Suppose I show you two objects, one round and the other pointed. Which one do you think is called “bouba” and which one is “kiki”? Your answer is usually immediate: round is “bouba”, pointed is “kiki”, of course. Since its discovery a century ago, this “bouba-kiki effect” might seem trivial. Quite the contrary, this phenomenon actually reveals several intriguing paradoxes with respect to what is known about language and our perception of the world.
The first paradox is that it suggests that, for certain words, there is a link between the sounds they contain and their meaning. Whereas, for most words in the languages ??of the world, there are none. The second paradox is that although it belongs to language, this effect is universal: it is found across a large number of different languages ??and cultures. Finally, my work in babies suggests a third paradox: although universal, the bouba-kiki effect does not seem to be present at birth. It would not be genetically coded (which could explain its universality), but rather learned. So how can we learn that “bouba” is round and “kiki” pointed, regardless of the languages ??and cultures we have?
We recently solved these three paradoxes at once by demonstrating that the bouba-kiki effect does not come from language, but from the physics of objects. We first showed that, on closer inspection, the word “bouba”, considered round, is actually composed of deeper and more continuous sounds than “kiki”, considered pointed. Then, we showed that, precisely, when they roll on the ground, round objects produce more serious and continuous sounds than pointed objects. It would therefore be by integrating these physical properties of objects into language that, for humans, “bouba” became universally round, across languages ??and cultures.
To analyze speech stimuli, we first turned to phonetics, a field of linguistics. Using a model that reproduces what the human ear hears and what our brain perceives, we were able to show that speech sounds like “bouba” or “malouma” contain more continuous and deeper sounds than “kiki” or “takete”, which are more discontinuous and high-pitched. When we talk about continuity, we are referring to whether the sound stops or not. You can hear that quite well: “bouba” and “malouma” are softer to the ear (and therefore continuous) than “kiki” and takete”, which are dry and jerky.
Let’s move on to the bass or treble aspect of the sound of “bouba” or “kiki”: there it is more difficult to hear. What we generally hear well is the main low or high pitch of the sound: it comes in particular from the vibration of the vocal cords for speech: our voice is more or less low or high pitched, we each have our own “fundamental frequency”. What the ear also hears are the bass-treble modulations of our voice according to the speech sounds that we pronounce. Our vocal cords vibrate and the sound is then modulated according to the shape of our mouth and the position of our tongue when we speak. It’s called the “spectral frequency” of sound. And it’s this precise bass-treble aspect, which is different between “bouba” and “kiki”, through the same voice.
The model we used first simulates how speech sounds are processed by the human ear. Then, for each speech sound, like “kiki”, our model extracts two cues that simulate some processing done by our brain. The first bass-treble index corresponds in our model to the balance between the low and high (spectral) frequencies of each sound.
The second index “calculated” by our brain is the continuity index, which mathematically corresponds to the difference in intensity between the loudest and quietest sound of each stimulus. Finally, this model combines these two indices to predict how much each speech sound should be perceived as rather round or rather pointed. These “roundness scores” from the model were then compared with those performed by nearly 400 adults (speaking different languages) for over 1,000 speech sounds. The model obtained scores very close to human judgments, which confirms that it is indeed these two indices (bass-treble and continuity) which are at the origin of the bouba-kiki effect.
We then turned to the field of the physics of objects and mathematics to, this time, analyze the characteristics of the sounds produced by everyday objects. We have shown that the sounds produced by round objects, when they roll on the ground, such as the speech sounds for “bouba” or “malouma”, are systematically more continuous and have a lower spectral frequency than the sounds produced by sharp objects of the same size. For continuity, that’s pretty good: a round, smooth rolling ball will tend to produce a continuous “whooooouuuuuu” sound, while a spiked ball is more likely to produce a discontinuous “tak tak tak” sound. tak”.
In video: A round ball produces a continuous sound
In video: A ball with spikes produces a discontinuous sound
For the bass-treble, it’s less easy to hear, but we were able to mathematically establish a relationship between the shape of an object and the spectral frequency of the sound produced by this object. More specifically, we have been able to demonstrate that, at an equivalent size, it is the perimeter of a shape (the path around the shape) that determines the spectral frequency of the sound produced by this object: the shorter the path, the more the sound that can be produced by this shape is serious, and vice versa. However, between a round shape and a pointed shape of equivalent size, the perimeter of a pointed shape is always greater.
It is because humans have integrated these physical properties of objects into language that “bouba” has become universally round. This finding is important because it shows that human language is not “sealed” from its environment. On the contrary, our language seems permeable to its environment, in particular to the characteristics of everyday objects.
Moreover, these results also have important consequences for the acquisition of language in children. Indeed, this shows that the language would not be learned independently of the rest of learning, but in constant interaction with other acquisitions. In other words, these results suggest that the sensitivity to certain physical properties of objects, which develops in the first years of children’s lives, may interact with the learning of their mother tongue.
Mathilde Fort, Lecturer in Psycholinguistics and Developmental Psychology, Grenoble-Alpes University (UGA), and Jean-Luc Schwartz, CNRS Research Director in Speech Processing and Cognitive Modeling, Grenoble Polytechnic Institute (Grenoble INP) .