Angry faces don't faze dogs — their brains recognize smiles but can't tell sad from mad

Among dog owners, it is practically received wisdom that dogs read the room — they creep closer when your voice softens and slink away when your face hardens.

But a look inside the canine brain tells a more nuanced story: dogs clearly recognized a smiling face, yet could not distinguish an angry face from a sad one.

A research team led by Dr. Raul Hernandez-Perez and Dr. Laura Cuaya of the University of Vienna published those findings in the international journal iScience. Researchers from Eötvös Loránd University in Hungary and the National Autonomous University of Mexico (UNAM) also took part.

The team placed unanesthetized pet dogs inside an MRI scanner and showed them photographs of human faces. To ensure the dogs were reacting to expressions rather than to the pleasure of seeing a familiar person, all faces shown belonged to strangers the dogs had never met.

In the first experiment, eight dogs were shown alternating images of smiling and neutral faces while their brain activity was recorded.

Viewing a smiling face triggered increased activity across a broad area of the right hemisphere. No region showed a stronger response to neutral faces.

The research team noted that the brain areas responding to smiles were associated with reward — the same regions that light up when a dog anticipates a treat or catches a familiar scent. The finding suggests that a stranger's smile may be processed in the canine brain as a signal that something good is about to happen.

Regions involved in processing visual information and handling social situations also responded simultaneously, indicating that smiling faces were not processed by a single brain area but engaged multiple regions at once.

In the second experiment, 12 dogs were scanned while being shown faces expressing happiness, sadness, anger and fear.

Using that data, the research team fed signals from the smile-responsive brain regions into a computer and tested whether those signals alone could identify which expression a dog was looking at.

The computer distinguished happiness from sadness with 69.8 percent accuracy, happiness from fear at 63.5 percent, and happiness from anger at 60.9 percent — all above the 50 percent chance level.

When negative expressions were paired against each other, however, the results were no better than chance: sadness versus anger came in at 54.7 percent, sadness versus fear at 50.0 percent, and anger versus fear at 48.4 percent.

The research team interpreted this as evidence that the brain regions responding to faces in dogs are tuned specifically to smiling faces, not to human expressions in general.

That finding alone might suggest dogs simply sort expressions into "positive" and "negative," so the team broadened the analysis to include signals from the entire brain rather than just the smile-responsive areas.

The research team paired the three negative expressions — sadness, anger and fear — into three combinations: sadness and fear, anger and fear, and sadness and anger.

Two of those pairings produced distinct brain responses. Dogs showed different neural activity when viewing sad faces versus fearful faces, and again when viewing angry faces versus fearful faces.

The third pairing — sad versus angry — produced no detectable difference in brain activity anywhere. In short, the two pairings that included fear were distinguishable; the one without fear was not.

The dogs' brains did show some capacity to differentiate negative human expressions, but only when fear was one of the two expressions being compared.

Why fearful faces received such distinct treatment remains unclear. The research team offered several possibilities based on prior studies.

Dogs have been observed to gaze at fearful faces but look away from angry ones. They also showed a stronger stress response to a fear-filled human voice than to a sad one.

Puppies responded noticeably to the scent of human fear but not to the scent of human happiness, suggesting that the ability to detect fear may be innate or may develop early in life.

As for why sadness and anger were indistinguishable, the team pointed to evidence that dogs prioritize body language over facial expressions when reading human emotions, consulting posture and movement first and faces only secondarily.

The research team cautioned against overreading the results, however. The findings do not mean dogs conceptually understand happiness or fear the way humans do, they said, adding that labels such as "happiness," "sadness," "anger" and "fear" are human categories and do not necessarily reflect what a dog actually experiences.

They also noted that differing brain responses do not in themselves prove that dogs actually recognize two expressions as distinct or behave differently in response to them.

The team acknowledged a further limitation: most of the dogs in the study were border collies, a breed known for closely following human cues and maintaining prolonged eye contact with people.

They also acknowledged that the use of still photographs rather than dynamic expressions, and the small sample size of 12 dogs, meant the results could be somewhat inflated.

The dogs that took part in the experiments were neither restrained nor anesthetized and were free to leave the MRI scanner at any time.

DOI: 10.1016/j.isci.2026.116900

Hernández-Pérez R, Concha L, Andics A, Dog brain representations of human facial expressions: Encoding happiness and differentiating specific negative expressions, iScience, 2026.