I think Adina Levin is underestimating Jakob Nielsen.
It's probably only a rough approximation of what's going on, but his collection of points adds up to a theory which pretty much does explain the "mystery" of why 3D interfaces haven't (and in their current form, won't) take off.
And it's a theory which you can use to make novel predictions rather than just fitting previous data. So the accusation that it's "backfilling" what we already know is a bit unfair.
Despite being a self-avowed "text person", I'd point out that clearly our knowledge and knowledge tools use a lot of graphics and 2D spatial reasoning (graphs, GUIs, typography and design, diagrams, facial expressions etc.) 2D space helps simplify and understand very well. It's only 3D which we avoid as confusing.
Games are the exception, of course, because they aren't trying to simplify. They're *trying* to make things more complex and challenging. But wherever we're trying to improve the efficiency / productivity of information work, we'll be trying to squash it into 2D.