
Do dogs understand the words we say, or only the tone we use? A 2016 brain-imaging study from Budapest suggests both matter: dogs appear to recognise familiar praise words regardless of tone, but the brain’s reward system responded most strongly when the right words came with a warm, happy voice. It does not show that dogs understand sentences the way people do.
The study also became a cautionary tale. A year after publication its authors corrected a mix-up of left and right brain hemispheres, which changed one of the headline claims. Here is what was done, what was found, and what it means for everyday life with a dog.
How do you scan an awake dog’s brain?
The research came from the Family Dog Project at Eötvös Loránd University in Budapest, led by Attila Andics, and was published in the journal Science in August 2016. The team used functional magnetic resonance imaging (fMRI), which tracks changes in blood flow to show which brain areas become active.
Thirteen family dogs, many of them border collies and golden retrievers, were trained with patience and rewards to lie completely still in the scanner while awake. They wore ear protection, were not restrained, and could leave whenever they wanted. That alone was a notable achievement; most animal brain imaging requires sedation.

Researcher Anna Gábor with Barack, one of the dogs trained to rest motionless in the scanner. Photo: Vilja and Vanda Molnár
The experiment: praise words, neutral words, two tones
A female trainer familiar to the dogs recorded Hungarian words in four combinations. Praise phrases, the equivalent of “good boy” or “well done,” were spoken once in an enthusiastic, praising voice and once flatly. Neutral filler words, the equivalent of “however” or “as if,” were spoken in the same two ways.
Mixing the words and the tone this way let the researchers separate two questions that are normally tangled together at home: is the dog reacting to what was said, or to how it was said?

Some of the volunteer dogs from the Budapest study, which included several border collies and golden retrievers. Photo: Enikő Kubinyi
What the scans showed
Three results stood out.
- Familiar words were picked out regardless of tone. Praise words produced a distinct response compared with the meaningless filler words, whether they were said warmly or flatly.
- Tone was processed separately. A region in the auditory cortex responded to the emotional quality of the voice, independently of the word.
- Reward needed both. The brain’s reward centre, the area that also reacts to food or petting, lit up only when praise words came with a praising tone. Warm tone with nonsense words, or praise said flatly, did not have the same effect.
Andics summarised the practical point for ScienceAlert: praise can work as a reward, but “it works best if both words and intonation match.”

A golden retriever resting on the scanner bed between sessions. Photo: Enikő Kubinyi
The left-right correction, explained
The original paper said dogs processed word meaning mainly in the left hemisphere and tone in the right, the same split seen in most people. That neat parallel with humans became the headline in much of the coverage.
In April 2017 Science published an erratum. The authors had flipped left and right in all of the dogs’ images while adjusting for the different way dogs and humans lie in the scanner. Corrected, the word effect sat in the right hemisphere and the tone effect in the left: the opposite of the typical human pattern.
The authors said their main conclusion still held, because dogs still handled meaning and tone in different hemispheres and combined them for reward. Not everyone agreed it was a minor change; a columnist at Discover argued that the human-like left-side bias had been a central claim. Both views are fair. What survives is the separation of words and tone, and the need for both to trigger reward. What does not survive is “dogs use the same side of the brain for language that we do.”
Common myths about dogs and language
“Dogs understand everything we say.” The study used a handful of very familiar phrases. Andics himself stressed to reporters that the team was not claiming dogs understand everything people say. Recognising a familiar sound pattern is not the same as following grammar or abstract ideas.
“Only tone matters.” This is the myth the study challenged most directly. Dogs did distinguish familiar words from meaningless ones even when the tone was neutral.
“One study settles it.” Thirteen highly trained dogs is a small sample, and the researchers called for larger studies. Later work from the same group and others has continued to explore how dogs learn object names, with large differences between individual dogs.
What dog owners can take from this
Use consistent words. A dog learns “good girl” faster if it is always “good girl,” not a rotating set of phrases. Match words and voice: if you want praise to feel like a reward, say it as though you mean it. And remember that body language, timing and treats remain powerful signals. Most trainers find that dogs read gestures at least as closely as speech.
Andics suggested that other animals living closely with people, such as cats or horses, might show similar abilities, though that idea has not been tested in the same way. For now, the Budapest study offers a fair conclusion: your dog may not follow your sentences, but it is listening to both your words and your voice.
Sources: ScienceAlert (2016); Science News: Dog brains divide language tasks; Discover: Dogs don’t process language with their left brains, after all; Science (2017) erratum; The Washington Post (2016); HuffPost (original report).

COMMENTS