Consider, inside the limits of the array's spatial sampling, you can create any shape of wavefront you want, given enough control over each individual driver input. For instance if you take a flat array and you set it up so that the time delays are such that the wavefront emerges in a circular shape, you have a point source (i.e. omni, most likely over only the front half of the array) pattern.
If you drive them all such that the impulses line up at 50 degrees, you have a far field plane wave that points to 50 degrees.
It's much like FIR filter design in that the delays and weights for each driver (including frequency weighting for constant beam widths across frequency) allow you to design a pattern you want.
For near field (under 5 times the height of the array) you can also do near-field focusing, in which the level is maximized at one point (or line) in space, and is less energetic both nearer and farther. I've done this myself by making a focus at 1.8 meters, and putting a lower level diffuse (i.e. reverb plus other stuff) into this driver, so you hear the "far away" effect until you're at about 2.2 meters away, then the source moves to RIGHT THERE ON YOUR NOSE (quickly over about 30-40 cm) and then back AWAY as you get closer.
This is, note, only true for an array with control over delay, gain, and frequency weighting per driver, and careful design. You can get something approximating (for straight in front of the array) a properly sampled array when the drivers are below 1/2 wavelength. Over 1/2 wavelength lobes happen. If you're shooting end-on in an array it's 1/4 wavelength. (any delay effectively moves a driver (or mike, works both ways) directly away from the point being observed, regardless of the angle of the array. This complicates aiming the array to some extent. Just more math.
I'm not trying to be mysterious here, there's actually a crapload of math involved, and I do mean a crapload.
A lot of this is rather counterintuitive, which is why I hesitate to try to explain thoroughly. Because of the differences in wavelength between bass and treble, the effects are very, very different than at radar frequencies, for instance.