Cerebrum Article

Decoding the Canine Mind

Our author, the Distinguished Professor of Neuroeconomics at Emory University and co-founder of Dog Star Technologies—a company using neuroscience to enhance the dog-human partnership—has put more than 100 dogs through a brain scanner. His article addresses a dog’s perception of the world, social cognition findings, canine mental health, and more.

Published: April 15, 2020

Callie in the MRI scanner while watching pictures of faces.

There is no official census for dogs and cats, but in 2016, the American Veterinary Medical Association estimated that 59 percent of households in the United States had a pet. Although the numbers of dogs and cats remains debatable, dogs continue to gain in popularity with 38 percent of households having at least one. Families with children are even more likely to have a dog (55 percent). With all due respect to cats, dogs have insinuated themselves into human society, forming deep emotional bonds with us and compelling us to feed and shelter them. Worldwide, the dog population is approaching one billion, the majority free-ranging.

Even though many people are convinced they know what their dog is thinking, little is actually known about what is going on in dogs’ heads. This may be surprising because the field of experimental psychology had its birth with Pavlov and his salivating dogs. But as dogs gained traction as household pets, in many cases achieving the status of family members, their use as research subjects fell out of favor. In large part, this was a result of the Animal Welfare Act of 1966, which set standards for the treatment of animals in research and put an end to the practice of stealing pets for experimentation.

How strange it is then that these creatures, whose nearest relatives are wolves, live with us and even share our beds, yet we know almost nothing about what they’re thinking. In the last decade or so, however, the situation has begun to change, and we are in the midst of a renaissance of canine cognitive science. Research labs have sprung up around the world, and dogs participate not as involuntary subjects, but as partners in scientific discovery. This new research is beginning to shed light on what it’s like to be a dog and the nature of the dog-human bond.

Dogs are Special

When scientists use animals in research, they often turn to species that are closely related to humans. “Close” is relative, as even chimpanzees and bonobos diverged from hominids at least 5 million years ago. Monkeys diverged about 25 million years ago, and to find a common ancestor with the dog—indeed with any carnivore—you have to go back 97 million years.

But this summary overlooks the very thing that makes dogs special: their evolution has been altered to make them more socially compatible with us than any other animal. They were, in fact, the first animal to have been domesticated. The million-dollar questions are when and where this happened. We know that dogs existed at the time of the first human settlements in the eastern Mediterranean. In the area known as the Fertile Crescent, their remains have been found buried alongside humans, and these have been dated to 11,000 years ago. Cats, for comparison, did not appear until 8,000 years ago and probably didn’t change into their modern form until 4,000 years later. It is fair to say that only dogs were present at the dawn of human civilization.

The world these early dogs and humans inhabited looked quite different from ours. Even though the last ice age was ending, the climate was still colder than now. This probably brought wolves (an ancestor of the dog) into more frequent contact with humans as the ice sheets retreated. One theory is that wolves and humans helped each other hunt. It seems increasingly likely, though, that the more social wolves began hanging around human settlements to scavenge for leftovers. It is not hard to imagine a curious wolf, probably a juvenile, approaching the edge of a tribe. A human, maybe a child who wouldn’t know any better, might leave some food on the perimeter. And a friendship is born. Eventually wolf-dogs, even if they didn’t hunt, could act as sentries, alerting humans to intruders.

The evolution of cooperation is what allowed humans to dominate the planet, and at the dawn of civilization, we extended our ability to cooperate with each other to another species: dogs. Although there is no fossil record of behavior, there is increasing genetic evidence for this sort of co-evolution. In 2017, a team of researchers found a correlation between sociality in dogs with variants of several genes that had previously been identified in Williams-Beuren syndrome (WBS), a rare genetic disorder in humans. A core feature of WBS is hyper-sociality. When the team evaluated dogs and wolves on tasks that measured sociality, they found two canine genes in the WBS locus that are associated with this hyper-sociality in humans.

These results suggest that the key evolutionary event that turned wolves into dogs was an amplification of genes related to sociality. If that is true, dogs may hold the key to helping humans achieve what can often be a struggle: to be more social, more generous, more loving, more forgiving.

