Dogs may be able to communicate by pressing buttons, research suggests

There’s no question that dogs can communicate with their humans. A dog who needs to go out might stand by the door and scratch at it, or bark. A hungry dog might circle a food dish. One who wants to play fetch might bring you a tennis ball.

But can a dog convey its needs by associating specific buttons with prerecorded words — such as “want,” “food,” or “play” — and use its paws to tap out messages?

Ongoing research, including a new study, suggests it may be possible. The study indicates that dogs are able to express themselves by pressing buttons in two or more word combinations that are neither random nor accidental, the authors said.

“If we know that they are using the buttons intentionally, they can use them in ways that seem smart, like a young child,” said Federico Rossano, an associate professor of cognitive science at the University of California at San Diego, who led the research. “This should lead owners to a renewed appreciation of the intelligence of their pets and help them provide for their dogs.”

Rossano and his collaborators have been studying this phenomenon, known as “button dogs,” to establish whether dogs can indicate their desires by using these buttons. Many dog people are believers and have flooded social media with videos and anecdotes of their button-pushing pups. However, this has been met with skepticism from those who study dog cognition and behavior.

The latest study, published in the journal Scientific Reports, is the second in a planned series aimed at ultimately answering the larger question of whether dogs spontaneously can “talk” to humans via the buttons. The first study, in which humans — not dogs — pressed buttons, found that dogs can grasp the meaning of words and respond appropriately.

Since 2022, the longitudinal study has been following several thousand dogs whose button presses are logged using an app accessible to the researchers and designed by FluentPet, which manufactures soundboards. From these, the scientists selected 152 dogs who were pressing two or more buttons in a sequence and analyzed them during a 21-month period.

The most common button sequences were the dog’s name followed by “want,” and then usually followed by another button that indicates what the dog was asking for, such as “food,” “play” or “outside.” Sometimes a time reference will come next, such as “now” or “soon,” according to the study.

The information was self-reported by dog owners as part of the ongoing larger data collection. Those selected for the multi-word combination study were not told about it to avoid bias, such as “having them change their training or interactions with the dogs and the soundboards,” Rossano said.

To determine whether the multibutton combinations were random, the researchers conducted computer simulations on the probability of two specific buttons being pressed one after the other over time given the total number of buttons, and compared these scores to the actual sequential pressing of those two buttons.

“This is how we know that most dogs in this pool were doing multibutton combinations in nonrandom ways,” Rossano said. “Note that nonrandomness can be caused by many things, including imitating the training they received, though the first analysis comparing button presses by owner and their dogs suggests that this is unlikely to be the main explanation.”

However, he acknowledged that they did find several dogs in the study whose button pressing was random, which “is important because if the data lined up as if they were all extremely systematic, it would seem very hard to believe,” he said, adding they saw “different degrees of nonrandomness,” but were able to determine who was “almost certainly” random and who was not.

Combined with previous findings, these results show “that the dogs understand the meaning of the more frequently used words on these soundboards and suggest that dogs are capable of using these soundboards to communicate with humans about their needs and wants,” he said. “It further raises the possibility that they might be communicating the way a 2-year-old human might, which is more sophisticated than previously believed.”

Moreover, the research “can help us get us a window into a dog’s mind, their intelligence and what they care about, potentially leading to an improved sense of companionship and strengthening our bond with our pets,” he added.

But Amritha Mallikarjun, a postdoctoral researcher at the Penn Vet Working Dog Center who was not involved in the research, pointed out that the button presses were “conditional on the action that occurs after the button press,” meaning that the dog’s first button press might have been random but led to more button pushes because of reinforcement.

“Dogs trained on buttons will often like to try things and will take a guess at what we want by performing a random behavior,” she said. “The button-trained dog presses ‘want outside’ one time just to try it. There is no real knowledge of what the concept of the verb ‘want’ is. ‘Outside’ usually means the dog’s person lets you in the backyard, and sometimes you hear your owner press ‘want food,’ so the dog associates the ‘want’ button with good things.”

Such reinforcement may lead a dog to press a button set more often, making the word combination seem deliberate, rather than random, she said. “The conclusion that dog button pressing is deliberate is pretty clear,” she said. “Every attempt from dogs to try and communicate with us is deliberate. Body language is deliberate, too. Once the buttons are a trained behavior, of course the dogs will use them to try and get things they want.”

Clive D.L. Wynne, a professor of psychology and the director of the Canine Science Collaboratory at Arizona State University, who was not involved in the study, said “language, especially English, relies on order of words in sentences to impose meaning, but not all ordering of words means language,” he said. “The most common concept pair they observed was ‘food-treat.’ Is that language competence or just repetition for emphasis?”

The study “suggests some buttons had some associations for the dogs, but to what degree we want to call that language is, I think, still an open question,” he added.

The study was conducted in collaboration with researchers from the University of California at Davis, Johns Hopkins University, the Universitat de València in Spain and CleverPet, the parent company of FluentPet. Some of the team members have served as consultants to CleverPet or worked for the company, but the scientists received no study funding from any commercial manufacturers of the button boards, Rossano said.

The researchers did not provide participants with the boards, which people can buy commercially, from less expensive “starter kits” to more complicated soundboards that cost several hundred dollars.

Also, “FluentPet did not know which specific participants we were including in our analysis nor which criteria we were using, so they could not bias our selection,” Rossano said.