In on-shell (OS) scheme, one of the renormalization conditions is that the propagator, say, a scalar theory   

$$\frac{1}{p^2+m^2-\Sigma(p^2)-i\epsilon}$$ 

must have a unit residue at the pole of physical mass $p^2=-m^2$. Some textbooks say this is to make sure the propagator behaves like a free field propagator near the pole. But why?