Stay tuned for our next paper, "Breaking the "Breaking the Bouba-Kiki Effect" Effect"

We find this effect is strong enough to override baseline sound-symbolic preferences, and speculate that it suggests a general cognitive principle of similarity alignment across modalities.

In sum, we report that adults prefer similar labels to refer to similar concepts. This is surprising, because things in the real world that are perceptually similar often have dissimilar labels, and vice versa (e.g., rat-mouse vs. house-mouse).

We found participants’ responses were more similar — both to the context label (toku) and to each other — when the creatures were perceptually similar, suggesting that the label-similarity effect generalizes to production.

We wondered if this effect might be an artifact of the experiment design (i.e. an alternative forced-choice task). So in Experiment 3, we asked participants to *generate* their own labels for these creatures.

Notably, the effect is strong enough to *reverse* canonical sound symbolic preferences. If participants hadn’t seen the toku before, they were more likely to say the spiky creature was the teki. But if they had seen the toku, this preference was flipped.

Participants preferred “teki” to map to the blobby creature when they had seen a similar-looking creature called a toku. This demonstrates a label-similarity effect in adults: people prefer similar labels mapping onto similar stimuli.

Then, we showed them two other creatures and asked them which was a teki.

Try it for yourself! In Experiments 1 and 2, we showed participants novel creatures like the following:

Here, we examine another source of non-arbitrariness: the similarity *structure* between labels and referents. For example, if you learn about a novel creature called a “toku”, what do you think a “teki” might look like?

But sometimes that mapping is non-arbitrary. One prominent example is sound symbolism (i.e. the bouba-kiki effect), where the linguistic form of the label “resembles” its referent. People have intuitions about some kinds of labels mapping onto certain shapes (e.g. “bouba” with round shapes).

What is the relationship between a label and the concept it refers to? Sometimes, the mapping between the two is arbitrary. For example, there is no more reason to call the thing that looks like a dog “dog” (English), as opposed to “chien” (French) or “gou” (Mandarin).

I was going to bring up Kaiser's work as well! Here's the abstract I was thinking of:
hsp2025.github.io/abstracts/31...