Breaking Bouba-Kiki: The Effect of Label Similarity on Naming Perceptually Similar Referents

When we encounter a novel label, what, if anything, do we assume about the appearance of its referent? Prior work has shown that sound-shape correspondences (e.g., rounded sounds pairing with rounded shapes) can guide how people map novel labels to objects. Here we examine whether people also have a bias to map labels that are similar in form onto perceptually similar concepts (and vice versa). Across three experiments, we show bidirectional evidence of this. In Experiment 1, adults were first introduced to a novel label–referent pairing (e.g., “This is a toku.” written above an image of an alien species) and then selected which of two new species best fit a second, novel label. When the new label was similar to the initial one, e.g., teki, participants reliably mapped it onto the species that looked more similar to the first one. In Experiment 2, we show that this label similarity effect is robust enough to override pre-existing sound-shape preferences. And in Experiment 3, we demonstrate that the effect also operates in production: after being introduced to a novel creature, participants generated more similar labels when asked to label a perceptually similar, but distinct creature. We propose a bias to assume that similarity in conceptual space maps onto similarity in lexical space when inferring new label-referent mappings, and we speculate that this bias may be an instance of a broader principle of similarity-based mapping across domains. Finally, we discuss potential implications for language learning and use more broadly.