Determining the Lenqth of Multiple Choice Criterion-Referenced Tests when an Answer-Until-Correct Scoring Procedure is Used

Abstract

When determining the length of a criterion-referenced test, results in van den Brink and Koele (1980) and Wilcox (1980a, 1980b) indicate that the problem of guessing might be more serious than may have been expected. Recently, however, a new method of scoring tests was proposed that corrects for guessing without assuming guessing is at random. Moreover, empirical investigations suggest that the underlying assumptions of the new scoring procedure will frequently hold. This paper indicates how test length might be determined when the new scoring procedure is used. The results indicate that test length might be substantially reduced when the new scoring rule can be applied.