Comment by stelonix

I believe you're looking at this like a classic programmer and that's why you're thinking "bouba.jpg" and "bouba.wav" are two different concepts. I used to think of the bouba kiki issue from that POV too. A good (and imo better) modern analog to the human brain thought process is a LLM that can definitely match different inputs to a single concept. Also the vocal cords have very specific shapes in the vowel continuum and that alone would help explain why for humans it's simply intuitive.

There's also another comment mentioning how it works for consonants, which I believe helps explain how it might work for dogs.