The bouba/kiki effect is the classic demonstration of sound symbolism: shown a spiky shape and a round shape, people across languages assign “kiki” to the spikes and “bouba” to the curves. The effect is real, replicable, and confined — a curiosity about a few phoneme-shape pairings, interesting but marginal to how language actually works.

It is not marginal (arXiv:2603.17306). Sound-meaning associations are pervasive, extending far beyond shape into every semantic dimension measurable. Phonemes carry statistical information about meaning — not deterministic mappings, but reliable biases. Words with certain consonant clusters are disproportionately about danger. Words with certain vowel patterns are disproportionately about smallness. These associations are recoverable from text alone, without any acoustic input, which means they are embedded in the lexicon itself, not just in the phonetics.

The associations are not limited to onomatopoeia (words that sound like what they mean, like “buzz” or “splash”). They pervade the ordinary vocabulary. The sound of a word carries information about its meaning even when the word is not iconic, even when the connection is invisible to the speaker, even when the word was borrowed from another language. The fabric of language is woven with sound-meaning threads that speakers neither notice nor control.

This challenges the strong form of the arbitrariness thesis — Saussure’s claim that the relationship between sound and meaning is entirely conventional. The weak form (most sound-meaning relationships are conventional) survives. But the strong form (all such relationships are conventional) does not. There is a non-arbitrary substrate beneath the conventions, a statistical bias in how sounds map to meanings that transcends individual languages and persists across the lexicon.

The word carries a shadow of its meaning in its sound, and the shadow was there before the meaning was learned.