Dog brain representations of human facial expressions: Encoding happiness and differentiating specific negative expressions.

Dogs can distinguish human facial expressions, particularly happiness, yet brain processes remain unclear. Using fMRI, we conducted two experiments in awake pet dogs. In Experiment 1 (n = 8), happy faces elicited a stronger response than neutral faces in a right temporal cluster extending to the caudate nucleus, including the rostral Sylvian gyrus. In Experiment 2 (n = 12), dogs viewed faces expressing happiness, anger, fear, or sadness. Using the Experiment 1 cluster as a region of interest, a machine-learning classifier distinguished happiness from each negative facial expression, but not between negative pairs, showing differential BOLD responsiveness to happy faces. Whole-brain representational similarity analyses revealed activity patterns differentiating angry vs. fearful faces (right mid ectosylvian and left splenial gyri) and sad vs. fearful faces (right rostral suprasylvian gyrus) but not angry vs. sad faces. This provides direct evidence that dog brains can distinguish between two negative facial expressions.