What a Crowdsourced Study Taught Us About How Dogs Learn

Thousands of years of selective dog breeding has created a fantastic diversity of domestic canine companions, from the workaholic border collie to the perky Pomeranian. In cultures around the world, humans bred different dogs to be good at tasks including guarding, hunting and herding. Later, in Victorian England, kennel clubs established breed standards related not only to their behavior, but also their appearance.

As genomic sequencing has become more affordable, scientists have begun to understand the genes behind physical features such as body shape and size. But understanding the genes behind dog cognition—the mental processes that underlie dogs’ ability to learn, reason, communicate, remember, and solve problems—is a much trickier and thornier task. Now, in a pair of new studies published in Animal Cognition and in Integrative and Comparative Biology, a team of researchers has begun to quantify just how much variation in dog cognition exists, and to show how much of it has a genetic basis.

To study canine cognition, the studies’ authors turned to publicly available genetic information from a 2017 study, and a large community science project, Dognition.com, in which dog owners tested their own pets. “These papers offer an exciting integration of two forms of big data,” says Jeff Stevens, a psychologist at the University of Nebraska-Lincoln who was not involved in the study.

Previous studies often compared cognition in one breed against another using small sample sizes of dogs from each. This study, by contrast, is the first to examine the variation in cognition across three dozen breeds, and the genetic basis of that variation, explains Evan MacLean, a comparative psychologist at the University of Arizona who oversaw the pair of new studies. MacLean says dog breeds may be an ideal way to study the heritability of cognitive traits because breeds—all part of the same species—represent close genetic relatives with an incredibly diverse range of appearances and behaviors.

To gather a sufficient amount of data on how dogs reason and solve problems, the researchers looked to the Dognition.com portal. The initiative, created by Duke University dog researcher Brian Hare, started with tests in the lab. Researchers developed methods to understand how dogs think. They then stripped those methods down, and simplified them for dog owners to do themselves. In an earlier project, the researchers tested dogs in the lab and compared their results to those from owners testing the same dog at home. The results were the same, giving them confidence that the results from the citizen science project were reliable.

To participate in this project, dog owners tested their pups on 11 standardized tasks used by animal behaviorists on a variety of species that reflect four aspects of cognition: inhibitory control, communication, memory and physical reasoning. One task that measured inhibitory control, for example, involved having an owner put a treat on the floor in front of the dog and then verbally forbidding the dog from taking it. The owner then measured how long the dog would wait before eating the treat. In a task to assess communication skills, the dog owner placed two treats on the ground and gestured towards one of them. The owner then determined if the dog approached the indicated treat. To assess memory, the owner visibly placed food under one of two cups, waited for a few minutes, and then determined if the dog remembered which cup the food was placed under. To test physical reasoning, the owner hid food under one of two cups, out of view of the dog. The owner lifted the empty cup to show the dog that there was no food and then assessed whether the dog approached the cup with the food underneath.

The participating dog owners reported their dog’s scores and breed, producing a dataset with 1,508 dogs across 36 breeds. The researchers analyzed the scores and found that about 70 percent of the variance in inhibitory control was heritable, or attributable to genes. Communication was about 50 percent heritable, while memory and physical reasoning were about 20 percent heritable.

“What’s so cool about that is these two traits that are highly heritable [control and communication] are those that are thought to be linked to dogs’ domestication process,” says Zachary Silver, a graduate student in the Canine Cognition Center at Yale who was not involved in the study.

Dogs are better at following humans’ communicative cues than wolves, and this is something that seems to be highly heritable, explains Silver. In contrast, there’s some evidence that wolves are better than dogs at physical reasoning.

Some of these traits are also influenced by environment and how the dog was handled as a puppy, so there are both genetic and environmental components. In fact, there is so much environmental and experiential influence on these traits that Gitanjali Gnanadesikan, a graduate student in MacLean’s lab and lead author of the new studies, cautions against the idea that these findings support certain breed restrictions or stereotypes. “Even the highly heritable traits have a lot of room for environmental influence,” she says. “This shouldn’t be interpreted as, ‘each of these breeds is just the way they are, and there’s nothing that can be done about it.’”

In the same way that women are on average shorter than men, but there’s quite a lot of overlapping variation within each sex, dog breeds also show a lot of variation within each breed that overlaps with variation among breeds.

Previous work has linked differences in inhibitory control to the estimated size of dogs’ brains. Comparative studies across many different species, ranging from tiny rodents to elephants and chimpanzees, also show that some aspects of self-control are strongly related to brain size. The bigger the brain size, the more self-control the animals seem to have, MacLean says.

Stevens notes that a lot of things—not just inhibitory control—correlate with brain size across species. And brain size, metabolic rate, lifespan, home range size are all correlated with body size. When many traits are correlated with each other, it is not clear which of these factors may underlie the cognitive differences. So there are a number of questions remaining to be explored.

After showing the degree to which different aspects of dog cognition are heritable, Gnanadesikan and MacLean used publicly available information on the genomes of dog breeds to search for genetic variation that was associated with the cognitive traits of interest. The researchers found that, like many other complex traits, there were many genes, each with small effect, that contribute to dogs’ cognitive traits. That is in contrast to morphological features in dogs; about 50 percent of variation in dog body size can be accounted for by variation in a single gene.

One of the limitations of the study is that the researchers did not have cognitive and genetic information from the same dogs; the genomes were breed averages. In the future, the researchers are planning to collect genetic data from the very same dogs that are completing the cognitive tests, to get measures of cognitive and genetic variation at the level of individual dogs. “This gives us a roadmap for places that we might want to look at more carefully in the future,” MacLean explains.