My honours thesis was in the field of psycholinguistics. During the research for that, there was a period when I spent three months listening to a couple of four-word sentences repeated thousands of times, examining them for speech production errors (I was looking at stuttering and methods to artificially induce it).
Often, when trying to characterise a speech error, I would isolate and loop a single word or phoneme. Do this enough, and you start to notice something very weird: if you loop an isolated phoneme, it very rapidly stops sounding like human speech at all.
I don’t just mean that it was unintelligible nonsense. Looped phonemes turn into clicks and hisses and whistles of the sort that you’d expect from an insectoid alien in a SF film.
It’s a startlingly clear demonstration of just how much of human speech perception is psychologically constructed. The words you hear have very little in common with the actual physical soundwave.