Prior categorization can produce more efficient ordering

Abstract

We hypothesized that prior categorization of an attribute might sometimes facilitate the ordering of stimuli, particularly when the number of stimuli is relatively large or some of them need to be remembered. In four experiments, participants were required to rank order either numbers or line lengths either using a selection sort or following prior categorization. The results indicated that prior categorization produced faster sorting when the number of stimuli to be ordered was relatively large and when the stimuli were not continuously visible.

Keywords:

• Categorization
• Category rating
• Ordering
• Sorting

Notes

¹There are many different ways to measure the accuracy of orderings. We chose Kendall's tau because it is relatively easy to compare across different numbers of items to be ordered (Experiment 2). Other choices could be justified. However, the choice does not appear to be critical here because no experiment featured much difference in accuracy between conditions. A parallel analysis using Spearman's rho coefficients produced identical results on tests of significance, although rho coefficients were generally higher.

²We also conducted an experiment in which respondents ordered sets of 40 visible numbers—the set size was chosen as the approximate cross-over point from Figure 3—using either prior categorization or selection sorting. We manipulated whether the numbers were rectangularly distributed or were subject to a marked positive (many small numbers) or negative (many large numbers) skew. The categorizing sort was relatively slower than the selection sort for the positively skewed distribution compared to the other two distributions but the effect was not large.