Reassessing scale effects on differential case marking