson, R. J., & Friesen, W. V. (1990). The Duchenne smile: Emotional expression and brain physiology II. Journal of Personality and Social Psychology, 58(2), 342–353. PubMed 2319446
- Girard, J. M., Cohn, J. F., Yin, L., & Morency, L.-P. (2021). Reconsidering the Duchenne smile: Formalizing and testing hypotheses about eye constriction and positive emotion. Affective Science, 2(1), 32–47. affcom.ku.edu
- McDuff, D., el Kaliouby, R., Cohn, J. F., & Picard, R. W. (2015). Predicting ad liking and purchase intent: Large-scale analysis of facial responses to ads. IEEE Transactions on Affective Computing, 6, 223–235. doi:10.1109/TAFFC.2014.2384198
See it on your own ad
EchoDepth pre-tests video ads on real viewers and reports the response second by second, with every score traced back to the Action Units behind it.