Everyone knows a yawn travels. What nobody had pinned down was exactly which muscles move when it lands on your face, or how that pattern differs from the one you produce when someone else laughs, or when you watch someone get hurt.
Researchers in Turin have now measured it. Filming 32 adults as they watched footage of other people laughing, yawning, and expressing pain, the team tracked facial landmarks frame by frame and isolated the specific muscle movements each reaction produced. Yawning spread to the most people. Wincing at someone else's pain spread to the fewest. The results appeared online in February in the journal iScience.
Yawns Traveled Furthest and Winces Traveled Least
The counts are blunt. Yawning caught 23 of the 32 participants, about 72 percent. Laughter caught 20, about 63 percent. Mirror pain, the involuntary grimace people make watching someone else get injured, caught 15, or roughly 47 percent. Fewer than half the room flinched. Twenty-nine participants produced a measurable reaction to at least one category, and the detailed analyses ran on those 29.
That ranking is worth sitting with. Laughter and yawning are the reactions people describe as socially contagious, and the data agree. Mirror pain is the one usually framed as empathy, and it was the least reliably transmitted of the three. The study does not test why, and the authors make no claim about anyone's capacity for compassion. It measures whether a visible facial movement occurred while watching a video.
The Muscle Codes Behind a Borrowed Expression
Facial researchers describe expressions using action units, a numbered inventory of individual muscle movements. The team identified a distinct signature for each reaction.
Contagious laughter ran on AU12, the lip corner puller, combined with AU25, lips parting. Contagious yawning recruited AU9, the nose wrinkler, alongside AU43, eye closure, and AU27, the full mouth stretch. Mirror pain showed up as AU4, the brow lowerer, with AU7, the lid tightener. In plain terms: laughter pulls the mouth wide, yawning wrinkles the nose and squeezes the eyes shut while stretching the jaw, and mirror pain narrows the eyes and pulls the brows down.
The overlap with existing literature is where it gets interesting. The authors' own survey of prior work lists laughter as typically featuring AU12 together with AU6, the cheek raiser, in the Duchenne smile pattern, usually with mouth opening at AU25 or AU26. Yawning is characterized by AU27, along with eye movements including AU43 and AU45, the blink. Mirror pain is messier, presenting upper face movements at AU4, AU7, and AU10 alongside various lower face actions. Several units therefore appear in more than one category, which is precisely why these reactions are hard to tell apart by eye or by software.
Rather than rely on labeling alone, the researchers at the University of Turin and Politecnico di Torino converted the movements into numbers, measuring normalized distances between facial landmarks so the same movement could be compared across differently shaped faces. They validated the approach with electromyography, recording actual muscle electrical activity. The processed data has been deposited publicly on Mendeley Data.
The Lower Half of the Face Did Most of the Talking
The clearest separation between the three reactions, and between all three and a neutral face, came from the lower face. Mouth and jaw geometry did the discriminating work.
That has a practical consequence beyond the lab. Automated expression recognition systems are largely trained on the six so-called basic emotions: happiness, sadness, fear, anger, surprise and disgust. Spontaneous social reactions like these are far more frequent in ordinary interaction and far less well characterized, partly because posed expressions are easy to collect and genuine contagious ones are not. The authors frame their main contribution as a curated set of action units and measurements that other researchers can use to build better detection of unposed expressions.
It also suggests a limitation for any camera-based system that watches only the eyes, or that has to work on a masked face.
A Sample of 32 and a Definition Nobody Agrees On
The study is small. Percentages calculated on a base of 32 move by roughly three points with every single person, so they should be read as description rather than as an estimate of how contagious these behaviors are in the general population.
Participants also viewed the stimuli without any contextual information, and the authors note that the intensity of these responses is modulated by social context. A lab video of a stranger is close to the weakest possible version of the effect. It is also not the only way to study it: yawn contagion documented in toddlers, in separate work by a different Turin group, was observed through everyday nursery interaction rather than screens.
There is a deeper unsettled question underneath. As the authors lay out, researchers do not agree on whether laughing and yawning count as emotional behaviors at all. Some treat them as emotionally laden signals shared across species that serve social communication and group cohesion. Others read them as physiological or neurocognitive events without universally recognizable emotional content, or as basic motor responses lacking intentional expression. Mirror pain is contested the same way, read by some as a cognitive act of perspective-taking and by others as affective resonance and motor simulation. Even the definitions themselves are disputed, with no settled line between emotional contagion and emotional mimicry.
Which means the paper answers a narrow question well. When a yawn or a laugh or a wince crosses from one face to another, here is precisely what moves. What that movement means is still open.
Key Questions Answered
What did the researchers actually do? They filmed 32 adults watching videos of people laughing, yawning, and expressing pain, tracked facial landmarks frame by frame, and identified which muscle movements distinguished each reaction.
Which reaction spread to the most people? Yawning, which produced a visible response in 23 of 32 participants. Laughter reached 20. Mirror pain reached 15.
Does a weak wince mean someone lacks empathy? No. The study recorded whether a facial movement occurred while watching a video. It did not assess empathy, and the authors make no such claim.
What are action units? A numbered system for describing individual facial muscle movements. Laughter here involved AU12 and AU25, yawning involved AU9, AU27, and AU43, and mirror pain involved AU4 and AU7.
Why does the lower face matter? Mouth and jaw movements gave the clearest separation between the three reactions and neutral expressions, which has implications for camera-based systems that focus on the eyes.
Is this settled science? No. Researchers still disagree on whether yawning and laughing qualify as emotional behaviors at all, and on how emotional contagion should be defined against emotional mimicry.
