Here's how I understand it. Think of air molecules as little grains, because that's what atoms basically are. If you push forward with a cone, the molecules seek the path of least resistance. Since resistance will be strongest directly in front of the cone, they will start to push outward until they all are experiencing approximately the same resistance, hence the spherical radiation pattern. BUT, let's say you have a bunch of drivers above and below. The ability to move to the left and right is mostly unchanged, but for any molecules trying to go up or down, there is increased resistance above or below because the molecules there are *also* trying to move. They encounter resistance above and below and/or bounce off each other so they have to go either forward or sideways, not up or down. So, with one driver, imagine an inflatable ball that is having air pumped in and out at a rapid rate, the sphere grows and shrinks. But imagine a cylindrical "ball" capped by the floor and ceiling. The cylinder moves in and out but only in two dimensions, not three. A single driver is far enough away from the floor or ceiling that the radiation is spherical until it hits the limits of the floor. But with a line array, the floor/ceiling are hard limits, the sound can't even head in their direction effectivly. So the sound "slides" along the floor and ceiling in a parellel motion, it can't really reflect off of it. Of course, things aren't perfect, so you might get *some* lower amount of turbulence and bouncing, even maybe a tad of comb filtering, but it should be extremely minimal and only because a series of cones isn't a *true* line source, but is just acting as one. In theory, since the vibrations are only in two dimensions, there'd be zero energy heading up or down, only out into the room, hence the 3dB radiation, not 6dB (pi R loss instead of pi R squared). That's why the volume would remain unchanged from floor to ceiling as well. Did that make sense?
