Reading Your Dog’s Emotion Depends on Whether You Are Looking at Their Face or Their Body

When you look at your dog and try to read how they feel, you are doing something more complex than you probably realise — and whether you are looking at their face or their whole body changes not just how accurately you read them but which emotions you misread them as. A study of 447 participants reading dog emotional expressions has produced the most detailed comparison yet of facial versus bodily expression channels in canine emotion recognition. The findings are specific, counterintuitive in places, and practically significant. 🐾

Researchers Kun Guo, Catia Correia-Caeiro, and Daniel Mills used dynamic naturalistic video clips depicting dogs in eleven emotion categories — six primary emotions (anger, disgust, fear, happiness, sadness, surprise) and five secondary emotions (appeasement, frustration, pain, positive anticipation, separation distress). Participants either viewed videos showing the dog’s face only or the full body including the face, allowing direct comparison of facial versus bodily expression channel accuracy across all eleven emotion types. With 447 participants and a naturalistic video design that goes beyond the static photographs most previous studies relied on, this is one of the most comprehensive dog emotion recognition studies conducted.

The Headline Finding and Why It Is Not the Whole Story

When averaged across all eleven emotion categories, bodily expression — the full face-and-body condition — produced higher categorisation accuracy than facial expression alone. This overall advantage for bodily cues aligns with what the pose estimation AI research covered in this series found: whole-body posture carries more diagnostic information than facial expression alone for classifying dog emotional states.

But the most scientifically interesting findings are in the category-specific breakdowns that the average conceals. Anger was recognised more accurately from facial expressions than from body posture. Surprise was similarly better recognised facially. Happiness and fear showed the opposite pattern — bodily expressions produced better recognition accuracy for both. The face and the body are not simply redundant channels carrying the same signal at different amplitudes. They are genuinely different channels whose diagnostic value varies by the specific emotion being expressed.

The misattribution patterns are equally revealing. Fear facial expressions were frequently mistaken for happiness — the wide-eyed, lip-retracted expressions that fear can produce overlapping visually with what observers associate with a happy dog face. Fear bodily expressions were more often mistaken for sadness — the lowered body and tucked tail of a fearful dog reading as depressed rather than scared. An owner who looks only at the face may tell themselves their frightened dog looks happy. An owner who reads the body may tell themselves the same dog looks sad. Both are wrong in different ways.

What Owner Experience Actually Improved ⚠️

The impact of dog ownership experience on emotion recognition was channel-dependent and emotion-dependent rather than producing general improvement across all conditions. Prolonged experience with dogs tended to improve recognition of fear facial expressions and appeasement bodily expressions specifically. It did not produce broad improvements across all emotion-channel combinations.

This is consistent with the findings from earlier research in this series on owner emotion perception. Experience builds familiarity with specific patterns in specific contexts — but it does not automatically produce the kind of comprehensive signal literacy that covers all emotion types across all expression channels. An owner with twenty years of dog ownership may read their own dog’s fear face accurately while still systematically misreading fear body posture as sadness in an unfamiliar dog.

The secondary emotion categories in this study — appeasement, frustration, pain, positive anticipation, separation distress — are particularly relevant for everyday dog management. These are the emotional states that owners most frequently encounter and most frequently misinterpret, and the study’s inclusion of them alongside primary emotions makes it more ecologically valid than research limited to basic emotion categories. Pain recognition deserves special attention: a dog in pain whose bodily expression is misread as sadness or lethargy may not receive the veterinary attention their signal warrants.

What This Means for Reading the Dog in Front of You 🐕

The practical implication from this research is not simply to watch the whole body rather than just the face — though that is part of it. It is to develop an understanding of which expression channel is most diagnostic for which emotion, and to cross-reference both channels rather than defaulting to whichever is more salient in the moment. A dog whose face and body are giving different signals — a relaxed facial expression with a tense, lowered body — is a dog whose body may be carrying the more accurate emotional information for some states and whose face is carrying it for others.

At Zoeta Dogsoul, this research adds further precision to what NeuroBond asks of owners. Attunement to the real dog means developing signal literacy across expression channels — not just the obvious vocalisations and gross body movements that owners naturally attend to, but the channel-specific diagnostic cues that carry the clearest signal for each specific emotional state. The Invisible Leash is only as accurate as the reading that informs it — and reading more accurately means knowing which channel to look at for which emotion, not just looking harder at the whole picture. 🐾

Source: Guo, K., Correia-Caeiro, C., & Mills, D. S. (2024). Category-dependent contribution of dog facial and bodily cues in human perception of dog emotions. Applied Animal Behaviour Science, 280, 106427. https://doi.org/10.1016/j.applanim.2024.106427