Although the effect of item wording on rating responses has long been acknowledged, much less attention has been directed toward the effect of rating anchors. This study tested assumptions that packing or skewing the response options of a scale reduces the common measurement problems of leniency and range restriction in instructor ratings. Packing the scale was found to reduce leniency, although this finding is dependent on the scoring procedure used. Packing the scale had no effect on the restriction of range problem. Further analyses determined that both the label and position of the label on the response scale influence respondent selection of a category, with position the more dominant. Similar results were found for both global and specific items.Rating scales are one of the most frequently used tools to measure attitudes and opinions. Research on rating scales has addressed issues of item wording, item ordering, and the optimal number of response options (Masters, 1974;Ory, 1982). The selection of response scale categories, however, appears so obvious that often little attention is paid to their wording. Despite this appearance of simplicity, various authors have suggested that the labels assigned to scale categories do influence raters. It has been speculated that leniency of ratings can be reduced by category labels (Guilford, 1936;Symonds, 1931), that the range of ratings across stimuli is influenced by the choice of category labels (Frisbie & Brandenburg, 1979;Lam & Klockars, 1982), and that the equal interval properties of the response scale depend on an appropriate choice of labels (Spector, 1976).The issues of range restriction and leniency are particularly potent inRequests for reprints should be sent to Trudy Dunham, Department of Educational Psychology, University of Minnesota, 178 Pillsbury Drive SE, Minneapolis, MN 55455.
DUNHAM AND DAVISONeducation. Instructor ratings are negatively skewed and differences in mean ratings across instructors are often relatively small (Cashin, 1985(Cashin, , 1987. Range restriction enhances the difficulty of distinguishing between instructors for decisions such as teaching awards, tenure and promotion, and merit pay. Leniency can be a problem to the extent that it reduces the probability of identification and rectification of instructional problems.No one, to our knowledge, has investigated whether the labels assigned to scale categories do influence the variability in the resulting ratings. The effect of labels on leniency, defined as the average rating assigned to instructors, was studied by Frisbie and Brandenburg (1979). They found that when every response category was labeled, the mean rating of instructors was lower than when only the end points were labeled. Lam and Klockars (1982) suggested that Frisbie and Brandenburg found less leniency when they labeled every scale category because their labels were a positively packed set of scale anchors.Scale anchors are often assumed to be balanced, that is, to be composed of labels forming an equal interval contin...