as regards Fmax in humans, I read an article in the mag that turned into "sound on sound", back in the early nineties, in which rupert neve himself told this story of a desk that his company built which developed a strange fault. he & the complaining engineer could quite clearly distinguish between two supposedly identical channel strips on this desk- one sounded more "open" than the other. daft word for it, I agree, but that was the best they could do.
upon examination, the "open" strip was found to have some defective soldering which resulted in a low-pass filter (in fact, one end of an inductor) being disconnected. mr neve reckoned that the difference should not have been audible, based on the received wisdom regarding Fmax in humans (even- perhaps especially- audio engineers).
this led him to conduct some blind tests during which he played sines & squares at 8 or 9 KHz..... as has been pointed out, the first distinguishing feature of the latter waveform is the third harmonic at 20-odd KHz, which ought not to be appreciable in it's own right at this frequency. but it was, in the majority of these tests.
I've tried it myself with some grado headphones. I'm not convinced I've eliminated other factors- IMD, level changes, the slight pitch change you sometimes get even with high end oscillators when changing frequency.....
but I do think that the assumptions about Fmax which determined- for instance- the sample rate for CD, DAT & so forth, were naive, & somewhat influenced by the economics of realising the technology at that time.