Do Dogs Hear Shapes? The Bouba/Kiki Effect in Animals

The Headphones Come Off, the Dogs Still Know

Fifty-one dogs sat in a room with their owners, each dog held while facing two panels, one showing a rounded black shape, the other a spiky black one, both on a light background^2,4. A recording played the made-up word "bouba" or "kiki" fifteen times over thirty seconds^3,4. To make sure the owners weren't somehow tipping their dogs off, with a shift in posture or a flicker of attention, the researchers ran it again with owners wearing headphones playing music, so they could not hear which word was being used at all^2,3.

The dogs still looked the right way. When "bouba" played, more of them lingered on the rounded shape; when "kiki" played, more of them concentrated on the spiky one, and the bias toward the round shape during "bouba" came through slightly stronger with owners deaf to the cue^2,7.

To understand why that small result landed as a genuine surprise, it helps to go back almost a century, to a much simpler question with no dogs involved at all. In 1924, Georgian psychologist Dimitri Uznadze gave ten people a list of nonsense words and six drawings, and asked them to match words to pictures⁸. He noticed, and wrote about with some puzzlement, that people converged on the same matches far more often than chance would predict, as if something in the sound itself were steering them⁸.

Wolfgang Kohler picked up the thread in 1929, showing people two shapes, one rounded, one jagged, and asking which was "takete" and which was "maluma." Without quite saying so outright, his results implied a strong pull toward pairing the jagged shape with "takete" and the round one with "maluma"^1,8. The modern version came in 2001, when neuroscientists V. S. Ramachandran and Edward Hubbard swapped in the words "bouba" and "kiki," and reported that about 95 percent of people picked the curvy shape as "bouba" and the jagged one as "kiki"^8,1. They floated an idea with real stakes attached: that this sound-shape pull might be a leftover clue to how language itself got started, evidence that naming things was never entirely arbitrary⁸.

Over the next two decades the effect held up across an impressive range of humans: 917 people speaking 25 languages and 10 writing systems in a 2022 study, toddlers as young as two and a half, four-month-old infants who hadn't learned to talk yet, and even, to a lesser degree, people blind from birth matching sounds to shapes by touch⁸. It looked, increasingly, like something wired deep into how brains link senses together, not a fact about any particular language.

That raised an obvious next question: if it's not about language, would animals have it too? In 2019, researchers ran the classic forced-choice test on great apes and found nothing. Humans showed the sound-symbolic matching; the apes, chimpanzees among them, showed no comparable preference at all⁵. That result mattered because it broke the easy story. If sound-shape mapping were simply handed down the primate family tree, our closest living relatives should have shown some trace of it, and they did not⁵.

The dog result, reported in October 2026 by Eleonora Biffi and colleagues at the University of Parma, working with Eotvos Lorand University's Attila Andics, landed as something close to the opposite finding^2,9. Dogs, with no language of their own, showed the human pattern anyway: rounded shape for "bouba," spiky shape for "kiki," with the headphone control ruling out the simplest explanation that owners were somehow giving it away^2,3.

It turns out dogs are not alone, and not even the most surprising case. Newly hatched chickens, one to three days old, have been shown to prefer the same pairings, a species about as far from human language as you can get⁸. And in a 2026 study of tortoises, ten animals exposed to "bouba" and "kiki" sounds alongside round and spiky objects showed the mapping flipped backward: they spent longer with the round object during "kiki" and the spiky one during "bouba," the reverse of humans, dogs, and chicks⁶. Andics, reflecting on the dog findings, suggested the matching may rest on "general mechanisms rather than complex linguistic representations," with an open question being "to what extent these links in dogs arise through statistical learning or more hardwired expectations"².

Back in that Parma lab, with the music playing in their owners' headphones, the dogs had no way of knowing any of this. They just kept looking toward the shape that somehow matched the sound.

What the Effect Actually Is

The bouba/kiki effect describes a cross-modal correspondence: a link the brain draws between information arriving through different senses, here sound and vision^8,1. It is not proof that "bouba" or "kiki" mean anything on their own. The leading explanation points to the physical mechanics of making the sounds themselves: "bouba" uses soft, voiced consonants and rounded vowels, while "kiki" uses sharper consonants and a high front vowel, and researchers have proposed that the mouth shape needed to say each word may map onto the visual shape people associate with it⁸.

