I still recall that when I purchased my first stereo cassette deck it came with two microphones. I remember placing them in the room about six inches apart and listening to the decks output via headphones, just monitoring people talking and general room noise/ambience. That sound was rather amazing, very realistic giving me my first impression of what binaural audio was. It makes me wonder if the dummy head is of any real value, or needed at all. I never got into purchasing or listening to any real binaural recordings though, except (as I recall) one of those flexible plastic records that came with an audio magazine. I don't remember it being that impressive though but I might be thinking of a sonic holography (flexible record) that I remember listening to and found equally unimpressive.
Like most here I prefer speaker to headphone listening, phones just don't image at all in the front. That being said headphones are excellent for monitoring recordings or late night listening. Electrostatics phones are the best sound wise, again great for monitoring or when doing restoration, every detail is crystal clear (almost exaggerated or magnified).
Rear speakers placed to the sides produce a similar effect as headphones, but with front speakers added the missing frontal quadrant is fully restored!
The ear is very sensitive to phase via headphones, for some it helps localise sound sources for others that doesn't seem to work, however I always find that any extra phasiness increases the spaciousness of the reproduced sound, even if that spaciousness is more like the ambience of a large room rather than distinct sound sources. Many laugh at the very idea of quad headphones, but I swear by them. The sound produced is much fuller than with conventional headphones, spread out much more (but still lacking anything from the front). Stereo listening through quad phones greatly benefits from enhancement via a simple quad decoder. My home brew decoder simply blends the rears out of phase and the fronts in phase. As you turn up the rear blend the sense of spaciousness greatly increases. The best effect is usually less than full blend. You then increase the front blend a bit to bring back the lost vocals. The overall effect is an enhancement of regular stereo, richer fuller even close to that of discrete quad via phones. JVC model 5944 has a phase switch for the back channels, when engaged adds much of that same type of spaciousness to stereo. And of course the Phase 2+2 headphones, with their ambience expanders and binauralators change the soundfield greatly but I can't say which setting is best they just all sound different, not necessarily realistic but still rather cool. Matrix decoders like the SQ full logic ones are not that effective over headphones. As you can imagine pumping back and forth between the front and rear headphone elements doesn't give much more than a regular stereo affect. A misadjusted decoder though is very easy to hear via phones, as so it is a useful tool for servicing/alignment purposes.