So, what he's suggesting is . . . you take your larger goals, and then you break them up into a sequence of smaller subgoals?
Sort of like a hierarchy?
Well... kinda. As pyite wrote above, the main idea behind Allens framework is that you should not have to think about a project or thing you want accomplished more than necessary. By making sure that your subgoals are actual physical actions that you need to do to move the project further along you dont have to "rethink" this step every time you read your todo-list, instead you just "do".
Say for instance that you have a ToDo-list with an item such as "plan new-years party", with the first subgoal being "find location". This is not a physical action, and everytime you see the list you now need to rethink how you are going to decide on one. Instead, Allen wants you to write something like "Call Bob to get ideas for party locations". Now when you look at your list (sorted under @phone so that you can do all phone-related stuff at the same time) you know exactly what to do next. Think hard once is the general idea.
I have a theory that it's impossible to prove anything, but I can't prove it.