Soundproof,
Please read my post before taking from it conclusions that I have NOT presented in it .
Surely the comment is about the ears - I have often referred that our ears are not microphones, but the end of track.
For me the interesting part seemed : Spectral differences provided by the head-related transfer function (HRTF) are the main cues used for vertical localization.
If the way the ear perceives the height is partially based on alterations of the spectral content, not in directional issues, than may be microphones can convey this information if they are accurate enough. (my humble interpretation)
If you think that this is impossible (may be it is) please educate us.
BTW, the Cocktail Party Problem has nothing to do with the height issue - it was only the title of the paper, and I had to reference it . Next time I will call it CPP to avoid misinterpretations.
Well - Ambisonics tries to present information that the ear can interpret correctly, by setting up an array of speakers around the listeners, that reproduce acoustic information that has been collected through a similar array. The goal being to present our sense of hearing with a fuller simulation of the real soundstage.
It might be useful to actually read about the HRTF which you highlighted, and how that works:
http://en.wikipedia.org/wiki/Head-related_transfer_function
Binaural heads are an attempt to emulate how we hear - but there's a limitation, as each of us interprets space according to the tools at our disposal - where our ear shape, and the shape of the pinnae, and the amount of practice we have identifying source-locations, determine our degree of precision. Which is why you might get better localization information if the binaural head used to record had ears attached that actually were casts of your pinnae ...
Worth thinking about.