For many thousands of years, we and our ancestors have been dealing with three dimensional objects in space. Being able to do this well and quickly no doubt often meant the difference between making a meal of a dangerous animal or it making a meal of you. We're well adapted to living in a three dimensional spatial world.
In terms of biology, we're no different than humans who lived and died before personal computers existed. We're still wired to intuitively respond to three dimensional objects. In terms of GUIs, I've said before that our perception of skeuomorphic, non-flat UIs is hardware-accelerated in our brains. Effects like shading and shadows that give the illusion of depth allow us to instantly recognize that a window is distinct from the background because it has a shadow or that buttons are meant to be pressed because they look like actual buttons. Flat interfaces that don't look like anything we're wired to instantly recognize require more cognitive processing for us to figure out what UI elements are actionable, and that takes a person's attention away from the task at hand for a brief moment. It adds up... I think the Reg article some time ago said that the subjects using flat interfaces were 22% slower than those using skeuomorphic ones, according to the study they were reporting upon. A particularly bad flat interface could be still worse if you have to hunt around to see what is an active UI element.
Looking modern or what some people think is attractive (I think flat interfaces are ugly) is not a good reason to make user interfaces (or themes that define their appearance) that are slower and less intuitive. Function is beauty.