Dogs Know When You’re Praising Them. That Doesn’t Mean They Understand Human Speech

A dose of caution with the results of an intriguing new study

The only one who really understands me. TatyanaGl / iStock

Being man’s best friend is no walk in the park. In addition to providing unconditional love, your dog must play exercise companion, get along with other domesticated pets, and even become your therapist by listening to your ceaseless one-way chatter. But what’s really going on in a dog’s head when you're talking their ear off?

A team of Hungarian neuroscientists has shown that dogs process certain aspects of speech similar to the way humans do, with specialized brain areas dedicated to discriminating both the meaning of words (what we say) as well as the tone in which they are delivered (how we say it). This shared neural circuitry between two species, the authors suggest, is evidence that the development of language in humans might have been a learned invention and not the product of unique changes in our brains. 

“When you use the command ‘sit’ you always use it in the same context. You are physically present, you look at the dog, and use a particular intonation—all while the dog is probably standing,” says Attila Andics, a neuroscientist at Eötvös Loránd University in Budapest, Hungary and lead author of the study. “But would it still work if you sang the command, if you weren’t there and said it via phone, or if the dog was lying down? We were really interested whether it was possible for dogs to actually take out the word meaning information from this—to separately process word meaning.”

In the study, Andics and his team brought 13 pet dogs to their laboratory, and trained them to remain still in an fMRI scanner—a claustrophobic space for even the most well-behaved canine. While their brain activity was monitored, the dogs listened to an audio recording with a variety of familiar praises (e.g. “well done,” “clever,” and “that's it”) as well as neutral conjunctions that are commonly used in everyday speech but had no relevant implications for the dogs  (e.g. “as if,” “such,” “yet”). To determine whether the dogs could discriminate between the “what” of the word versus how we say it, each word was spoken in either a praising intonation—characterized by a higher, more varying pitch (i.e. how you might talk to a cute toddler)—or a neutral intonation.

The results, published this week in Science, revealed that regardless of tone, speaking words of praise to pooches elicits a significant activation of the brain's left hemisphere. Past studies have shown a similar pattern of left hemispheric dominance in human speech processing and consider it to be a neurological consequence of the evolution of language. The authors of this latest study argued that this similarity could have implications for both what dogs understand and how humans process language—though scientists and science writers have cautioned not to take these conclusions too far.

The team also found that regardless of the word meaning, when the dog listened to neutral tones, there was increased activity in a region of the right hemisphere in the pup’s brain. According to the authors, this region is specifically designed to process generic acoustic information. A subsequent analysis showed that the lower frequencies characteristic of neutral tones elicited the greatest responses—a result that also mirrors auditory processing in humans, macaques and previous dog studies.

The combination of praising words and tones, however, activated very different parts of the brain. Andics and his team found that regions typically associated with reward in humans showed significant increases in activity only when praising words were delivered in a praising tone. This means that the dogs exhibited an ability to isolate and separately process word meaning, but they were then able to integrate both in the reward centers of their brain.

It’s easy to conclude that, because dogs process verbal and nonverbal information in speech similar to the way we do, they likewise understand what they are hearing. But this would be a vast oversimplification. The study “shows that dogs are capable of identifying strings of phonemes that form meaningful speech commands, rather than solely relying on the command's intonation,” says David Reby, a psychologist at the University of Sussex whose previous behavioral studies in dogs also identified parallels in speech perception between humans and dogs. “It does not, however mean that dogs are capable of understanding human language.”

Terrence Deacon, a professor of biological anthropology at University of California, Berkeley, agrees: “The fact that a dog hears a sound and responds to it preferentially with left hemisphere activation is not a surprise to me,” he says. “But our brains are handling language in a way that is radically different than this dog is handling the sound of words. Dogs don’t have an elaborate semantic network in their heads—interpreting a word as being in relation to a whole system of other words—like we do. That we can understand a complex sentence or any number of sentences is a uniquely human trait that is not being demonstrated here.”

Dogs and MRI
The trained dogs sit around the MR scanner. Image from Enikő Kubinyi

Additionally, some experts have criticized the study’s methodology. “The Hungarian paper is wildly overinterpreted,” says Gregory Berns, a neuroscientist at Emory University who has conducted numerous fMRI studies on dogs. “Because of the small sample size, one dog could easily sway the results to suggest a hemisphere bias. And even if there is a bias, it is small and there are other more likely explanations than concluding something about language processing.”

For example, because the researchers didn’t test the dogs for left hemisphere activity when they weren’t listening to words, there is no way of knowing whether the elevated activity is due to word processing, or a generally higher activation level. “We see lateralization all the time in our dog studies, as well as human studies, but we don’t usually ascribe any functional significance to it,” Berns says.

And even if the hemispheric bias is a real phenomenon, Berns says it's impossible to determine whether it is a consequence of generalized word processing or because the words that they used to test the dogs were praise words. “Human imaging studies have long shown that the left hemisphere tends to be more active to positive emotions. This could explain the Hungarian results, without invoking lexical processing,” he says.

But if the analyses are accurate, what do they say about the initial development of language in humans?

Many prevailing theories of language evolution hypothesize that a significant change in the human brain enabled mankind to start using and understanding words. However, given that the hallmark left hemisphere bias in the processing speech is not uniquely human, Andics postulates that the use of words was a human invention that arose independent of any significant brain changes.

“When humans invented the wheel, we were able to do many things we didn’t do before,” he says. “But I don’t think we gained some neural mechanism that allowed us to invent the wheel, and I think it’s the same here. The invention of words was a great cultural invention of humans, but it had nothing to do with brain changes.”

Yet evolutionary anthropologists and other experts maintain that this interpretation glosses over the complexity of human language processing. “The human brain utilizes systems that have been there for a long time,” says Deacon. “It utilizes them in different ways perhaps, but it there is not a total reorganization of the brain ... It’s not that language is a cultural invention and that brains just do this, it’s that language has recruited these systems that were originally there to do other things.”

Others don't agree that specialized speech processing was an original feature in dogs, asserting that they may have evolved this ability as they became domesticated.

“Because dogs are domesticated animals, this striking parallel may reflect the effect of artificial selection on their perceptual capabilities," says Reby. "In other words, over thousands of years, we may have selected them to efficiently process our spoken commands, and this parallel may represent a case of convergent evolution rather than an ancestral latent ability. Testing tamed—but not domesticated—animals, such as wolves should provide an answer to this fascinating question.”

Moving forward, Andics and his team will continue to explore various aspects of speech processing in dogs: how they differentiate between speakers, how they learn new words, and even how they might process a combination of words in syntax.

“Dogs are a great model for comparative studies on language processing because they care about language and care about human speech, but at the same time are distantly related to humans,” says Andics. “We believe that this is a very first step in a very interesting new direction for research.”

Get the latest Science stories in your inbox.