Two related questions for the techies:
- How could the difference in ideal headphone and loudspeaker FR help us deduce the role of the room and reflections in preference?
- Is recorded music mastered primarily for loudspeakers or headphones, or does that not make any difference?
The first question is pretty complicated. I'm not sure what you mean or how to answer in the context of the discussion, so I'll start from scratch.
Ideally the transduction of an electrical signal to its acoustic counterpart would show no difference at all when measured. Frequency response should therefore be flat, i.e., show no deviations in amplitude. This should be the case for both loudspeakers and headphones, but it isn't because speakers play in a room and headphones are worn on your head.
When you talk about reflections, what you're referring to is time-domain behaviour and the way the acoustic signal propagates in its surroudings since there's a delay between the speaker firing and the wavefront hitting your eardrum (your body, really, since your torso, head and ear shape influence what ends up at your eardrum, with bass rattling the rest). Same thing for the headphones and IEMs but on a smaller scale. Headphone manufacturers have to account for the resonances of the entire circum/supra-aural cup, for example.
So again, ideally by the time the wavefront hits you, there should be no change in amplitude at all. But there will always be some difference because, by the time you hear it, given the speed of propagation, you're perceiving multiple slightly delayed versions of same wave. The hearing system works according to a time-based averaging function which will take into account the interference patterns of the multiple arriving wavefronts within a small window. Think of it as the time required for your ears to go through their own mechanical to electrical transduction.
The reasons why the preferred curve is different for headphones and loudspeaker is due to the physical difference in listening situations.
The discussion in this thread is about the effects of vertical reflections on listening with speakers. Assuming a boxlike geometry, vertical reflections from the floor and ceiling will arrive at your ears at the same time. This means whatever effect they have will be perceived the same way, and will be far less noticeable than that of left/right sidewall reflections, which will arrive at slightly different times and cause slightly different interference patterns at each ear.
Since loudspeakers aren't perfect transducers and incur energy losses, and are also of a certain size and shape, the discussion turned to the effects of multiple driver crossovers on vertical polarity, which will affect FR according to the height of your listening position. Smooth vertical polarity, like smooth horizontal polarity, will ensure minimal differences between different listening positions. Actually getting smooth polar response means solving an extremely difficult design problem.
Reflections in themselves are just multiplied versions of the direct sound from your speakers. They serve to increase the amount of apparent source locations. Given a well-designed speaker with flat FR and smooth off-axis response, the content of reflections won't differ significantly from the speaker's direct sound. Any abrupt on-axis or off-axis FR anomalies will emphasize certain areas of the spectrum and cause you to limit the listening position to a smaller range of locations or outright change what you end up hearing. It's worth thinking about the multiplication of apparent sources and the connection to reverberation and perceived spaciousness. Too many reflections and perceived sound will be indistinct. Too few, and the sound will shrink to pinpoints on a line situated between your speakers (only under anechoic conditions; no amount of acoustic panels in your living room will even come close to what this is like, although you can screw up balance by relying too much on one kind of absorber, like foam).
With headphones, the above set of issues occurs in the small distance between the driver and your ear. All of the same things happen, but on a small scale. The main thing to consider here is that preference is much harder to establish because of the individual dimensions of your ear. There's no one way to tailor a headphone such that it compensates perfectly for everyone's ears, although the goal is more or less clear: flat FR. Still, just as the circumstances have shrunk, so has the perceived sound shrunk to the distance between the two drivers. The best headphones have managed to minimize the effects at best, though the flipside is that many other problems that loudspeakers have are obviated.
With mastering, the workflow is pretty archaic. The basic assumption is that mastered music at end of the process will sound good on any medium. This is a roundabout way of saying that most mastering engineers will use both speakers and headphones, audition their mixes in a number of environments, and then make adjustments. There are very few studios that are interested in the acoustical differences between headphones and speakers. From a methodological, scientific perspective, the methods are random and haphazard. But then so are most listening circumstances, so it works out.
Some have created specifically binaural music for headphones. Chesky Records have created a set of filters to replicate the experience with speakers. It's not an active area by any means.
These days signal processing is cheap, and some of the issues of stereo sound and hardware limits have been addressed with DSP and software. When this area gets really good, hardware will likely only be required to hit a basic set of necessary parameters (like maximally flat FR) and the digital side will take care of the rest.