Recently I happened to stumble across a fascinating 1941 article called "How I Broke the Rules in Citizen Kane," written by Gregg Toland, the film's cinematographer. One of the "rules" (or Hollywood conventions, as he prefers to think of them) he broke was to use mostly large apertures producing shallow depth of field. Typical apertures at the time were between f2.5 and f3.2; he would regularly use f8. (How he managed to do this is an interesting story in and of it self, and makes the article worth reading.) This let him have much larger depth of field; with wider-angle lenses everything from the foreground to the background was in focus.
To this day this is an uncommon technique in feature films. In fact, we've become so accustomed to a shallow depth of field in film productions that in the last decade or so animated works and video games now emulate this. There's no need for or advantage to shallow depth of field, since in these circumstances there's generally no lens involved; it says a lot that people would go to extra work to emulate what has usually been considered a lens deficiency.
This reminded me of seeing a shallow depth of field in an unusual circumstance the other day that really made it jump out. At consumer electronics stores in Japan lately they now have demonstrations of 3D video systems of the sort that use a pair of LCD glasses. Watching an animated feature playing on one, I noticed that it used the "shallow depth of field" technique. However, this produced a rather odd sensation in a truly 3D image. The image felt more realistic than a standard 2D image in that I had a much stronger sense of being to move my point of focus between foreground and background objects and know just where they were in the 3D space in front of me. But as I did this, I immediately found myself discovering that what was in focus didn't shift with what I was looking at, as it does in the real world. This being both unreal and unconventional gave me quite a jolt.
I wonder where we'll go with 3D, particularly when recording through a lens. Will 3D productions widen the depth of field to avoid the unreal jolt? Or will they stick with the narrow depth of field and let that become an almost unquestioned convention in 3D productions as it is in 2D ones?