Since 2001, many studies have reported that people are much more likely to give the made-up name “bouba” to a curved shape and “kiki” to a spiky shape. The effect has recently been shown in chicks and possibly macaques, too.

Eleonora Biffi at the University of Parma, Italy, and her colleagues decided to test whether pet dogs have the same tendency.

In the first experiment, 51 owners sat with their adult dogs, holding them during the experiment. They were shown a rounded shape on a panel on one side of the room and a spiky shape on the other. When they heard the word “bouba” 56 per cent of the dogs lingered on the round shape. When they heard “kiki”, 63 per cent concentrated their gaze on the spiky shape.

To make sure the dogs weren’t influenced by their owners, the team repeated the experiment with the owners wearing headphones playing music so they couldn’t hear the words. The outcome was the same, with a slightly stronger bias towards round shapes for the word “bouba”.

“If animals without human-like language also spontaneously associate sounds with shapes, this suggests that at least some of the mechanisms underlying sound symbolism may not be specific to language,” says Biffi. “Instead, they may reflect more general principles of how the brain integrates information from different senses.”

However, it is unclear to what extent the effect depends on dogs having experience of listening to human speech, or the fact that dogs have co-evolved with humans for thousands of years. To investigate this, Biffi says the team is now collecting data on puppies and wild canids, including foxes, and also wants to see if the effect exists in cats.

“If cats also show the bouba–kiki effect, this would provide further evidence that sound symbolism may reflect general perceptual mechanisms, rather than adaptations specifically related to dogs’ long history of cooperation and communication with humans,” says Biffi.

Attila Andics at Eötvös Loránd University in Hungary says the study suggests the matching of speech sounds to shapes relies on general mechanisms rather than complex linguistic representations. “An exciting next question is to what extent [these links] in dogs arise through statistical learning or more hardwired expectations,” he says.