I think people confuse the term hiearchial regression, because different statisticans use it for totally different methods. One form is when (normally in the context of OLS) you add variables in blocs based on a theory of importance. At each stage you are testing if the new variables add signficance to the model. The second form of hiearchical regression (called multiway by many) analyzes regression in the context of nested data generating a series of paramters at different levels of analysis. Lowerer order parameters explain variation within higher level parameters.

They are totally different methods, but the same term is used to describe both.