A reexamination of the stress erasure convention and Spanish stress