On Representing Evidence