bull_dog: it looks to me like these parameters map out the squares within the 4D space within which the circles are drawn. In other words, they define where in the 4D space the samples should be taken (X1) and over how large an extent (X2-X1). Since this type of noise is deterministically pseudo-random, i.e. it always comes out exactly the same given the same input parameters, you have to alter (X1, Y1) to get different chunks of noise. You'd pick (X2, Y2) to be (X1, Y1) plus however much of the 4D space you want to consume. If you grab a lot of it, it's going to be more chaotic. If you grab a small portion, it's going to be smoother.
It seems to me that really these parameters could be (X1, Z1, Radius1) and (Y1, W1, Radius2), but I admit I can't think of an example where this would be all that useful.
Thanks JTippets for this post!