Not every language shows the effect equally. A 2021 study found Mandarin Chinese, Turkish, Romanian, and Albanian speakers matched sounds to shapes at lower-than-chance rates on average, though the authors were cautious about ruling the effect out entirely in those languages⁸. Roman alphabet orthography appears to give the effect a slight, modest boost, but a weak one⁸. Brain imaging has also found a signature of it: a 2019 fMRI study found stronger prefrontal activation when sounds and shapes were mismatched than when they matched, along with effects detectable in early auditory and visual cortex⁸.

The Dog Study, by the Numbers

• 51: dogs in the first Parma/Eotvos Lorand experiment, held by their owners while facing two shapes^2,3,4
• 15 times over 30 seconds: how often the recording played "bouba" or "kiki" during each trial^3,4
• 56%: of dogs lingered longer on the rounded shape when "bouba" played^2,3,4
• 63%: of dogs concentrated their gaze on the spiky shape when "kiki" played^2,3,4
• 95%: of humans, in Ramachandran and Hubbard's original 2001 report, picked the curvy shape as "bouba" and the jagged one as "kiki"⁸
• 70: dogs cited in some reporting as the total study sample, though how that figure relates to the 51-dog first experiment is not clearly explained in available accounts^10,11

Who's Who

• Wolfgang Kohler: German American psychologist who popularized the takete/maluma version of the effect in 1929, building on earlier work by Dimitri Uznadze^8,1
• V. S. Ramachandran and Edward Hubbard: neuroscientists who introduced the "bouba" and "kiki" labels in a widely cited 2001 paper and linked the effect to theories of language evolution^8,1
• Eleonora Biffi: researcher at the University of Parma who led the 2026 dog study²
• Attila Andics: researcher at Eotvos Lorand University in Hungary who collaborated on the dog study and commented on its implications for canine cognition²

What's Still Open

It isn't settled whether dogs' sensitivity to the effect comes from thousands of years living alongside humans and hearing human speech, or from something more general about how mammalian (or even non-mammalian) brains integrate sound and sight. Biffi's team is now testing puppies and wild canids, including foxes, along with cats, to try to separate those possibilities².

The animal evidence doesn't line up neatly by species relatedness. Great apes, humans' closest living relatives, showed no comparable effect in a 2019 study⁵. Chicks, only one to three days old, show the standard human-like pairing⁸. Tortoises show it reversed⁶. Reports differ on how many dogs were tested in total, with some accounts citing 70 across the full study and others describing the 51-dog figure as the clearest count for the main experiment, without a clear explanation of how the numbers relate^10,11,2.

Timeline

1. 1924Dimitri Uznadze documents consistent nonsense-word-to-shape matching among ten participants⁸.
2. 1929Wolfgang Kohler presents the takete/maluma version of the effect^8,1.
3. 1947Kohler revisits the association in Gestalt Psychology¹.
4. 2001V. S. Ramachandran and Edward Hubbard introduce the "bouba" and "kiki" labels and link the effect to language evolution^8,1.
5. 2013Ozge Ozturk and colleagues find four-month-old infants show the same sound-shape bias as adults⁸.
6. 2019A study finds humans but not great apes show sound-symbolic congruency detection⁵.
7. 2019First fMRI study of the effect finds stronger prefrontal activation for mismatched sound-shape pairs⁸.
8. 2021A large study finds weaker-than-chance effects in Mandarin, Turkish, Romanian, and Albanian speakers⁸.
9. 2022Cross-cultural study of 917 participants across 25 languages confirms the effect broadly⁸.
10. 2026-07-14A study reports a reversed bouba/kiki mapping in tortoises⁶.
11. 2026-10-13New Scientist reports the Parma/Eotvos Lorand dog study results².
12. 2026-10-14"Hearing shapes: The Bouba-Kiki effect in adult domestic dogs" is published⁹.

Pictures