Models for the Adjustment of Rating Scales

In a category rating experiment using a graded series of squares we changed both stimulus range and frequency during the experimental session. Range-frequency theory (Parducci, 1965; Parducci & Perrett, 1971; Parducci & Wedell, 1986) predicts the same rate of adjustment to the new stimulus range and to the new stimulus frequencies. Its closest competitor, consistency theory (Haubensak, 1992a, 1992b), predicts same or different rates depending on the forgetting parameter, f. The experimental findings support the consistency model, for adjustment to the new stimulus frequencies was complete but adjustment to the new stimulus range was not. Range-frequency theory fails to predict such a dissociation. According to range-frequency theory (Parducci & Perrett, 1971), category ratings such as very large, large, medium, small, and very small, are governed by two principles, the range principle and the frequency principle. The range principle postulates that subjects judge each stimulus relative to its position between the endpoints of the stimulus range. The frequency principle postulates that each stimulus is judged according to its percentile rank. The two principles conflict when stimuli from different parts of the range are presented with unequal frequencies. The actual judgment is assumed to reflect a compromise between the two principles. Although the theory predicts mean judgments accurately, its implicit assumption that frequency effects are produced by the subjects’ tendency to use the available categories equally often, or to assign an equal number of stimuli to each category is open to question. According to consistency theory (Haubensak, 1992a), frequency effects arise from the subjects’ tendency to assign the first stimuli to the middle categories rather than to the extreme ones. These stimuli are likely to come from the more frequent ones if drawn at random. Thus, the subjects' scales shift towards the region of the most frequent stimuli. Because subjects, if not instructed otherwise, have a tendency to repeat the same responses to the same stimuli, the effect of the first judgments persists throughout the experiment. (Hence the name of the model.) Both theories provide a satisfactory account of category rating data but they differ in their predictions when both stimulus range and frequency change during an experimental session. The range-frequency model predicts the same rate of adjustment to the new stimulus range and to the new stimulus frequencies, whichever that rate will be. The consistency model in contrast leaves room for a dissociation between the two rates of adjustment. To test the models' predictions, we performed an experiment with an unexpected change of both stimulus range and frequency. The experiment is described in the following.