Bunny, centre, at home with her owner, Alexis Devine, who says the dog has learned 92 words to date. Photo / Ruth Fremson, The New York Times
Bunny, an internet-famous sheepadoodle, has brought attention to a new area of study within animal cognition: the use of assistive technology for language acquisition.
Alexis Devine said that she knew early on that Bunny, her sheepadoodle puppy, was destined to talk.
A 40-year-old artist and jewelry designer in Tacoma, Washington, Devine had pored over literature on canine cognition, communication and training in the months leading up to Bunny's arrival.
Through her research, she came across the Instagram page of a speech pathologist named Christina Hunger, who was documenting how her own dog, Stella, was beginning to develop an English vocabulary.
Stella had a soundboard made up of circular buttons, each of which dictated a word when pressed. By pawing the buttons, which together formed loosely structured sentences, Stella was supposedly communicating in English.
Hunger, 27, had been working for several years with assistive technology — in particular, alternative and augmentative communication (AAC) devices — to help nonverbal children acquire vocabulary and communicate without speaking.
It had long been common practice for speech pathologists to restrict the vocabulary on children's AAC devices, the idea being that too many words would overwhelm them. But the conventional wisdom among communication experts had been shifting in favor of letting users of the devices demonstrate their own capabilities by giving them as many words to play with as possible.
One wouldn't assume a baby was speechless if he hadn't uttered his first word by 12 months, Hunger reasoned. So why should that logic be applied to those merely incapable of speech — whether that be a child who suffers from catatonia or a creature devoid of the necessary organs of speech, like a dog?
Hunger began experimenting. Most AAC devices were either too expensive or unsuitable for canine use, so she chose the cheapest option she could find online: a four-pack of recordable answer buzzers.
The box arrived at her San Diego home a week after Stella. Hunger decided that a button that said the word "outside" would be the best place to start when it came to walking and house training. Within a few weeks, Stella was regularly and routinely pressing the button to be let out.
Devine had read about Stella on Hunger's blog. So when Bunny showed up in October 2019, her own first button — "outside" — was already waiting by the door.
Dogs have learned many tricks in the 20,000-odd years since they are believed to have first been domesticated. Most can respond to basic commands like "sit" and "stay." They can recall terms like "treat" and "walk." Some have demonstrated a rather human capacity for quickly picking up the names of new objects and storing them for future retrieval.
"Domestication is likely to have affected dogs' brain positions so they can interact and socialise with humans better," said Claudia Fugazza, a researcher in the department of ethology (that's animal behaviour) in Eotvos Lorand University in Budapest, Hungary. "They are probably more predisposed to interact with humans as social partners."
All of this is to say, it's clear that dogs can follow a wide array of human social cues. But outside of movies and TV shows, dog owners have seldom claimed that their pets possess the ability to speak.
"Bunny can now speak 92 words," Devine said on a Zoom call in April, her dog just in frame and blending in with the fluffy rug beneath them. Bunny is almost 2 years old now, and her language acquisition might rival that of a human toddler. (The typical human 2-year-old can use at least 50 words with ease.)
According to Devine, Bunny can use the buttons on her soundboard to form four-word phrases. She can ask questions. She can, and often does, tell people to shut up — or, in the words of her buttons, "settle down."
"For a long time, Bunny was talking almost exclusively about poop," Devine said. "But toddlers do that too, right?"
With 6.6 million followers on TikTok and 818,000 on Instagram, Bunny has become the poster girl for Hunger's canine AAC movement. "Alexis is amazing at social media," said Hunger, who has nearly 800,000 of her own followers on Instagram, most of whom seem to be there for the dog content.
Most of the dogs (and their owners) dabbling in this area — and there are many; just search the hashtag #hungerforwords — don't have Bunny's social media paw print. Passersby frequently recognize her on walks. "There was one instance where a car did a U-turn in traffic and stopped in the middle of the road and rolled down their window to say hello," Devine said.
In early 2020, about six months after Bunny learned "outside," Devine was contacted by Leo Trottier, a product developer who works in the pet industry. He was hoping to work together.
In 2016, Trottier, a doctoral candidate with a master's degree in cognitive science, introduced CleverPet, the world's first game console for dogs. But after a failed attempt to raise funds for the product on Kickstarter, he abandoned the project.
Three years later, when Trottier discovered Hunger's work, he saw an opportunity for collaboration. While Hunger and Devine were using simple prerecorded sound buttons they'd found on Amazon, Trottier was developing FluentPet, an AAC device designed for dogs, and he was looking for beta testers. (Hunger had signed a book deal with HarperCollins around the time of FluentPet's beta release and declined Trottier's request to collaborate.)
Trottier reached out to Federico Rossano, a professor at the University of California, San Diego, to help him — in Rossano's words — "science up" the project.
Rossano, a cognitive researcher who has worked extensively with a range of species, was skeptical at first. But he ultimately saw an opportunity to study dogs' capacity for languagelike abilities in a systematic, rigorous way, with the potential to draw results from a participant pool unlike any he'd been given access to before.
At the same time, Devine, whose jewelry business had slowed significantly during the pandemic, was given the further incentive to become an affiliate influencer for the product, meaning that she would receive upward of 8 per cent of every FluentPet sale made through a referral link to the website from her Instagram page.
