Researchers led by Raúl Hernández-Pérez and Laura Cuaya, now at the University of Vienna, scanned the brains of twelve pet dogs — mostly border collies, chosen for being unusually good at holding still and reading social cues — while the dogs looked at photos of unfamiliar human faces showing happiness, anger, fear, and sadness. The dogs stayed awake and unrestrained inside the machine the entire time.

Happy faces were never the hard part.

Every dog’s brain lit up in the same specific region — a network stretching into the caudate nucleus, tied to reward — every single time a happy face appeared, clearly separate from any negative expression. That part confirmed what most dog owners already assumed. The real question was whether a dog’s brain treats “anger,” “fear,” and “sadness” as three different things, or just lumps them all into one generic “not happy.”

A whole-brain analysis, using machine learning most of us would need the paper’s supplementary materials to properly understand, found a specific pattern that separated fear from both of the other negative expressions. Anger and sadness, on the other hand, produced patterns close enough that the analysis couldn’t tell them apart. To a dog’s brain, a furious face and a heartbroken one land the same. A terrified one does not.

The likely explanation isn’t complicated. Fear and anger tend to demand something from a dog immediately — faster reactions, quicker heart rate, a decision about whether to get closer or get out. Sadness usually doesn’t ask anything of the dog in the moment, so there was less evolutionary reason for a separate, dedicated brain response to it.

“I would like people to see dogs as emotional beings,” Cuaya said of the findings. “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.” Every dog owner who’s had a bad day and found a dog quietly leaning against their leg already knew the outline of this. The MRI just found where it lives.