Dog Breed Is Not an Accurate Way To Predict Behavior

A new study that combined genome sequences from over 2,000 dogs with survey data from a further 18,000 pooches has come to a conclusion that upends over a century of established thinking on dog behavior, finding that a dog’s breed contributes little to its behavioral tendencies.

The six-year study was a huge undertaking involving researchers from across the US, led by senior author Elinor Karlsson, director of the vertebrate genomics group at the Broad Institute of MIT and Harvard.

Dogs and behavior

The new study, published in Science, was the culmination of a project that began years ago as an investigation into compulsive dog behaviors, Karlsson told a press conference. Dogs are an excellent model for human behavior – they are closely bonded into human society and are even treated for behavioral disorders with human psychiatric drugs. Karlsson hoped that understanding the reasons why dogs whine, bark and play fetch at a gene level might be informative for understanding the genetics of human behavior too.

Her initial efforts to understand dog genomics were hamstrung by a lack of access to doggy DNA. Karlsson found that whenever she shared the topic of her research with others, there was only one outcome. “They immediately pulled out their cell phone, showed me a picture of their dog and started to tell me everything about their dog's behavior,” she says. Karlsson realized that owners’ enthusiasm was, in reality, the perfect way to access more dog information.

Her team put together a website, called Darwin’s Ark, where owners could (and still can) upload information about their dogs to the site, answering a survey of over 100 questions about their dog’s appearance and behavior.

Karlsson, despite her status as an expert in canine genetics, had never actually owned a dog. She therefore recruited Marjie Alonso, then-executive director of the International Association of Animal Behavior Consultants, to find out more about canine behaviors. Alonso’s insights helped develop the structure of Darwin’s Ark, designing questions that would cut to the core of dog behavior, without getting occluded by the personality and intention that owners often graft onto their dog’s actions.

What’s in a breed?

The study’s initial data presented the associations between the genomic variation among single breed dogs and the responses that owners of these dogs submitted to the survey. The division of domesticated dogs into discrete breeds has become commonplace, but modern breed classifications are under 200 years old. That’s 50-80 generations of dog, versus the thousands of generations packed into the more than 10,000 years of dog domestication. What effect has this relatively short span had on dog genetics? Strong selection of breeds for looks – think massive Great Danes, wiry Airedale Terriers or folded Shar Pei faces – were evidenced in the team’s findings. Physical traits, like size and fur length, showed strong links to different breeds. Some motor patterns also showed links – Alaskan Malamutes were more likely to howl, while Labrador Retrievers were, unsurprisingly, more likely to retrieve.

But the great majority of behavioral traits showed little differentiation between breeds and no behaviors were breed exclusive – while Labradors were rated as being the least likely to howl, 8% of owners said that their Labs did so. To Karlsson, this wasn’t surprising: “The thing about complex traits is that selecting on them takes time. And so, the idea that they'd been created within the last 160 years when these breeds came up, didn't make any sense.”

The team compared their responses to breed standards as defined by the American Kennel Club, who give each breed a three-word character profile (Shiba Inus are active, alert and attentive, Chihuahuas are charming, graceful and sassy) and the Encyclopedia of Dog Breeds. The Encyclopedia proved slightly more in line with the team’s findings. Survey questions were divided into eight “factors”, which captured key aspects of dog behaviors. Nine of ten rankings from the Encyclopedia correlated with at least one factor from the study. But all of these findings had a major flaw – owners’ reports of their pets’ behaviors are likely influenced by the very breed stereotypes that Karlsson and team were trying to outfox. This is where the mutts came in handy.

Mendel’s mutts

Half of the dogs submitted to Darwin’s Ark were mixed-breed mutts. The team took a sample of these mutts who had no easily guessable lineage and sequenced their genomes to work out their exact breed makeup.

This analysis showed some game-changing findings. While owners of purebred Golden Retrievers were more likely to say their dogs was friendly to strangers, actually having Golden Retriever ancestry didn’t make mutts more or less likely to growl at the postman. The team conducted an in-depth genome-wide association study (GWAS) involving these mutts. This analysis showed that, on average, a dog’s breed predicted just 9% of its resulting behavior.

For a few factors, some breed ancestries did have an impact. For example, having Collie ancestry made a dog more likely to be biddable (responding to human commands) while dogs with some Shar Pei in them were immune to the temptation of a thrown toy. But other behavioral factors, which conventional wisdom has decided are linked to breed, showed almost zero link. Breed mattered to an insignificant degree in deciding whether a dog was sociable around other dogs or whether it was easily stimulated across different contexts.

Heritability of dog traits

So, what does influence behavior in dogs? It’s important to remember that the study doesn’t suggest that genetics as a whole is unimportant in determining behaviors. The heritability of a trait is the extent to which genetics influences variation in that trait – the heritability of biddability, for example, was 30%, suggesting that nearly a third of variation among dogs is due to their genes.

What the study does say is that these influences are more likely to have been built up over the thousands of years prior to breed emergence, rather than more recently. To Alonso, the message from the study is clear: “Dogs are individuals.” If you are choosing a dog, breed isn’t totally unimportant – it is probably a bad idea to get a husky if you can only spare 15 minutes a day for walkies – but, breed isn’t the be all and end all. “I don't think that we should really be deciding that breeds are the things that will tell us whether or not we're going to be happy with a dog or a dog is going to be happy with us,” Alonso says.

The study could have implications for legislation around “dangerous breeds” of dog. While the team explained that aggression itself was difficult to measure, their proxy factor of agonistic threshold – essentially how likely a dog was to react negatively to a stimulus – showed almost no link to breed. In respect to genetics, Karlsson says, a breed-based restriction “doesn’t seem to make a lot of sense to us.”

But more widely, these findings could have major implications for how we pick and handle man’s best friend. Rather than expecting to adopt a Labrador and end up with a greedy, friendly family pet or predict a smart, independent and stubborn dog when buying a West Highland Terrier, owners should work more on a dog-by-dog basis. “You’re adopting an individual,” sums up Karlsson, “that you are going to have to learn how to live with.”

Reference: Morrill, K, Hekman, J, McClure, J et al. Ancestry-inclusive dog genomics challenges popular breed stereotypes. Science; 376. doi: 10.1126/science.abk0639