That's probably as good an explanation as anyone could give without delving deeply into technical details.
At this point, mantle dynamics are still not very well-understood. We know that the plates are moving, we know that the mantle is convecting and that there is some connection between the two, but there's still lots of disagreement about how much of the mantle is convecting, how large the convection cells are, and how exactly mantle convection relates to plate motion and other features (i.e. hot spots) that we see on the surface. The problem is mainly one of access - seismic imagery only penetrates so far, and resolution drops dramatically with depth, so there are large areas of the mantle that are almost completely unknown. And of course, we have no way to directly access the mantle. Things would be so much easier if we just had that machine from The Core...