The most contentious issue in the field concerns the animals’ true aptitude with language, and what it signifies. Photo: Dougie Wallace

Want to talk to your dog? Science is finding a way

A few promising—if somewhat salacious—experiments with apes and dolphins in the 60's didn't amount to much. Recent breakthroughs using machine learning, however, offer us fresh prospects of decoding animal languages
April 1, 2020

I adopted a dog a few months ago. Suka is a retired sled dog, an old colleague from my time working at a husky kennels in the north of Finland.

I love dogs but this is the first time I’ve owned one. And it struck me, upon her arrival, how strange it is: to cohabit and coexist so closely with another sentient creature, and yet to have no clear way of communicating.

At the start, it was like having a house guest to stay who spoke no English: a little awkward. She was respectful of our space, tried not to get in the way. Greeted us pleasantly when we met one another in the hallway, but then seemed to run out of steam.

Friends asked if she remembered me, and the truthful answer was: I couldn’t tell. What was clear was that, at a fundamental level, we didn’t understand each other. Worse: that she didn’t understand this was her home now, and that she was to stay here, with us, forever.

It broke my heart, those first few days: her sweet docility, her subdued manner, her excitement on seeing other dogs, which quickly faded when they turned out to be strangers.

One thing that must have reassured her was that I used the same basic commands familiar from her kennels: a mongrel mishmash of orders and words of endearment pieced together from various sources. Some are more useful than others. “Gee” and “haw”—generic commands for “right” and “left” used with working animals—could have limited mileage out on walks. There are scraps of English—“come!”; “sit!”—and some Finnish too: hyvä koira! (good dog!); alas! (down!).

That was, essentially, all the common ground we had. Other than that, we were limited to body language. We watched each other carefully, reserved judgments. I moved slowly, a caricature of unthreatening behaviour. When I opened my hands to her in invitation she would approach, but not come close enough to stroke. She wagged her tail, but just that little bit too slowly to be sincere. Probably, I supposed, out of a willingness to please.

I chattered to her anyway, to put her at ease. “Are you hungry?” I asked her, as I stood at the stove. She looked up at me with her big brown eyes. “Are you tired?”

She sat at my feet, then slid onto her belly, never taking her eyes off mine.

“Bored?” I said. “Homesick?” She listened politely, uncomprehending. If only, I thought. If only you could talk.

*** This is a sentiment that I must share with every pet owner in the land, and many scientists working in the field of interspecies communication too. Philosophers have, traditionally, been scathing of such a notion. Descartes, for example, believed there was simply nothing to find out about what animals were thinking. While he famously sought out the foundations of knowledge in the experience of the human mind (“I think, therefore I am …”), he insisted that animals lacked any mind to speak of, and that their bodies were, in effect, automata functioning on instinct alone.

Yet for those who live and work among animals every day, this notion has never rung true. His, and many others’, reduction of animals to their mechanistic parts has done little to deter the human desire to connect with other creatures. Throughout the 20th century, and continuing into the present, a huge amount of time and money has been invested in exactly this: teaching animals to speak.

A fool’s errand, some would say. Well, perhaps. It’s true that the field has, to some extent, self-combusted. But not before producing some fascinating results.

Scientists first focused on teaching animals to produce the sounds of human speech. In the 1940s and 1950s, Catherine and Keith Hayes—two married research psychologists in Florida—had some early success when they adopted a newborn chimpanzee, taking her into their home and raising her as a human baby. Viki managed to produce three words: “mama,” “papa” and “cup”—sometimes with assistance, as the Hayes moved her lips for her—but only after six, nearly seven years of intensive training. (The experiment was cut short when Viki died of meningitis.)

Apes have been a key focus of study, thanks to their shared genetic heritage with humans and assumed similarities. (Although not always: animals as disparate as whales and elephants have been taught to mimic human speech sounds.) But despite their intelligence, apes’ disinclination—or perhaps, given their different physiology, inability—to join us in spoken dialogue was a source of frustration for years. Until in 1967, Allen and Beatrix Gardner, another psychologist couple (and couples do seem to loom large in this field) hit upon the idea of channelling chimpanzees’ natural preference for gesture and body language.

