TANSY WAS NOT into sports. The little border collie, a rescue, didn’t care for agility trials or flyball. But her adopted family—with two other border collies already in the house—played them all the time.
Border collies, the elite athletes of the canine universe, are working dogs. They go a little nuts without something to do. After a little consternation, Tansy's new owner Robin Queen, a linguist at the University of Michigan, got some advice: sheep. And why not? Border collies are, after all, sheepdogs. As soon as Tansy caught sight of some livestock, “it was the first time she showed evidence of understanding something about the world,” Queen says.
That's how Tansy got into competitive sheepdog trials, a sport in which a handler and dog manage a half-dozen sheep through various tasks. Despite their name, sheep are not sheepish and often act on their own closely held ideas about where to go. Keeping a flock on track can require dogged persistence. It's difficult and takes a lot of practice. “We were a little bit unusual in that we had very little dog experience and certainly no livestock experience,” Queen says. “People like us don’t tend to stick it out for very long because it’s hard, and you don’t get a lot of fuzzies very fast. It’s hard to control a dog around sheep.”
To exercise that control, sheepdog handlers typically use a specialized whistle. Yes, literally a dog whistle. Dogs might get up to half a mile away, so you need something loud but with finesse. With a whistle, handlers deploy a small lexicon of commands. Two medium blasts, for example, means "walk toward the sheep." A single low note means "go clockwise around the herd."
Queen later competed with her dog Hamish and another pup, Ky. That was a decade ago. Then Tansy went to the big meadow in the sky and Queen got Zac, with a plan to elevate their game together. About then, Queen started to notice something. In talking to other handlers and listening not just to the lexicon of commands they used but how, and how the dogs responded, she realized: These aren’t just orders. In fact, those whistles sounded a whole lot like a language.
Ten years after Queen started competitive sheepdog trials, she at last felt ready to turn what had initially been an escape from academia into a scholarly pursuit. Her hobby had become a research project. She presented these ideas at a linguistics conference in July, where another linguist named Gretchen McCulloch livetweeted them. (A caveat—this work isn't peer reviewed or formally published yet, so take it, for now, as tantalizing rather than definitive.)
People have been trying to parse how dogs and people communicate with each other for a long time. Obviously they do—but hypothetically the form and content go way beyond sit and stay—and say something broader about language and animal cognition.
Border collies in particular have been central to this research. In the early 2000s a team of researchers at the Max-Planck Institute for Evolutionary Anthropology in Germany heard about a border collie named Rico whose owners said he knew the names of 200 different objects, mostly toys. So the researchers, led by an evolutionary psychologist named Juliane Kaminski, went to see Rico and tested him.
Not only could Rico, under experimental conditions, retrieve specified objects with no clues other than hearing their name, but he could also infer the name of a third, unfamiliar object when presented with it alongside two of his toys. And he still remembered the name of the new object when they tested Rico a month later. "Apparently Rico's extensive experience with acquiring the names of objects allowed him to establish the rule that things can have names," Kaminski's team wrote. "Consequently, he was able to deduce the referent of a new word on the basis of the principle of exclusion." The dog was, it seemed, performing what developmental psychologists call “fast-mapping,” or figuring out the names of new things with the speed and acumen of a 3-year-old human. (Rico: Smarter than your toddler. Would meaningfully communicate with again. 13/10.)
After reading the Kaminski team’s work, South Carolina psychologists John Pilley and Alliston Reid acquired an 8-month-old border collie named Chaser (awwww) and, through intense training over the next three years, taught her the names of more than 1,000 different balls, plush toys, frisbees, and other tchotchkes. Not only that, but Chaser also learned commands that she could recombine with the nouns into sentences she’d never heard—like “nose the lamb” or “paw the ABC” (a cloth alphabet block). And she could learn names by exclusion and remember them, just like Rico.
Rico and Chaser are heroes in the animal cognition community. It’s not that they learned to associate relatively arbitrary human sounds with specific objects. Lots of animals have done that—grey parrots, chimpanzees, sea lions, even pet dogs. Big whoop (or squawk, or eep, or grunt). It’s that these dogs, and others in later research, were able to do much more. They could fast-map and learn the names of things through reference cues such as pointing. Even chimps and bonobos, humans’ closest genetic relatives, can’t do that. “What’s unique about humans is that we cooperate and share information,” says Evan MacLean, a biological anthropologist at the University of Arizona. “We have a joint frame of reference.”
So, not unique at all, because apparently we share that frame of reference with puppies. Canine social structures are much more like human social structures than chimpanzee social structures—maybe because humans and dogs essentially grew up together, evolutionarily speaking. “Dogs are much more like young children,” MacLean says. “If the dog is looking for something and you provide cooperative information, the dog uses this information.” Cooperation leading to the development of language is one of the leading hypotheses for what makes human beings unique among the animals. Except not so unique, because dogs can cooperate with us.
Good dogs. Would go back 15,000 years and domesticate again. 11/10.
About the same time as animal cognition researchers and evolutionary anthropologists were getting spun up about Rico, linguists weren’t as encouraging to Queen. She had tenure at Michigan, but colleagues still tried to wave off the idea that dog handlers had language-like communications with dogs. Queen was still thinking of it as an avocation only.
But independently she had noticed the cooperation connection, and she couldn't get it out of her head. As she writes in her presentation notes, dogs make it easier for people to handle livestock and people make the dogs’ jobs easier with a flexible communication system. We help each other. “This is most especially apparent in the whistles shepherds use to help their dogs do their job," Queen writes.