What It’s Like to Be a Dog

So what is going on in a dog’s head? The traditional approach, pioneered by Pavlov, is to measure a dog’s behavior under different circumstances and try to infer why they do what they do. But consider a common example: teaching a dog to fetch. Some dogs, like retrievers, may do this instinctively, but others do not. Is this because the non-performers don’t understand what is being asked of them? Or is it that they understand but would rather do something else? It is all too tempting to project a human explanation onto the dog, to anthropomorphize. The fetch example also highlights an important point: dogs, like people, are individuals. We must be careful in generalizing about dog findings, as there is no such thing as a generic dog. Just like there isn’t a generic human.

Because of the limits of interpreting behavior, my colleagues and I turned to the use of brain imaging to figure out what dogs are really thinking. When we began ten years ago, our approach was different from most animal research. Instead of treating the dogs as research subjects, we treated them as if they were voluntary participants, affording them the same basic rights as human volunteers. We did not use sedation or restraints. Instead, we developed a training program that taught the dogs to walk into a functional magnetic resonance imaging (fMRI) scanner, place their heads in custom-designed chin rests, and lie comfortably while scanning their brains (music video). Since then, we have trained over 100 dogs for fMRI. Many of them have been participating for their entire lives and have become so used to the scanner that it is hard to get them to leave!

Before getting into how a dog’s brain works, it should be understood, if obvious, that dogs do not have the same amount of neural infrastructure that humans do. As a rule, larger animals have larger brains. The encephalization quotient (EQ) accounts for the relationship between brain and body size, such that an EQ=1 means an animal has an average brain size for its body weight. Humans have an exceptionally large EQ of about seven, while dogs are a bit better than your average mammal, with an EQ of 1.2. However, we can see from an MRI of a dog brain that even though it is smaller than a human brain, all of the same basic structures are present. This is true for large regions like the cerebral cortex and the cerebellum, as well as for smaller, subcortical structures like the brainstem, hippocampus, amygdala, and basal ganglia, which have important roles in movement, memory, and emotion.

Dogs also have large olfactory systems, comprising about two percent of the total brain weight (compared to 0.03 percent in humans). Where dogs fall short is in the cortex. Apart from being smaller, there are fewer folds, which means less surface area and fewer neurons. The frontal lobe, which in humans occupies the front third of the brain, is relegated to a paltry ten percent in dogs.

The commonality of brain structures is true across all mammals. While there may be differences at a microscopic level, we all carry around the same basic hardware. Scientists and philosophers continue to debate whether a dog’s experience is the same as a human’s, but the commonality of brain structure suggests a certain commonality in function as well. Dogs have a hippocampus because they have to remember things, too. They have an amygdala because they get aroused and excited and scared, just like we do. They may even suffer similar mental problems (more on that later).

We have discovered many things about dogs’ perceptual experience of the world, but the ones that are most interesting are in the domain of social cognition. The first question many people ask is, “Does my dog love me?” Without getting into the nuances of love, the question gets to the heart of the dog-human relationship, namely, what are a dog’s motives? Is it all about food, or can dogs experience positive emotions for purely social reasons? To answer the question, we used fMRI to measure activity in a structure at the heart of the brain’s reward system: the caudate nucleus.

Before scanning, we trained the dogs on a simple association between toys and rewards (video). Each toy was held in front of the dog for ten seconds and then followed by either a treat or by their owner popping into view and praising them with, “Good dog!” The toy set up a state of expectation, which we could measure in the caudate. We found that 13 of 15 dogs had equal or greater activation for praise than for food. Is that love? We don’t know, but it does show that most dogs have brain systems highly tuned to social rewards, and some even respond more to their owner’s praise than food itself.

How does this social bond form? Humans, like most primates, are born ready to bond with their parents and other members of their social group. Faces carry a wealth of social information and, in the 1990s, neuroscientists discovered that primates have an area of their visual systems dedicated to processing faces, called the fusiform face area. To see if dogs have equivalent areas, we showed pictures and videos to dogs while they were in the MRI scanner. We showed faces (dog and human), objects, scenes, and scrambled images. And just as in humans, we found an area of the dog visual system that is strongly and specifically activated by faces. We called it the “dog face area.” Like the praise experiment, this demonstrates that dogs have more in common with us than we realized, and that they have the basic tools to process human faces.

