Direction of Wording Effects in Balanced Scales

Abstract

This study investigated the degree to which statistical item selection is effective in reducing direction-of-wording effects in balanced affective measures developed from relatively small item pools. A pool of Likert-type loneliness items and their reversals was partitioned into balanced subsets by random and statistical item selection. Factor analyses of the alternate forms indicated that selection of items with high item-total correlations produced an instrument that showed the clearest direction-of-wording effects, with the positively and negatively keyed items loading on separate factors. The second factor was small in magnitude but clear and consistent. The clarity of the effects decreased as the internal consistency of the item subsets decreased. Possible explanations for these findings are discussed and directions for future research are suggested.