What's the most important speaker attribute...and


why?
That ineffable "musicality" is IT for me. The ability to reproduce human voices that sound, well, palpably human and distinct (not all raspy, if male, or sibilant, if female).
Instruments sound like they do "live."
In a nutshell, you don't gush to your friends, "Hey,I keep hearing new things all the time," but rather, " Billie sounded soooo beautiful," or, " the counter tenor blending with the bass---magical!"
If it doesn't sound beautiful, if the instruments and voices aren't naturally seductive---you're not listening to music anymore, no matter how impressive the stage width, depth, or heighth, or how dynamic and uncompressed it is. (My system is currently in large scale flux, from front-end to speakers; I just went to SF EAIIs [from Piega C8LTDs], Pass Labs electrs, Audio Note cdp, but I'm swapping out passive preamps this week [Preeminence & con-john pfr] seeing if they make significant improvements).
128x128jayme
Tonal accuracy, as most say in one way or another, and especially in the midrange, as Scott says, is certainly the first thing to get right. For me, though, low distortion is more important. Almost all speakers make a lot of noise and have much higher levels of distortion than electronic components. Driving the distortion level down helps to make voices sound like real life by eliminating the sonic products that you wouldnt hear in an unamplified live performance, like, e.g., material coloration from almost all cone drivers.
Hi, Paul:

Our posts are not in conflict with each other -- accurate mid-range reproduction means, ipso facto, that the distortion is very low. Otherwise, the mid-range ain't accurate...

You are quite correct that distortion is far higher in speakers than other audio components, but since speakers are transducers, there will always be some inherent non-linearity. I personally think that time- and phase-accurate speakers sound better, assuming their drivers are of high quality, than speakers that are not, and they often have the virtue of lower distortion as well.
Well, I didnt think I was disagreeing with you in any material respect, but it depends on how you define accuracy. It is possible to measure a speaker's frequency response, and get a flat reponse, but still have high levels of distortion. Of course, you can include low distortion in your definiton of of accuracy, but I think it best to keep the concepts distinct.

Take a look at the measurements of the Mirage MRM-1 at soundstage.com, equipment archives, loudspeakers,and scroll down, or just try this url

http://www.soundstagemagazine.com/measurements/mirage_mrm1/

Not real flat, but not a bad frequency response, yet, to me, unacceptable levels of noise and distortion in the bass (where almost all speakers have a problem) and midrange. No real correlation with the FR, they are separate measurements.

I know from personal experience that when I play a 50hz tone through my Harbeths, I hear a 50 hz tone coming out of them. No audible distortion. But when I play the same tone through a pair of AE Aegis Ones, a pretty good little $300/pair speaker, though the spl meter reads about the same, I hear something that sounds more like 58-60 hz because the output includes the fundamental plus some doubling and other distortion components.

The reason I say low distortion is more important is that we adjust to different tonal balances in different venues all the time, so moderate FR inaccuracies are not always troubling (some are), but distortion is always annoying. A frequency response inaccuracy means the notes are played a little too loud or a little too soft (if in the fundamentals part of the midrange). You may not notice, or may not care. But distortion means the wrong note.
Sdcampbell pretty much nailed it...you could have the most liquid top end and the tightest, deep bass known to man...but if the midrange isnt seamless...lotta work for nothing...which is why I assume that in 2-ways for example...the trend towards higher x-overs...3khz and up...is becoming more common...would also agree...the hi end has gravitated towards a more mellow, laid back midrange presentation...which isnt adherently bad...basically what Vandersteen has pioneered since the 70s...albeit without the clever boxless design, phase integrity, and value the Vandies represent(maybe this insnt a good comparison!)...at any rate...the "spatial" midrange production all too common tends to be generated from the speakers and less from the recording itself...again...not a bad thing...but coupled with an exaggerated "hot" treble...can sound artifical over time...just an observation...also...Scott...thanks to your informative responses to my posts and many others....always enjoy your insight...
Many good comments above. I like the description of one of the masters of our trade, Mile Nestorovic. He would always say "The speakers should disappear!" For me this is a requirement. I think that was Jayme's point at the start.

What physical attributes or designs will accomplish this? There are many answers. Major considerations, that often involve compromises, are the bass lower frequency limit and the maximum acoustic level (loudness, room size). (this brings us to what music are we playing, and how loud? - as mentioned by Eldartford).

Interesting that no one has mentioned "sound stage" or "imaging" in those terms. But what I like is for this 3-D hologram to appear at the "front" of the room, and I forget that there are speakers, and the music transports me to another dimension.... This is an illusion, of course, and it requires my participation. Very subjective stuff.

This illusion can be convincing when the speakers reproduce the voice and musical instrument sounds, as described by others above; and also do not call attention to themselves due to distortions, noises or introduced sounds (such as resonances), or compression of dynamics.

Cheers