The lexicon—the “words” available for use—is small (maybe a dozen commands). But, Queen says, the whistles have what are called sign relations. They can be symbolic, where the sound doesn't have any connection to its meaning. But they can also be iconic (where you can sort of tell what they mean from their form). And even more language-like, they can also be indexical, where the meaning changes depending on how you use them. But here’s the really cool part: Shepherds vary the whistles’ rhythms, pitches, speed, and volume, and "each of those variations provides different kinds of information about what the dog should do,” Queen says. That’s called “prosody,” and it’s a key part of human language.
Shepherds don’t think of their dogs as little furry people. They understand them as dogs.
ROBIN QUEEN, UNIVERSITY OF MICHIGAN
Just as you might speak more loudly and clearly if you think someone doesn’t understand you, a shepherd will more clearly and slowly blow a command if the dog seems to hesitate. Higher pitches attract attention. A faster whistle tells the dog to speed up, even if they haven’t been trained to do it. (That’s “iconic meaning.”)
When commands have to come faster or more urgently, handlers simplify and remove the parts of the shared language that they don’t need. Queen says this is an example of “metapragmatics,” or speakers understanding how to use their speech. This communication system has none of the “who’s-a-good-dog-yes-you-are” cooing that you might hear between a dog owner and its pet. “Shepherds don’t think of their dogs as little furry people. They understand them as dogs,” Queen says. “It’s this really interesting question of, how do you communicate with a species that doesn’t share your communication system, that doesn’t share your kind of mind?” The answer, roughly, is that anything that might convey whether the handler thinks the dog is doing well or poorly gets cut. “Those parts of language that the dog can’t understand—because it’s not a human—come out.”
Handlers even start to acquire a certain style and élan as they get more experienced. Human and dog learn each other's idiosyncrasies and styles. As Queen talked to more handlers about how they thought about what they were doing, she got better at it herself—and came to appreciate other people's skills all the more. “As people learn to do this, they become much more aware of the nuance,” Queen says. “They become much more able to understand, in a sense, the conversation going on between an accomplished shepherd and a dog.”
Just like when you learn a new language.
Nobody knows if the dogs understand what the whistles mean in a metacognitive sense. Lots of animals execute complex behaviors rigidly, instinctually. When birds flock, they’re following other birds’ pointing and directionality, but not (perhaps) with intention. The collective behavior is an emergent property. A dog bred to point at prey orients toward it almost reflexively but doesn’t use the same behavior to merely indicate objects of interest, like a favorite toy back home. “There are all these things that animals do that are rigid, computer-like. What’s special about cognition is that it’s flexible,” MacLean says. “There are lots of examples of animals that have seemingly complex behaviors, and you do one teeny thing to change the situation and the whole thing falls apart.”
Queen says that handlers often impute emotional or cognitive significance to their dogs’ actions when they’re working the herd—they’re being good, or they know they’re misbehaving, or they want to help. “Presumptions about what the dog is thinking,” says Alexandra Horowitz, a researcher at Barnard who studies dog cognition and olfaction, “sound exactly like the kind of attributions made by companion dog owners and haven’t been subjected to real, empirical scrutiny.”
A skilled herder can read the dog’s signals, but is the dog trying to convey information, or is the person just really good at reading the dog?
EVAN MACLEAN, UNIVERSITY OF ARIZONA
On the other hand, we humans read each other’s behaviors and impute emotional and cognitive content to it all the time. A shared communication system helps us confirm, sometimes, our intuitions about the meaning of those behaviors. But it doesn’t always work.
One thing nobody disputes: Even the smartest border collie doesn’t talk back. “Dogs are flexible at interpreting these signals from humans, but they don’t seem to be as flexible for producing them,” MacLean says. “A skilled herder can read the dog’s signals, and maybe that’s the dog communicating back, but there’s a lot we don’t know about that system. From the dog’s perspective, is this intentional communication? Is the dog trying to convey information flexibly to the person, or is the person just really good at reading the dog’s behavior?”
Sometimes during a sheepdog trial, in the midst of competition, a dog hears a command and comes up short—stopping suddenly, ears pricking up, maybe even looking back at the shepherd. Usually the handler interprets that as surprise, as the dog suggesting that maybe the handler messed up: “Is that really what you meant?” The truth is, Queen says, “We don’t really know much at all about dogs’ cognitive architecture other than what we can deduce from their behavior.”
That’s why the cooperative aspect of all this is so important—maybe even the key to how language and cognition evolve. Dog domestication probably goes back at least 15,000 years, but the breeds most familiar today are only a few hundred years old at most, and not all of them have shown the, ahem, cognitive ability of a border collie. (Yorkshire terriers, for example, failed most of the same skill acquisition tests.) But flip that around: Human breeders were able to activate a phenotype for language, or linguistic understanding, in just a couple centuries.
Sure, maybe it was an accident. They were breeding for temperament or skills, and linguistic ability essentially came along for the genetic ride. But still! “Like, this core component of what makes humans special, we can bring it about in the blink of an eye, evolutionarily,” MacLean says.
If you can imagine a cognitive, linguistic bond with one non-human species, you can imagine it, perhaps, with others—octopuses, let’s say, or crows and ravens, all famously intelligent in weird, non-human ways. So it should be easy to further imagine communicating at a high signal rate with some other human, even one who doesn’t think about things the way you do.
It’s worth it. It’s worth figuring out how we all talk to each other. Queen and her partner now manage a small farm with livestock. She’s hoping her newest dog, Scout, will be ready for sheepdog competition in about a year, and only slightly regretful that she still relies on a mechanical whistle instead of using her fingers, like some of the old pros do. “Honestly, it’s like magic. When it works well, it’s like nothing I’ve ever experienced,” she says. “When you and the dog are working as a team, it’s just glorious.”
Actual communication, cooperating in service of a greater objective. “A lot of shepherds refer to it as ‘grace,’” Queen says. “Like, it is the epitome of grace.”