While we humans identify people by their appearance, dogs may rely on their sense of smell. In an early fMRI study, we presented dogs in the scanner with five scents: their owner, an unfamiliar person, another dog in the house, an unfamiliar dog, and their own scent. Human scents were obtained from underarm wipings and dog scents from the area that dogs like to smell—their butts. Although we expected to find the strongest response to the smell of other dogs, in fact we found that the scent of the owner elicited the greatest activation in the reward system of the dog’s brain. This means that dogs cannot only identify us by smell, they seem to like the smell of their human best (to the extent that reward system activation means they like something.)

Dog brain (left) and human brain (right). Images are not to scale as the dog brain is approximately one-tenth the weight of the human brain. (Dog brain image courtesy of Thomas Fletcher, University of Minnesota.)

Auditory Processing

What about dogs’ ability to understand human speech? Here, we have to be careful in what we mean by “understand.” Dogs seem to understand basic commands like “sit” and, to varying degrees, “come,” but that does not mean that they understand words the way humans do. We use words as symbolic placeholders. We are also very noun-centric. There are roughly ten times as many nouns as verbs, in part, because we label everything. A dog, however, may find actions more salient than names. Humans know that the word “ball” represents a whole class of objects, and its precise meaning derives from how it is used in a sentence. When a dog hears the word “ball,” do they conjure up an image in their mind’s eye like a human would? Maybe “ball” to a dog means the act of retrieving something, or maybe dogs pick up salient information by the tone of our voices when we say the word.

As a first step toward answering these questions, we taught some of the MRI dogs the names of two new toys. To do this, the owner would point to a stuffed animal and say its name, for example, “monkey.” When the dog moved toward it, they would get a treat. Gradually, we removed the pointing. When the dog learned the name of one toy, we then introduced a second. After they learned that, they had to make the correct choice by name when both were present. Before they were deemed ready to scan, a dog had to demonstrate their knowledge by being 80 percent accurate in picking the correct toy on command, much like the famous dog, Chaser, who was reported to know the names of 1,000 toys. With the dog in the scanner, the owners spoke the names of the toys. As a control condition, they also spoke gibberish words that the dogs hadn’t heard before. When this type of experiment is done in humans, real words activate language areas more than fake words, presumably because humans immediately recognize gibberish and stop trying to extract meaning from it. But in the dogs, we found the opposite. The gibberish words caused more activation in auditory areas than the real words. These areas extended beyond what is considered primary auditory cortex, and so we think they represent rudimentary language processing areas.

This tells us two important things. First, dogs can discriminate between words they have heard before and those they haven’t. Second, their reaction to novel words is different from humans’. Instead of immediately recognizing that they have no meaning, dogs pay close attention to novel words, perhaps to figure out what their human is trying to communicate. This response may derive from their hyper-sociality and desire to please. (However, you can be sure that a dog will learn to ignore you if you constantly speak gibberish).

What about complex emotions, like guilt? Although many people believe their dog knows when they have done something wrong, researchers continue to debate whether dogs have the capacity to experience emotions like shame or guilt. Unfortunately, we can’t use fMRI to look for a neural signature of guilt in a dog, in large part, because we haven’t found one in humans. (This may seem surprising, but it has been devilishly hard to find reliable neural markers of human emotional states in general.) However, we have found evidence for something like envy in the dog’s brain. In this experiment, the dog had to watch their owner feed a realistic statue of a dog. As a control condition, the owner placed food in a bucket. We found evidence for amygdala activation, which is a neural marker for arousal, when the fake dog was fed. Although not quite the same as envy, arousal might be a response to envy. This wasn’t universal, though. Only the dogs who displayed aggressive traits toward other dogs had this amygdala response. Again, this highlights the individuality of dogs.