In June 2020, Trottier and Rossano started They Can Talk, a research project and an online forum for participants. "Initially, we just thought we'd have a few participants from across the San Francisco and San Diego area," Rossano said. But after lockdowns began in early 2020 and TikTok's popularity rose, thousands of bored homebodies began to wonder whether their pet could talk like Bunny, too.
Currently, the study has more than 2,500 participants. Buying the FluentPet product isn't required in order to participate, but there is an incentive on the study's website. (Prices range from US$29.25 for a tester kit to US$195.95 for a 32-button set.)
"We have a data-sharing agreement," Rossano said. "I am the scientific lead of the project, and the analysis and findings will be reported in scientific papers."
To avoid a conflict of interest, Rossano is not being paid for his work on the study. Ideally, he would prefer for the research to operate as independently as possible from FluentPet, but a study of this size required the company's sponsorship.
"I am a scientist, and as far as I am concerned, my job is to assess whether these devices are revealing cognitive abilities that are novel and unexpected or whether this can all be explained through simple learning mechanisms common across several animal species," Rossano said.
For at least 200 years, researchers have reported several instances of nonhuman animals demonstrating remarkable languagelike capabilities. One such example, which has loomed large over the field of comparative cognitive studies since the early 20th century, was the case of a horse named Clever Hans.
Hans appeared capable of responding to simple arithmetic calculations with accuracy. For instance, when asked "What is 2+2?" he would tap his hoof four times.
But when psychologist Oskar Pfungst analysed the horse in 1907, he concluded that Hans was merely responding to humans' cues rather than showing that it could understand human speech. The "Clever Hans Effect" has since pushed scientists to develop methods that remove human presence and influence from animal cognition studies in order to avoid false positives.
In the late 1950s, primates became the focus of studies on the linguistic abilities of nonhuman animals, particularly chimpanzees. Intent on teaching spoken language to young chimps, scientists quickly hit a roadblock: Nonhuman animals don't possess the vocal apparatus to open their mouths and say, "Hey, you."
In the decade that followed, comparative cognitive scientists began to take inspiration from disability studies, shifting their attention toward manual languages like American Sign Language.
The idea that visually conveyed language contained the same potential for expression as speech was still relatively new when chimpanzees, bonobos, gorillas and orangutans began demonstrating their ability to learn and use ASL-based gestures. In the 1990s, the idea that dog development could mirror a toddler's began to gain traction, but the research into canine communication remains very rudimentary.
"Canine neuroscience is a relatively novel field," said Fugazza, the researcher in Budapest. In 2017, Gregory Berns, a professor of neuroeconomics at Emory University, led a training program that taught dogs to walk into an fMRI scanner without sedation or restraint. With the dogs inside, their owners listed the names of surrounding objects and toys, alongside occasional gibberish.
The scans showed that the dogs' brains could quickly discriminate the words they knew from the unfamiliar and nonsense but that dogs seemed to make no distinction between words that differed by a single speech sound (for example, "paw" versus "pow").
Unlike chimps, dogs have been subject to a process of artificial evolution as a result of domestication. Across the past two decades in particular, research has shown that dogs possess a nuanced and social understanding of humans' social cues. One could attribute this to the "domestication hypothesis" — the idea that dogs' social behaviour has been moulded to satisfy human sensibilities.
"We are really interested in a recent finding which showed that there are certain facial movements in dogs that humans find very attractive, which has led to dogs evolving a facial muscle that wolves don't," said Juliane Kaminski, a lecturer in comparative psychology at the University of Portsmouth. "They're puppy-dog eyes, basically. There's a facial expression dogs produce when they raise their eyebrows, and this is a movement that resembles sadness in humans and that they seem to find extremely attractive in dogs' faces."
Rossano said that instead of asking whether dogs can understand humans, "we can ask whether they can learn to communicate with humans using human signals."
Still, Judith Schwarzburg-Benz, a philosopher and senior researcher at the Clever Dog Lab in Vienna, wonders how much we can truly learn about canine communication through a human lens. "What we get is a very reduced picture," she said. "I think we can only get glimpses into the mind and learning processes, like very specific questions."
Indeed, at what point could one say with satisfaction that Bunny can talk? Would Bunny have to fulfill every item on a linguistic checklist, or only a certain number? And how would that be determined?
"If dog AAC is going to be as big as I think it is, as I think it can be, it's going to take a lot of people working in different arenas from different angles to come at it from all sides," Hunger said.
On May 4, HarperCollins published "How Stella Learned to Talk" (now a New York Times bestseller) alongside the rollout of Hunger's own buttons, which are being distributed on a mass scale, both online and in big-box stores. (A box of four costs US$28.40.)
"I think this has the potential to change our relationships with dogs forever," Hunger said.
For Devine, the communication with Bunny comes back to a personal connection. She said she recently heard Bunny pressing the "ouch" button on her board. A few minutes went by before Bunny pressed "stranger" and "paw," then stretched her arm out toward her owner.
"I felt between her paws and found a thorn in there," Devine said. "Anytime she chooses to communicate with me in a way that is not her natural communicative method, it feels really special. If she's going out of her way because she trusts me and wants to engage, then I just know that she loves me."
Written by: Emma Madden
Photographs by: Ruth Fremson
© 2021 THE NEW YORK TIMES
New York Times: Our are devices distancing us from real connections?