Their success with Washoe, a chimpanzee who lived in a trailer in their garden to whom they taught sign language (and who went on to teach some to her own adopted child), attracted huge public interest and heralded a golden era of ape language studies in the 1970s. Ann and David Premack worked with Sarah, a female chimpanzee, teaching her to communicate by way of plastic objects; Duane Rumbaugh and Sue Savage-Rumbaugh used a keyboard labelled with abstract symbols to communicate with Lana, a chimpanzee, and Kanzi, a bonobo. (Kanzi, who initially learned this experimental form of communication by observing his adopted mother, a laboratory ape who had not herself shown any exceptional aptitude, has been of particular interest—apparently able to discuss events in the past and future, and vocalise speech-like sounds as he uses the keyboard.)

Of all the experimental ape subjects, perhaps the most famous is Koko the Gorilla. From 1972 until Koko’s death in 2018, the American researcher Francine Patterson trained Koko in an adapted form of sign language and claimed that the gorilla had a working vocabulary of more than 1,000 words (and, furthermore, understood another 2,000 spoken words).

Koko’s gentle interactions with her many celebrity callers (including William Shatner and Robin Williams), her love of kittens, and unfulfilled wish for a baby of her own, found her many fans around the world—and saw her make the cover of National Geographic twice. But perhaps more than any other animal language experiment, Koko’s case has come to serve as a lightning rod for scientific scrutiny and scepticism.

The results of many—if not most—interspecies experiments have been extremely contentious. In some cases, safety and ethical issues have arisen. Critics have highlighted the cruelty intrinsic to removing an infant ape from its family and peers so as to immerse it in human culture. Sometimes projects also fell apart due to lack of funding or changing research priorities, leaving their subjects without purpose or properly trained handlers, facing a lifetime in captivity. Other test subjects (most often chimpanzees) became aggressive and injured their trainers.

In other, weirder, cases, the intimacy between researchers and subjects itself has raised suspicion. The case of Peter the dolphin—who cohabited in a semi-flooded lab with trainer Margaret Howe Lovatt in the 1960s, as she intensively coached him to vocalise through his blowhole—became notorious, after it was revealed that Lovatt regularly “relieved” the dolphin when he became sexually aroused during training sessions so as to keep him focused. (“It wasn’t sexual on my part,” she clarified later. “Sensuous, perhaps.”) When the experiment ended, Peter was moved to a small tank, and is said to have committed suicide: he swam to the bottom, and refused to rise again to breathe. A vet later attributed the death to losing “the love of his life,” Lovatt.

Salaciousness aside, the most contentious issue in this field is over the animals’ true aptitude with language, and what it signifies. Linguists have declared that Koko’s fluency in sign language—thought remarkable to a layperson—was far from the “mastery” it was often portrayed as. Her language employed little to no grammar and was subject to no small amount of interpretation by Patterson and her staff.

A transcript of a conversation between Patterson and Koko, broadcast on AOL Live in 2000, shows this in action:

[su_quote]Dr Penny Patterson: Koko, someone wants to ask, “Do you like people?” Do you like people, gorillas, dogs, cats? Which do you like? Koko is pointing to the catalogue. You like reading! I was asking you about living things, like people or gorillas or cats, or maybe even elephants! Or birds! She’s looking at her empty nut bag. Do you like people? No? Well, what do you like? Koko is brushing herself. You do you like scratching. Like when I scratch your back? Or brushing? What do you like? Koko has picked up the videotape. I think she likes watching tapes. We do that a lot in the morning! So, you like watching movies, huh? …Koko is going to the mirror with my brush to brush her hair. She’s purring. She’s happy. Looking in the mirror and brushing her hair makes her happy. Koko: Toilet. PP: Honey? What? How about another question? What’s your favourite colour? K: Lips, fake hair. PP: Oh, you’re being funny![/su_quote]

The spectre overhanging all investigations of this nature is Clever Hans, the famous “counting horse” of Berlin who found fame at the turn of the 20th century for his incredible feats of intelligence. Apparently able to multiply, divide, tell the time, spell, and recognise artists and composers from their work, Clever Hans attracted huge crowds, and answered not only his trainer, but anyone who asked him a question—even strangers.

Initial investigations by German authorities over 18 months cleared Hans’s maths teacher owner of creating a hoax; the mystery was solved when the psychologist Oscar Pfungst realised that Hans could only perform when his questioner already knew the answer. It wasn’t that Clever Hans wasn’t clever, he was clever in a different way: he was so attuned to human behaviour that he could react to involuntary facial signals invisible to human observers.

Critics of ape language studies claim that Koko, Washoe and their ilk are simply echoing the input of their owners back to them—having learned that repeating certain behaviours gets rewarded. If this is true, the apparent profundity of some of their recorded statements perhaps most resembles that spooky sense of significance aroused during the reading of horoscopes or tarot: it says more about us than it does about them.

