Personally I think the problem we have regarding lack of realism in a playback-system comes mainly from these things;
- Recording
- Acoustics
- Dispersion
- Capacity
To my ears I find that good speakers with controlled on/off-axis behavior sounds remarkably similar in near-field where the room and dispersion characteristics matters less. That's expected due to the fact that we then emphasize the frequency response and minimize the time-related issues that comes from the room/speaker interaction. If we continue to explore the realism aspect, I'd put money on the notion that a real-life trumpet in a anechoic chamber would sound the same as a loudspeaker with equal frequency response and enough capacity in that anechoic chamber, if the recording allowed for it.
In a room where the dispersion characteristics will affect how everything sounds there's no way to make a speaker sound like a real trumpet/piano/guitar. If you reduce the room's impact on sound, you (should) get closer to the real thing.
Of course, the easiest way to perform this experiment is to take both the instrument and the speaker outside and try.
I seem to remember Dutch & Dutch taking the speakers and some musicians to an anechoic chamber to compare in a blind test, but don't know how that turned out.
@Martijn Mensink ?
At a show we did a blind live vs. recorded demo. We recorded several instruments in the anechoic chamber and played them back over a single 8c. The same musicians were present in the room and played the same piece or motif live in the room. Very interesting demo and a lot of people were fooled.
In the end though, I think there's no way to fool a trained ear with a demo like this. In line with what you're saying, every instrument has its own radiation pattern. Where do you place the microphone to accurately capture its timbre? Every loudspeaker has its own directivity pattern, which differs from the instrument. If there are any reflections in either the recording space or the playback space you'll have incongruent cues.
I haven't read the rest of this thread and the current subject seems a bit off-topic, but I'd be happy to share my view on realism. If you wish to create a really convincing illusion that you're listening to the real thing, you'll have to look at technologies such as binaural recording and playback, higher-order ambisonics and BACCH. Remarkably realistic illusions are possible. Just not with traditional stereo in a normal living room.
With respect to the importance of time-domain behavior my position may surprise some people. In our Dutch & Dutch 8c - a DSP empowered active loudspeaker - I linearized the phase response by means of an FIR filter. The filter fixes the phase-distortion caused by the two 4rd order Linkwitz-Riley filters at 100 and 1250 Hz. From a theoretical point of view flat phase response is better than a rotating phase response. We could linearize it without paying any real price, so I did it.
If you look at the science, it's not so clear-cut that normal phase-rotation caused by a 4rd order crossover is audible, or if it's audible if it's a serious problem. There are indications that it might be audible, but probably only under controlled conditions with very specific signals. Like I said, we did it because we could. Not because it's a huge improvement in sound quality. The impulse response looks very nice though!
On the other hand, many of our customers have reported to me that the 8c sounds better in linear phase mode. Those reports come primarily from our mixing and mastering customers, who listen in relatively dead studio environments. These have all been sighted tests, but these people are trained listeners who usually have very good sounding systems in good rooms. Therefore I think there's probably some truth in these reports. Having said that, I personally find it quite difficult to hear a difference. I've been able to tell a difference in a single blind listening test on some test signals we made a couple of years ago, but only under more or less ideal conditions. Nothing statistically significant with music.
At the time I was a pretty well-trained listener, but I was primarily trained in timbre and artefacts. Those were things I was trying to perfect with the 8c back then. I never really trained myself to hear phase distortion. Today I'm sure I'd fair worse than back then, because I haven't really trained my ears much lately. Listening is a skill you lose if you don't keep practicing. When our next speaker goes into the development phase I'll get back to training again.