It is an incredible accelerant in top-down 'theory driven' learning

Is it? People claim this but I really haven't seen any proof that it is true.