A complementary approach toward decoding emotional states uses machine learning to mine brain data obtained while a person (or dog) watches videos with different types of emotional content. Building on early results of decoding content of visual images from the brain, this new approach suggests a map of emotions in the human visual system, including states like anxiety, awe, fear, disgust, joy, and adoration. Machine learning techniques require a lot more data than conventional fMRI experiments provide—typically hours in the scanner for each subject. This would seem impossible for a dog, but with each visit to the scanner, we have found that the MRI dogs get more and more comfortable with the environment. We have several dogs who are content to lie there, watching whatever content we create for them. Preliminary results suggest that it is possible to decode brain states in some dogs. In the language study, for example, we were able to decode which word was spoken from about half of the dogs’ brains. As we extend this approach to more complex stimuli, we may soon be able to decode emotional states and learn what makes them so hyper-social and lovable.

Dogs and Mental Health

If dogs have evolved to be man’s best friend, is it possible that they also suffer from some of the same mental disorders as people do? Growing evidence suggests the answer is yes, and this is all the more reason to take a closer look at what is going on in dogs’ heads.

Human mental illness is diagnosed largely by symptoms. According to the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM-5), depression is characterized by depressed mood, diminished pleasure, slowed thinking, fatigue, feelings of worthlessness or guilt, and thoughts of death. The only objectively measurable symptom is weight change. Similarly, generalized anxiety disorder is associated with excessive anxiety and worry, restlessness, fatigue, decreased concentration, irritability, muscle aches, and sleep problems.

Dogs, of course, cannot speak, so they can’t report whether they’re feeling sad or anxious. Although neuroimaging may soon change things, we currently have to rely on dogs’ behavior to infer what they are feeling. For example, when dogs are scared, they behave in characteristic ways, which include trembling, hiding in closets or under furniture, chewing or scratching doors to escape, pacing, barking, whining, and defecating or urinating in the house. When these occur in the context of being left alone, they are often labeled “separation anxiety.” Aggression is another frequently misunderstood manifestation of emotional states in dogs. What humans label as aggression may be a normal part of a dog’s behavioral repertoire, which includes barking, growling, and biting. Any dog can bite, and most will do so if provoked sufficiently. However, when they bite, dogs can cause serious injury, especially to children.

Interestingly, dogs with behavioral problems often improve when they are treated with human medications for depression and anxiety. Serotonin and norepinephrine reuptake inhibitors, like fluoxetine (Prozac), are some of the most commonly prescribed drugs in veterinary behavioral medicine. Others include benzodiazepines, tricyclic antidepressants, beta-blockers, and even lithium. Indeed, the psychopharmacopeia for dogs is nearly the same as for humans. The fact that these medications work in dogs speaks to common biological mechanisms of mood regulation. And unlike humans, dogs are not susceptible to placebo effects (although their owners might be, by expecting improved behavior.)

Notwithstanding their emotional quirks, dogs are used in a variety of capacities to help people with disabilities. Service dogs are trained for specific tasks that a person cannot do by themselves, which might include picking up items, opening doors, and alerting to sounds. A psychiatric service dog might be trained to detect the onset of psychiatric episodes, or to turn on lights for someone with post-traumatic stress disorder. In contrast, emotional support dogs are not trained for specific tasks, but used for companionship, to alleviate loneliness, and to aid in the treatment of depression and anxiety.

While service dogs are afforded certain protections under the Americans with Disabilities Act, emotional support animals are not (although they may be covered by other laws, like the Fair Housing Act and Air Carrier Access Act.) Because service dogs often require extensive training, the cost may be prohibitive for many people, up to $50,000. Most dogs are not cut out for this kind of work, so there is a need to identify those that are and not waste resources training those who will not be good service dogs. Brain imaging may play a role here. In a study of 50 dogs-in-training, we were able to predict with 91 percent accuracy whether a dog would or would not graduate service dog training. In particular, we found that amygdala activation was negatively correlated with success, suggesting that dogs that are prone to arousal—either because they are anxious or simply want to play—are not good candidates for service dogs.

It is worth keeping in mind, however, that dogs are not simply treatments to be prescribed for various conditions. Like people, dogs have a wide variety of skills and personalities. And while there are some differences between breeds in any particular personality trait, there seems to be as much variability within a breed. The key to a strong dog-human bond is in the match between dog and human, but this may be as hard to predict as the match between two people. Future research, both with brain imaging and other physiological measures, may soon shed light on the canine side of the equation.

Financial disclosure

This article first appeared in the Spring 2020 issue of our Cerebrum magazine. Click the cover for the full e-magazine.