On the reformulation of the finite element method