skip to main content skip to footer

Ability to Make Consistent Judgments in Paired Comparisons as a Predictor of Academic Achievement (Preliminary Report)

Gulliksen, Harold O.; Saunders, David R.; Tucker, Ledyard R
Publication Year:
Report Number:
ETS Research Memorandum
Document Type:
Page Count:
Subject/Key Words:
Attitude Measures, Grade Prediction, Paired Comparisons, Predictive Validity, Response Style (Tests), Test Reliability


The purpose of this memorandum is to direct attention to some recent results on tests of response consistency in making paired comparisons. The evidence obtained so far strongly suggests the possibility of increasing many of our better multiple correlations with academic grades by .10 to .15, using a paper-and-pencil test lasting 30 to 60 minutes. The relation between the number of circular triads in a set of judgments given by a student and his grade-average has been studied in two quite different groups--one from Temple University and one from West Point. For both groups it was found that the relation between response consistency and grades was curvilinear, the best grades being made by the students who scored at about the 75th or 80th percentile of consistency. The overall linear correlation between consistency and grades worked out to about .30 to .35 However, this correlation became zero when only the more consistent half of the population was considered and rose to .40, despite the restriction of range, when the less consistent half of the population was considered. The overall correlation between the response-consistency scores and such predictors of academic achievement as the ACE or the WPAT was substantially zero.

Read More