When Koko the gorilla signed “Cat, cry, have-sorry, Koko-love,” was she really expressing regret over the death of a beloved kitten? Or are we projecting onto her what we want to see?

*** In 1974, Irene Pepperberg—then working on a doctorate in theoretical chemistry at Harvard—found herself bewitched by a documentary on animal communication. In it, chimpanzees grappled with sign language, and researchers charted the dialects and vocal learning in birdsong.

Having grown up around talking parakeets, she was seized by the idea that parrots too might be taught to communicate with researchers—through speech. Three years later, having completed that chemistry doctorate—but all the while devoting every spare moment to studying the latest work in human-animal communication—she decided to switch scientific fields. She bought Alex, an African grey parrot, from a pet shop, and began his training in earnest.

To begin with, she says, she met with a great deal of resistance; applications for funding failed because it was assumed that birds were too, well, bird-brained to converse in any meaningful way. But over time Alex proved them all wrong. Pepperberg would spend the next 30 years patiently teaching Alex, through a carefully-devised method of conversational modelling, to identify shapes, colours, numbers, materials and letter sounds, and to answer questions that required comprehension of spoken English. “What shape red?” she might ask him, offering a tray of various objects. To which he might answer: “three-cornered,” meaning triangular. Or: “What’s different?” while proffering two keys of different appearance, to which he might answer: “colour.”

Alex died in 2007, but videos of their interactions—all of which took place and were recorded in transparent and controlled circumstances—are unsettling to watch. The most exciting developments were when Alex opened up new avenues of his own accord. Once, while apparently demonstrating a remarkable ability to synthesise meaning by pairing phonetic letters with their sounds. He requested a reward (“want a nut”), and when it did not appear, he spontaneously leapt from phonics to spelling, loudly demanding “N, U, T.” Similarly, towards the end of his life, he showed an ability to add numbers together while eavesdropping on another parrot’s counting tutorial—perhaps, though I might be over-interpreting, as a way to show off his superior skills.

This was a spectacular display of cognitive ability, especially for a bird with a brain “the size of a shelled walnut,” as Pepperberg puts it, working in, essentially, his second language. But, as she cautioned me, I mustn’t think of Alex as “speaking English.”

“Two-way communication is what we called it,” she explained to me by phone. I would never have been able to sit down for an interview with Alex, or to converse beyond a limited range of scenarios. But, she adds, “parrots can learn enough that we can test them the way we test young children; you can certainly communicate well enough to examine a lot of cognitive processes.” What Alex offered us—as do his successors in Pepperberg’s programme, Griffin and Athena—is a window into the black box of parrot thought, and a reminder not to underestimate them.

Even with Alex, whose linguistic abilities are yet to be matched, “it wasn’t, and I don’t think it would have ever gotten to the point of, full-blown language,” said Pepperberg. “But the point is, what these animals acquired was very sophisticated… It’s really impressive. Symbolic representation”—meaning using the sounds of English words to represent objects, numbers, and so on—“is not simple.”

Sadly, perhaps, long-running experiments like Pepperberg’s—and the Ape Cognition & Conservation Initiative in Iowa, where Kanzi the bonobo, now 37, lives—are thin on the ground. The field of human-animal communication collapsed under the weight of its controversies during the 1980s. But, says Pepperberg, it would be better to put all this behind us, and move towards a revival. “It’s so fascinating to be able to examine the mind of another species,” she says, “and inter-species communication is a really, really, valid way of doing it.

“Most of these studies were precipitously ended before they should have been. People lost their funding, they couldn’t continue their work, and now [due to tighter rules on the use of animals in research] it’s not legal. So, unless there’s some special dispensation, we’re never going to find out what’s possible.”

*** If anything, Pepperberg’s caveats and caution lead us only deeper into interesting philosophical inquiries. What exactly does it mean to speak, or to communicate? Might we ever really be able to make ourselves fully known to a chimpanzee, dolphin or parrot, or to understand them in turn, when our emotions, social organisation and perception of the world is so at odds? When Wittgenstein said: “If a lion could talk, we would not understand him,” this was what he was getting at. Without a common frame of reference, true communication may be impossible.

But the asking of such questions leads us to some interesting places. In 1961, the neuroscientist and dolphin researcher John C Lilley attended a high-level meeting of scientists (including Carl Sagan and Frank Drake), an early iteration of the Search for Extraterrestrial Intelligence programme (SETI). Lilley, who had been invited on the basis that an expert in interspecies communication would be best placed to advise on the prospect of communicating with alien life-forms, so impressed those present that they dubbed themselves “the Order of the Dolphin.”

