It means that it is calculated by sending multiple rays from many random points instead of one. Rays starts randomly from circular area located in camera position and all crosses in focal length. This method gives most realistic results. Quality doesn't depend on depth. It depends only on amount of rays calculated from each image point. More samples = better quality. This method needs a lot of computing power because whole image (or every image point) has to be calculated hundreds times. Alternative way of calculating DOF is to use z-buffer and calculate effect after rendering image. It's faster but looks much worse.