Brain scans reveal what dogs see when they look at a human face

A frightened human face does not look the same to a dog’s brain as an angry or sad one. Brain scans now suggest canine responses to our expressions carry more detail than a simple division between good and bad.

Researchers scanned awake pet dogs as they viewed photographs of unfamiliar people showing happiness, anger, fear, sadness, and neutral expressions.

The experiments revealed a particularly strong response to smiling faces in a right-sided brain network involving the temporal cortex and caudate nucleus.

A broader analysis then found something less expected. Fear produced patterns of brain activity that differed from both anger and sadness.

“Humans are really good at picking up small details on the face, and so are dogs,” said first author Raúl Hernández-Pérez of the University of Vienna. “This makes dogs very interesting. By studying their brains, we can understand what adaptations have developed to support their remarkable social and emotional understanding of humans.”

The researchers began with eight dogs and compared their brain activity while they viewed happy and neutral human faces.

Faces in general activated visual and temporal areas. Happy faces, however, produced a stronger response in a large region of the right temporal cortex that extended into the caudate nucleus.

The caudate is associated with reward processing, making its involvement especially notable.

“The activity in the caudate nucleus is particularly interesting,” said senior author Laura Cuaya of the University of Vienna. “It suggests that happy human faces may have emotional significance for dogs.”

The team then asked whether the same region simply responded to emotional intensity or specifically to happiness.

A second experiment included 12 dogs. The animals viewed human faces expressing happiness, anger, fear, and sadness while researchers recorded their brain activity.

Machine-learning analysis could distinguish happiness from each of the three negative expressions using activity from the happiness-sensitive brain region.

Average classification performance reached 0.698 for happiness versus sadness, 0.609 for happiness versus anger, and 0.635 for happiness versus fear.

When researchers compared the negative expressions within that same brain region, performance dropped close to chance.

Sadness versus anger produced an average of 0.547. Sadness versus fear reached 0.500, while anger versus fear reached 0.484.

That result initially left open a simpler explanation. Dogs might process smiles differently while grouping all negative expressions together.

The researchers therefore expanded their analysis across the entire brain.

That wider comparison revealed distinct activity patterns separating fear from sadness and fear from anger.

Fear and sadness differed in a region called the right rostral suprasylvian gyrus. Fear and anger produced an even larger difference involving the right mid ectosylvian gyrus and part of the left splenial gyrus.

No significant pattern separated sadness from anger.

The results therefore do not indicate that dogs distinguish every negative facial expression equally well. Fear stood out most clearly from the other two.

The researchers stressed that different brain responses do not prove that dogs experience or interpret emotions exactly as people do.

Words such as happiness, anger, fear, and sadness describe the human facial expressions shown in the photographs. The scans cannot establish what meaning individual dogs assigned to those faces.

Nor do different neural patterns necessarily mean every dog could behaviorally identify each emotion.

Still, the findings indicate that canine brains contain information beyond whether a face appears positive or negative.

A tendency toward right-sided processing also appeared across the experiments.

The strongest happy-face response occurred in the right hemisphere. Most regions that distinguished negative expressions were also on the right, although part of the anger-versus-fear response extended into the left hemisphere.

The team cautioned against treating that pattern as definitive. Some happy-face activity also appeared on the left but did not reach the statistical threshold.

Dogs offer researchers an unusual comparison with humans because their close social relationship with people did not arise from a recent shared primate ancestor.

Instead, domestication brought dogs into repeated interaction and cooperation with humans.

“When we compare humans with primates, we are looking at abilities that may have come from a common ancestor,” Cuaya said. “With dogs, the story is totally different. They followed a very different evolutionary path, yet through domestication, they developed the social abilities needed to understand and cooperate with us.”

The dogs viewed unfamiliar people rather than their owners. That reduced the chance that familiarity with a particular person influenced the brain responses.

The canine participants were also family pets rather than free-ranging dogs.

Dogs living on streets or in other environments experience different relationships with humans and may respond differently to human social signals.

The research involved only eight dogs in the first experiment and 12 in the second.

Most were border collies, a breed previous research has associated with strong attention to human social cues. The findings therefore may not generalize equally to every breed.

The small sample also creates a risk of overfitting in machine-learning analysis.

Researchers tried to limit that problem by collecting 10 functional scanning runs from each dog, using cross-validation, and performing 100,000 statistical permutations. They could not eliminate the possibility entirely.

Another limitation comes from the faces themselves.

Dogs saw static two-dimensional photographs. Real interactions include changing facial expressions, body movements, vocal signals, and scent.

Future work could combine those different signals and compare how dog and human brains respond to the same emotional information.

The findings give researchers a more precise way to investigate how dogs process human social information.

Rather than asking only whether dogs detect positive and negative emotion, future experiments can examine which expressions generate distinct patterns and how those signals combine with voices, movement, and scent.

Broader samples could also test whether breed, upbringing, or living conditions affect canine responses.

Understanding those differences may help clarify how domestication shaped dogs’ ability to live closely with humans.

“I would like people to see dogs as emotional beings,” Cuaya said. “They have evolved with us to understand our expressions and cooperate with us, and recognizing that can change how we communicate with and care for them.”

These resources explore how dogs attend to emotional faces, use human social information and process facial and bodily signals across species.

Species-dependent gazing behavior of emotional facial expressions in dogs (Canis familiaris) and adult humans (Homo sapiens): Researchers compared how dogs and people visually inspected emotional dog and human faces. Dogs concentrated heavily on the eye and nose regions, while gaze patterns in both species changed according to the expression and species being viewed. (Journal of Comparative Psychology, 2026)

Domestic dogs as a comparative model for social neuroscience: Advances and challenges: This review examines dog neuroimaging research on faces, bodies, emotion perception and the dog-human relationship. It identifies the temporal cortex as an important part of canine visual social cognition while highlighting statistical power and methodological challenges in dog fMRI research. (Neuroscience & Biobehavioral Reviews, 2024)

Visual perception of emotion cues in dogs: a critical review of methodologies: This review assesses evidence that dogs respond to visual emotional cues while examining limitations such as artificial stimuli, anthropomorphic assumptions and an emphasis on faces rather than whole-body signals. It provides useful context for interpreting experiments based on static photographs. (Animal Cognition, 2023)

Dogs functionally respond to and use emotional information from human expressions: The authors review evidence that dogs do more than notice human emotional expressions. Dogs can use information from faces and body posture when making decisions and solving problems, connecting emotion perception with their everyday social behavior around people. (Evolutionary Human Sciences, 2023)

The neurobiological basis of emotions and their connection to facial expressions in non-human mammals: insights in nonverbal communication: This review examines the neurobiology of emotion and facial expression across nonhuman mammals, including dogs, and discusses methods such as AnimalFACS for objectively describing facial movements. It places canine emotional communication within the wider study of mammalian social signaling. (Frontiers in Veterinary Science, 2025)

Research findings are available online in the journal iScience.

The original story "Brain scans reveal what dogs see when they look at a human face" is published in The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.