Nasa went on to finance Lilley’s dolphin lab on the Caribbean island of St Thomas—the setting of the forbidden romance between Peter the dolphin and Margaret Howe Lovatt—and Sagan visited several times (once meeting a dolphin who demanded “more!” when Sagan stopped scratching its belly). But, as Lilley’s research grew increasingly wacky, rocked by controversy over sexual impropriety and Lilley’s dosing of dolphins with LSD, Sagan’s interest waned.

What was really crucial, the SETI scientists decided, was to understand what dolphins were saying to each other. Only then might they truly get an insight into what an alien language might comprise, and how we might comprehend it—and they us.

Recent breakthroughs using machine learning offer us fresh prospects of doing just what SETI wanted: decoding the messages between members of other species. In 2016, researchers at Tel Aviv University detailed how a modified voice-recognition programme had helped them to classify the vocalisations of captive fruit bats into four types of complaint: jostling for the best sleeping positions, arguing over food, protesting over mating attempts, and generally bitching about their nearest neighbours.

Con Slobodchikoff, the American behaviourist, has spent years studying the alarm calls of prairie dogs—showing them to be remarkably complex, not only identifying the type of danger (be it hawk, dog or person) but encoding aspects like size and colour. After collaborating with a computer scientist to produce a prairie dog “translator,” he recently founded Zoolingua, a company “dedicated to developing technology that will allow us to communicate with animals”—starting, he says, with pet dogs. The tech entrepreneurs Aza Raskin (formerly of and Mozilla) and Britt Selvitelle (a founding team member at Twitter) are also using artificial intelligence to study language use, this time among elephants in the Congo Basin as part of the Earth Species Project.

At the Wild Dolphin Project, a not-for-profit based in Florida, information theory—a mathematical field used in SETI for the detection of exoplanets—has been repurposed to detect and classify dolphin whistles. With the help of a mobile device known as the Cetacean Hearing and Telemetry project (Chat), researchers are working on a common Esperanto-style, go-between “language” of whistles that dolphins might be capable of mimicking; they celebrated a major breakthrough in 2013 when a Chat machine detected use of a whistle representing the word “sargassum,” a type of seaweed, in real time, spitting out an audio translation through a set of earphones into the ears of a diver.

“‘Are we alone?’ has been a question asked at SETI for a long time now,” as Laurence Doyle, a SETI investigator working on the Chat project, said at the time of the breakthrough. “But the fact is we’ve got a million languages on planet Earth that are not human.” Forget sending messages to the stars, alien minds are flying by overhead every minute, swimming through the waters of the harbour, sleeping in front of the fire in my own living room.


Over a period of weeks, Suka and I slowly came to accept one another’s presence. At first it was strange having another creature in the house during the day. I wasn’t used to working with anyone else around. But she was undemanding company, understanding of my slovenly freelancer ways, and we quickly developed our own habits: meal times, walk times, regular routes.

Communication quickly ceased to feel like a problem. Her requests were not difficult to interpret. The dog equivalents of “want a nut,” I guess: want meat, want treat, want out. I could provide all of these things. She learned quickly too: she doesn’t need English to understand she’s not allowed on the sofa. Routine, it turned out, is a form of common understanding.

But one day, a month or so after she arrived, something was different. Wrong. She stopped eating, later vomited on the kitchen floor. We sat up with her, then took her to bed with us. When I woke again in the night, she wasn’t there.

I found her downstairs on the living room floor. She wagged her tail when she saw me, in that slow, uncertain way of hers, but it was clear something was seriously wrong. Blood stained the rug and the backs of her legs.

We rushed her to the emergency vet, who had an answer for us quickly: pyometra, an infection of the uterus—not uncommon, but a critical case. It will have been brewing for days, said the vet. She’s so quiet, so grimly stoic, that we had no idea. “There’s a strong possibility she’ll die on the operating table,” said the vet. “Do you want to go ahead with treatment?” Please, I said. Please.

Suka survived, but she might easily not have. When she was released, 48 hours later, she looked awful: drugged up, absent, looking right through me. I spent two nights with her on the kitchen floor, watching her sleeping, worrying as she refused all offers of food. At 3am, I found myself, again, wishing for connection, a way to talk. I made a deal with myself: it doesn’t have to be language, I decided. We don’t need to debate. Just give us two-way communication, like Pepperberg’s parrots.

Because I know exactly what I need to say.

Are you in pain? Where does it hurt? And how much?