Discrimination and reliability in Q-sort personality descriptions.

Abstract

Using a 76-item Q sort set, 11 sorters made 4 sorts of each describing the personalities of each of 5 political figures. The reliabilities of both unforced and forced sorts were determined and the number of discriminations present in each unforced sort was computed. The heterogeneity of the distributions employed for the unforced sorts indicates that no particular form of distribution was consistently preferred by the sorters. It was found that as more discriminations were made in Q-sort situation, the test-retest reliability of the sort tended to increase. The data suggest that the maximum possible number of discriminations can be made among a set of items with no appreciable loss of reliability. The recommendation is therefore offered that a rectangular distribution be employed for forced Q-sorts.".