Like many devoted dog owners, Alexis Devine spends hours every day sitting in her living room talking to her dog, Bunny. The peculiar thing is that Bunny “talks” back. Scroll through Devine’s TikTok page and you’ll see a stream of videos that follow the same general pattern. Bunny stands next to a collection of buttons on the floor, raises a paw, and presses down. The prerecorded buttons sound off in the order she presses them: “More, Scritches, Now.”

People are fascinated by Bunny and her ability to “talk.” She has 5 million followers on TikTok, and the likes on each video are in the hundreds of thousands. There are parody videos on TikTok and existential jokes on Twitter about Bunny’s sentience. Devine has been enjoying the parodies. “Most of the memes are really funny, I got a good laugh out of them,” she says.

Along with Bunny’s demands for scritches, Devine, an artist and self-identified nonexpert in dog science, fields hundreds of questions from humans every day. One question persists among fans and skeptics alike: is this dog really “talking”? Inspired by Bunny’s videos, researchers at the Comparative Cognition Lab at UC San Diego are trying to find out. They haven’t gotten anywhere close to an answer yet, but they’re gathering a lot of data along the way.

Bunny’s journey started when Devine saw videos from Christina Hunger, a speech-language pathologist who has been teaching her dog Stella to use a board full of buttons with words prerecorded on them. The board is an augmentative and alternative communication (AAC) device — an umbrella term for tools from boards with symbols on them to speech-generating devices — which is typically used by nonverbal people to communicate without speech. Inspired by Hunger, Devine diligently trained Bunny from puppyhood and started setting up her own system one button at a time. Bunny is now 15 months old, and her system has expanded into a mat with over 70 buttons.

After Devine’s videos started picking up traction in early spring, Federico Rossano, director of the Comparative Cognition Lab at UC San Diego, started discussing them with people in his department. They began planning a project to study Bunny and other dogs like her who are learning to use the buttons. They hope to determine scientifically whether non-humans can really use something like language to communicate. There are now over 700 participants, including dogs, cats, and even horses, and Rossano says the growing number is almost certainly due to Bunny’s popularity drawing people in.

Each participant receives instructions on how to set up their buttons, starting with words like “outside” and “play.” Cameras are constantly pointed at their personalized boards, and that footage is sent to the lab where researchers comb through and code what they see. “We want to make sure we’re not just getting cherry-picked clips,” says Rossano.

Rossano and his colleagues plan to use the footage to understand different aspects of animal cognition and communication — not just whether they can communicate using something like language, but also how that communication might work. One of the first things they’re looking at is how quickly the animals are learning to use the buttons. That means basic data collection, like figuring out the speed at which a dog learns to associate a button that says “outside” with going outside. Rossano’s hope is that with a large pool of diverse participants, they may be able to start drawing connections between factors like breed or age with learning speed.

They’re also looking at how much the animals seem to be exhibiting properties that are generally claimed to be uniquely human, like temporal and spatial displacement, or the ability to make observations and form narratives. When Bunny asks “Where, Dad” does that mean she has a sense of spatial displacement, where she is aware of “Dad” and acknowledging that he is not present in the room with her? When another dog presses “Water, Outside,” is that an observation about the rain, or is it a request?

One of the most interesting recent introductions to Bunny’s board, at the prompting of researchers, has been words that are related to concepts of time, including “morning,” “evening,” “yesterday,” and “tomorrow.” There’s not much known about how dogs might conceptualize time. Lisa Gunter, a research fellow at Arizona State University who has worked with dogs with separation anxiety, thinks dogs likely have a concept of duration, “but who’s to say how they would describe it.”

The next step, anticipated for winter 2021, is to send researchers to the animals’ homes to conduct more controlled experiments. Will the dogs be able to produce the same seemingly remarkable behaviors with outside researchers that they regularly display for their owners? These experiments will be critical in drawing any conclusions about exactly how much they understand.

When Bunny presses “Settle, Sound, Ouch,” she might be using a novel string of known words to tell someone to quiet down, or she might be pressing a random series of buttons while confirmation bias on our part does the rest of the work. Even Devine says that she thinks Bunny’s “speech” is primarily operant conditioning, where Bunny has made an association between pressing a button and something happening. A true understanding of language goes beyond simple associations, and involves pulling unique combinations of words together into narratives.

Bunny and her cohort are part of a long legacy of the search for human-like communication and cognition in animals. There are famous non-human primate examples like Kanzi, the bonobo who has memorized hundreds of symbols on a special keyboard. There are also dogs like Chaser, who could remember the names of over 1,000 objects. The researchers at UC San Diego are less interested in how many symbols or words Bunny can memorize, and more in how her vocabulary might lend to meaningful communication with humans.

One of the most infamous cases of animal-human communication is that of Clever Hans, the 20th century horse who could apparently provide answers to simple math questions by tapping his hoof. Upon further investigation, it turned out Hans wasn’t doing any arithmetic but was instead reading subtle cues from whoever was questioning him to know when to stop tapping.

Researchers are wary of falling into the trap of the Clever Hans effect. Rossano says the videos of Bunny are interesting, “but we need to be very careful about what we think is going on. There’s a lot of risk about making bold claims.” He wants to gather as much data as possible, and until experiments determine how much humans influence their companions’ actions, he won’t be drawing any hard conclusions about Bunny’s capacity for language.

Even if it turns out Bunny’s button-pressing isn’t exactly robust communication, Rossano thinks the research is on the right track in comparison to past experiments, where animals were taken out of their natural habitats. “Dogs are enculturated naturally, they live with humans,” Rossano says. This connection makes them particularly handy subjects in research, especially when it’s being conducted in their own homes.

Thanks to our shared history that reaches back thousands of years, dogs already have a significant understanding of human communication and expression. “That’s what makes them really different than doing primate research or doing any sort of research where the animals aren’t intimately involved with us,” says Gunter. “The amount of time the dogs have to just watch us and learn, I think it can’t be understated.”

Because dogs are learning from us all the time, and they have their own established ways of communicating, Gunter worries that excitement over projects like this might detract from the relationships we already have with dogs. “I don’t want it to take away from all the fantastic ways that we share our lives with dogs, and how they talk to us all the time about what they want and how they’re feeling.”

But Rossano thinks dogs’ built-in communication — nonverbal vocalizations, gestures, sniffs — is not exactly easy to learn for most humans, with the exception of experienced trainers. He wants to see if dogs will be able to learn and combine words in order to better communicate their wants and needs, with less guesswork on the side of their humans.

Gunter thinks people’s outsized reactions to Bunny videos may be a reflection of our nervousness when it comes to fully providing for our companions’ needs. A dog with the potential to communicate with us in a new way could push us to accept that animals “have their own thoughts, wants, needs, desires,” she says. “I think that likely means that we’re gonna come up short sometimes.”

“And even if people have an existential crisis about it,” Gunter says, “maybe that just means we look at the world from their perspective a little bit more.”

Devine considers Bunny’s perspective often, while also maintaining a healthy skepticism about Bunny’s level of understanding. “I don’t think she understands it in the way we understand it at all really,” she says. But she still finds the process engaging for both of them, saying it’s brought them closer. Bunny uses her buttons all day long, but if she ever becomes disinterested in the buttons, Devine says, “then that’s it, it’s fine. It’s all about our relationship.”