The noise scales itself by multiplying the position input by the inverse (reciprocal) of the scale. When the position input has a large value (i.e. it is far from 0,0,0), small changes to the scale have large absolute changes to the position because it is a multiplication effect. Therefore whatever function you use has more effect on the noise shape at larger distances from the origin, regardless of what function you use.
It is possible to cancel out this effect by dividing part of your function by the distance from the origin, but you need to make sure that there is also a constant component to your scale function which you don't apply the divide to. Difficult to explain without providing an example... but hopefully you can see what I mean.
Matt