Wow. Interesting discussion. Sorry to be getting in late on this one. I've skimmed all the previous, but forgive me if some of my thoughts have already been said.
There does seem to be some cyclic redundancy to the discussion. Tryg, I appreciate what you are expressing but I earnestly urge you to focus not on the sound system per-se, but on the mixing challenges faced by motion picture soundtrack artists. 3 speakers across the front is enough for spacious audio, but let me hazard an explanation on why what you are experiencing is less than that.
I have a stereo recording of Miles Davis who's music I really like, but who's mix I detest. The sound engineer elected to put the bass guitar exclusively in the left speaker and the drums exclusively in the right (all other instruments image in the center). This makes the two instruments in question sound less lifelike than if they imaged even as half left and right respectively. From all your posts above, I suspect that a similar lack of liveliness is what you dislike about dialogue from a single speaker.
Does that mean all Hollywood sound engineers are mixing 'bad' soundtracks? Heck no! Remember (and this is key) they are faced with the challenge of mixing a soundtrack that will match the on-screen image AND do so regardless of where a person is sitting (relative to center). So while it would be possible to mix dialogue as a phantom center image from the left and right speakers, it just doesn't 'work' for anyone seated even slightly off center.
No matter how many speakers we put in the front soundstage, the goal of coordinating sound to visual cues will always result in sound from just one speaker, IF we want to accommodate people off center. Sony's cinema-only sound system SDDS (of which I am personally a fan) can have 5 screen channels. Even still, dialogue will often come from one of the five speakers not only to 'lock' the dialogue to the exact left-to-right position of the visual that is producing the sound, but also to ensure that the same sound (in terms of tonality) is heard by all. If they were to mix a line of dialogue as a phantom image between center and left (for example), people off to the right will get the comb filter effect which may affect intelligibility. This is one of the prime reasons to use 5 screen channels: more creative possibilities for horizontal positioning of sound with less phantom imaging. As such, you get pin-point and consistent sound, no matter where you are seated in the horizontal plane. For this and many other reasons, 8 channel SDDS (when used correctly) is really something special.
Enough about that...
The posts about the comb filtering artifact of two speakers used for the same channel are not without merit: it is a very real problem. To answer question 1 of your above post, most center speakers based on the ubiquitous mid-tweet-mid design do not eliminate the comb filtering effect (in fact, they exemplify it) BUT, the closer the drivers are sandwiched together, the farther off center you have to be before it becomes an real issue.
Even so, several upper end receivers have two speaker outputs for the center channel. They do this for people who have HUGE screens and can't put the speaker behind it. If the screen is big enough, a center speaker above or below may be so far away from the visual cue, that it strikes our brains as 'wrong'. In that situation it MAY be preferable to put a center above AND below the screen so that the sound images in the center of the screen for most every one in the room, without too much in the way of artifacts (unless you have a balcony http://www.avsforum.com/ubb/smile.gif
Anyway, based on what you've said, I'd like to know if you have auditioned the Mirage OM-C2 center speaker. I reviewed it sometime ago and was substantially impressed (it is a really boss speaker). I mention it here because I think it's very unique and exclusive Omni-polar design will give you the sort of sound you crave in a single speaker.
Secrets of Home Theater and High Fidelity http://www.hometheaterhifi.com
[This message has been edited by Brian Florian (edited 12-01-2000).]