Toward a Graded Model of English Phonology