A COMPARATIVE TEST OF MAGNITUDE ESTIMATION AND PAIR-COMPARISON TREATMENT OF COMPLETE RANKS FOR SCALING A SMALL NUMBER OF EQUAL-INTERVAL FREQUENCY-RESPONSE ANCHORS

This study investigated whether previously-found differences in two sets of recommended five-point equal-interval response anchors could have been caused by scaling too many stimuli at one time. One set of recommended anchors, produced by magnitude estimation, was compared with a set produced by Thurstone Case III pair-comparison treatment of complete ranks. Subjects (N = 110) completed magnitude estimations and rankings of 8 frequency expressions. Few of the scale values produced by magnitude estimation differed significantly from the means obtained in previous studies or from the "ideal" values expected of exactly equal-interval anchors. However, this outcome was not true of the Case III results (they were seriously discrepant). Implications for future research are discussed.