Decomposing the focus effect: Evidence from reading