Depth of field is the area in front of a camera where objects are in focus. Objects closer or farther are out of focus in some proportional manner.
In some tests I'd define a 3D camera with a specific depth of field that matched the distance to the object I wanted to be in focus. Say I had an array of view ports onto documents, processes, or images, the one selected would come forward to the focal distance and be in sharp focus. I tried using the effect both coupled and decoupled with actual distance. To me, simply blurring the element without having the effect being tied to distance did not seem nearly as effective. I also played around with having selected, but not active objects coming forward of focus.
I also did some tests with "constellations" of documents connected by lines. The lines could indicate a path, a hierarchy, connectiveness, or dependency. The browser would allow you to spin the viewpoint, zoom into and out of groupings. Depth of field here, as well as aerial perspective (greying and blueing far away objects) made for a very nice effect, and seriously improved navigational awareness.
These were all animations, not actual interactive demos. Very exploratory in nature. Never got enough interest to go deeper into it.