Andy - Tonality stays intact far above 20kHz, and plays some role in perception. I have auditioned mics / preamps flat to 50K and 40 and 30 and 20kHz. I along with most people can hear the difference. Go figure. That fact is not questioned in the recording community.
You brought up limit of eardrum vibration above 20kHz. That is a subtle question about which I know only a little. The eardrum is connected via the ear bones to the cochlea which is a hydraulic-mechanical receptor. The eardrum is an impedance matching device from the low-impedance air to the high-impedance cochlea. The drum (transformer) is locked to its output side (cochlea), which limits its upper frequency response to around 20kHz . . . unless partially decoupled in various ways, but lets call it locked.
But, that's only part of the story. The cilia - inner hairs in the cochlea which generate the auditory neurons - are embedded in a basilar membrane which changes its own charactristics to bias the cilia for differing functions, which include signal sampling (reading sequenced impulses such as every 100th peak, etc.) Now, the basilar membrane is known to respond to external stimuli including (but not limited to) inputs from the forehead and mastoid process skull bones, the phase relationships of those inputs being important for basillar bias functions. Hmmm.
We are barely scraping the surface here. The point is that the inner functioning of the ear mechanism itself is vastly complex, even before the neuronic energy is sent to the brain for processing. And, by the way, the left ear signal goes to the temporal lobe and on to cerebral processing, whereas the right ear signal goes directly to the auditory cortex to register its primal and emotional content and link to smell, taste, memory and the endocrine system. The multiplexed (left-right) signal begs to be reconciled for complete satisfaction.
The more I study and make connections, the more I am convinced that everything matters, and that dismissing some aspect is arrogant. Note that the engineering community's dismissal of phase importance is based primarily on ABX testing. Note that such testing may have less to do with the hearing under test than to other brain biases. Note that executive - judgement - choice functions take precedence over more subtle functions. I posit that the test set-up (distinguishing and choosing A or B with consequences to success-failure, professional reputation, etc.) may be usurping the brain-power required for pure auditory absorption of the musical (or just sonic) content. The ABX test is the scientific gold standard, but it is highly suspect to me.
