The ear localizes sound by two mechanisms: Arrival time and intensity. Arrival time is a bit more dominant below roughly 1.5 kHz and intensity a bit more dominant above, but both are important.
If you are seated off to one side of the centerline, the output from the speaker on that side will of course have the "winning" arrival time. Usually it also wins intensity, so the center image is typically pulled over to the near side rather dramatically.
The ideal would be a set of speakers where as you move off to one side, the speaker on the other side actually gets louder. This way one speaker wins arrival time and the other wins intensity. We can do this by using speakers that are louder on-axis than off-axis (but still have smooth off-axis response) and toeing them in severely, so that the axes criss-cross in front of the listening position. This way as you move off-axis to one side, you're moving more on-axis of the farther (later-arrival-time) speaker so that it's actually getting louder. This is the set-up I usually use at audio shows (such as the one in Vegas I just got back from).
A fairly wide-pattern line source speaker will also give you decent soundstaging over a wide listening area because the SPL falls off more slowly with distance than from a point source, and therefore the typical loudness advantage of the closer speaker isn't as great as with a point source speaker.
A very wide-pattern point source speaker can also work well, but the center vocalist is more likely to shift towards the near speaker than with a wide-pattern line source or with the crossfiring arrangement described above. This is because the loudness discrepancy is usually greater than either of the first two cases.
Very wide pattern line source speakers can be set up actually facing each other and thus create a very wide sweet spot. Beveridge lensed electrostats were designed to be used this way, and the Maggie 3.6 and 20.1 with their extremely narrow line source ribbon tweeters can be used this way.
Note that as you take steps to improve the soundstaging over a wide listening area, you may be degrading the imaging a bit for the listener seated exactly in the middle of the sweet spot.
Duke
