Hill compared three questionnaires (all items positively worded, all items negatively worded, and mixed) on internal consistency and means. Although they used only 10 items in their analysis, they found that negatively worded items, when mixed with directly or positively worded items, provided lower internal consistency and mean scores that were significantly different across the three types of questionnaires. They concluded that it might not be advisable to use negatively worded items for the control of acquiescence response bias because such a practice impaired the response accuracy. Chamberlain and Cummings (1984) compared reliability for scores on two forms of a course evaluation instrument. They found that score reliability was higher when all positively worded items were used.