I saw Avatar in 3D and while the effects were fun to watch, I thought that the expressiveness of the depth-of-field shots was actually reduced.
In 2D cinema, the viewer's eyes adapt to the shot that the director intended, which often involves focusing on the foreground or the background... once in a while a shot puts both in focus, but it's rare.
The visual cue of depth of field is why cinematographers continue to use it even though it's been technically feasible for a long time to simply make all objects in the shot appear in focus.
In Avatar, the critters zoomed toward the audience and the effect was immersive, but I thought many of the close-in shots lacked the intimacy that good 2D shots can often attain.
3D vision is a perceptual use of depth of field to heighten the brain's ability to discern objects. There is similarly no reason why the human visual system couldn't have evolved to perceive objects in focus over a wider range of distances.
In fact, the human perception of a constant, detailed visual field is an illusion that the brain stitches together from the data collected by the narrow beam of detailed focus directly in front of our gaze.
So I think doing accurate 3D requires simulating the way in which the brain picks out objects at different depths, but this can't be done simply by offering one depth of field for all viewers, since our eyes can't bounce back and forth alternately focusing a near and far object and creating the increased acuity we get in normal vision.
3D film is unreal in that it requires us to suspend our normal method of discerning space, and watching it is something we must learn to do, just as we must learn to interpret 2D depth of field as both an attentional and spatial cue.