Do easy topics predict effectiveness better than difficult topics?