70 Dogs Look at the Round Shape When They Hear 'Bouba'

Play a dog the nonsense word "bouba" beside a rounded shape and a spiky one, and it tends to look at the round one. Play "kiki," and its eyes drift to the spikes.

I think this matters far beyond the novelty of it. The comfortable view is that language is a special human faculty that arrived more or less whole, and that tricks like this are party pieces. My view is the opposite. Much of what we call language is older animal cognition, the brain's habit of tying what it hears to what it sees and touches, wearing a suit and carrying a briefcase.

What the dogs did

The work was led by Eleonora Biffi and colleagues at the University of Parma in Italy and published in Proceedings of the Royal Society B ^[1][2]. The bouba-kiki effect itself is old news in people: ask them which of two shapes is "bouba" and which is "kiki," and they match the soft sound to the round form and the sharp sound to the spiky one.

The team recruited 70 domestic dogs for two experiments ^[2]. In the first, 51 dogs took part with their owners present ^[2]. In the second, 19 other dogs were tested while their owners wore headphones playing music, so nobody in the room could hear the test words and nudge the animal by accident ^[2].

Each dog saw a rounded shape and a spiky shape while a recording of a person saying "bouba" or "kiki" played. Software then tracked the direction of the dogs' heads in video recordings, as an estimate of where they were looking ^[2].

In the first experiment, 56 per cent of dogs lingered on the rounded shape when they heard "bouba," and 63 per cent concentrated their gaze on the spiky shape when they heard "kiki" ^[1]. The pattern also appeared when owners could not hear the words, with a slightly stronger lean toward the rounded shape for "bouba" ^[1]. The first group also showed a general bias toward spiky shapes, which did not appear in the headphone group ^[2].

Those are not dramatic numbers. Fifty-six per cent is a tilt, not a landslide. But a tilt that shows up in a species that never evolved speech is the whole point.

The best case for the sceptics

Let me give the doubters their due, because they have a real argument. Dogs spend their lives surrounded by human talk. A dog could have learned, through years of listening to people, that certain kinds of sounds tend to go with certain kinds of things. On that reading the dogs are not revealing an ancient wiring, they are showing how much a clever animal picks up from living with us.

The study cannot rule that out. It does not establish whether the association comes from learning or from inborn expectations, and Attila Andics of Eötvös Loránd University said the balance between statistical learning and hardwired expectations is an important next question ^[1]. The study also tested only where dogs looked while they heard certain sounds. It did not test whether dogs understand words, use language, or share the human package of abilities ^[1][2].

So no, your dog does not speak. Nobody serious is claiming it does.

Why I think the roots run deeper

But weigh the dog result against its neighbours. In February 2026, separate research found that newly hatched chicks also tended to link "bouba" with rounded shapes and "kiki" with spiky ones, in birds tested one day after hatching ^[3]. New Scientist notes earlier reports in chicks and possible evidence in macaques ^[1].

A chick one day out of the egg has had almost no time to learn anything about anyone's speech. A dog has had plenty. Humans have had everything. When the same sound-to-shape habit turns up across birds, dogs and people, the cheapest explanation is a shared one, not three separate accidents of upbringing. That is my judgement, not a finding, and the researchers are careful not to claim it outright.

Biffi's own reading points the same way. She said that if animals without human-like language spontaneously link sounds to shapes, some mechanisms behind sound symbolism may not be specific to language, and may instead reflect general ways the brain combines information from different senses ^[1]. That is a modest sentence, and I think it is a large one.

Why? Because the human-module view needs words to have been arbitrary at the start.

Nothing about the sound "bouba" is supposed to resemble a round object, which is why a French speaker and a Korean speaker use different noises for the same thing. If even a few sounds carry a built-in pull toward particular shapes, then the first words did not have to be conjured from silence. They could lean on a bridge between hearing and seeing that was already standing.

What I find affecting is how ordinary the machinery is. Every animal with ears and eyes has to work out which noise belongs to which object. An animal that failed to link the rustle with the thing in the grass would not leave many descendants. If language rode in on that ability, it would be a very old tool put to a new job, not a miracle.

I come to this with a personal stake. I live with someone who is losing words, and it has made me curious about what sits underneath them. It is a note, not evidence. The evidence is the dogs and the chicks.

What this does not buy us

I should be plain about the limits of my own argument. A link between a soft sound and a round shape is a small thing. It is not grammar, not syntax, not the ability to say that yesterday the round thing was elsewhere.

The leap from "brains everywhere link sound and shape" to "language is animal cognition" is mine, and a long chain of other work would have to hold it up. The dog study does not settle whether the association is a shared inheritance of distant ancestors, and by itself it does not show that language grew out of this particular link ^[1].

The accessible reports also leave gaps. They describe the design and headline results, but not the full statistics, trial counts or exclusion rules. How robust the effect is across other sounds, shapes, languages and independent replications remains to be shown. Until someone repeats it, 70 dogs is a promising start, not a verdict.

So why stake out a view at all? Because the alternative, treating every such result as a curiosity until the full proof arrives, is how a field avoids thinking. The pattern across species is already there to be argued about, and I would rather argue it than file it under "fun."

What to watch

The next experiments are the ones that can hurt my case. Biffi said the team is collecting data on puppies and on wild canids, including foxes, and wants to test whether cats show the effect ^[1]. If puppies with little exposure to speech lean the same way, the learning explanation weakens. If foxes, who have never lived at our tables, do too, it weakens further. If cats shrug, we will have learned something worth knowing about which animals bother to listen.

Either way, the next time you say "bouba" to a dog, watch where it looks. It may be telling you that you are less special than you thought, and in better company.