A dog looking at a human face does not just register happy versus unhappy. According to a new brain-imaging study from the University of Vienna and collaborators, a dog’s brain treats fear as something distinct from anger and sadness, while anger and sadness produce patterns that are nearly indistinguishable from each other. The finding, published August 10 in the open-access journal iScience, is the first neural evidence that dogs differentiate among specific negative human expressions rather than lumping them into one category of bad.
The asymmetry is the surprise. Fear stood apart in the brain’s activity patterns. Anger and sadness did not. That matches what behavioral studies have long suggested, and it hints that for dogs, fear may be the negative emotion most worth detecting in a human face.
Two experiments, fourteen dogs
The team, led by first author Raúl Hernández-Pérez and senior author Laura Cuaya, ran two functional MRI experiments on awake, unrestrained family dogs that had been trained to lie still inside scanners. Experiment 1 used 8 dogs, mostly border collies, scanned at the National Laboratory for Magnetic Resonance Imaging at UNAM in Mexico. The dogs watched unfamiliar faces showing either happy or neutral expressions. Happy faces produced a large cluster of activity, roughly 25 cubic centimeters, in the right temporal cortex extending into the caudate nucleus, a region associated with reward. The authors read this as evidence that a smiling human face carries emotional significance for dogs, not merely visual novelty.
The low-level details were controlled: the happy and neutral images did not differ in brightness, contrast, hue or saturation, so the brain response could not be dismissed as a reaction to a brighter or higher-contrast picture.
Experiment 2 scaled up to 12 dogs across two scanning sites, UNAM in Mexico and Semmelweis University in Budapest, and used a different stimulus set, the Karolinska Directed Emotional Faces, showing strangers expressing happiness, anger, fear and sadness. The dogs had deliberately never seen these people before, so familiarity could not drive the results.
Fear is special
Within the negative emotions, the interesting structure emerged. When the researchers used representational similarity analysis, a method that compares whole-brain activity patterns across the brain rather than in single regions, they found that fearful faces produced a distinct pattern that separated from both angry and sad faces. The differentiation showed up in the right mid ectosylvian gyrus and the left splenial gyrus for anger versus fear, and in the right rostral suprasylvian gyrus for sadness versus fear.
Angry versus sad faces, by contrast, produced no significant clusters at all. The dog brain appears to draw a sharp line around fear, while anger and sadness occupy overlapping territory.
A machine-learning analysis told a related story. Using the happy-face cluster from Experiment 1 as an independent localizer, a classifier could reliably decode happiness against each negative expression, but it could not decode one negative expression from another. The two approaches together, one based on single regions and one on whole-brain patterns, converge on the same conclusion: the canine brain represents the difference between positive and negative valence clearly, and within the negative range, fear carries its own signature while anger and sadness do not.
The authors frame the finding in terms of what the brain is optimized for. From a dog’s perspective, reading a human face may be less about cataloging every emotion and more about detecting threat. Fear is the expression most directly tied to danger in the environment, and it may be the one worth a dedicated neural representation.
What the study does not show
The authors are careful about generalization. The majority of the dogs were border collies, a breed unusually attentive to human social cues, and the paper explicitly cautions against extending the findings to all breeds. The stimuli were static photographs, not the dynamic, multimodal social signals dogs encounter daily. The sample of 12 dogs is small for machine-learning analyses, and although the researchers used strict cross-validation and permutation testing, they acknowledge a residual risk of overfitting.
The study also deliberately used unfamiliar faces, so how familiarity with a specific caregiver modulates these patterns remains untested. And the results do not claim that dogs understand emotions the way humans do, only that their brains represent certain facial expressions differently.
The team is now moving on to a related question: how dog brains respond to a vocalization believed to be the canine equivalent of human laughter. Whether cats would cooperate with such experiments is a separate matter; the researchers have indicated they have no plans to try.
What this study adds to the larger picture of dog cognition is a sharper map of what a dog’s brain actually does with a human face. Dogs can tell happy from neutral, and they can tell fear from anger and sadness, at least at the level of whole-brain activity. That is more resolution than science previously had, and it suggests the dog’s reading of human emotion is not a simple good-bad dial but a more selective instrument, one tuned especially to the signals that matter most for survival.
Sources:
- Hernández-Pérez R, Concha L, Andics A, Bernal-Gamboa R, Cuaya LV, “Dog brain representations of human facial expressions: Encoding happiness and differentiating specific negative expressions,” iScience, 2026, 116900. DOI: 10.1016/j.isci.2026.116900
- Science News, “Sad or afraid? Dogs may be able to read your face” (Aug 10, 2026): https://www.sciencenews.org/article/dogs-understand-emotions-humans-face
- ScienceDaily (Cell Press release, Aug 11, 2026): https://www.sciencedaily.com/releases/2026/08/260811011132.htm
- The Guardian (Aug 10, 2026): https://www.theguardian.com/world/2026/aug/10/dogs-distinguish-fearful-angry-sad-human-faces-brain-scans
- Phys.org (Cell Press release): https://phys.org/news/2026-08-mri-dogs-distinguish-human-sadness.html