[Edit: this perhaps should have been "Changing semantics based on visual attributes is not"].