A THEORETICAL FRAMEWORK FOR THE STUDY OF ITEM DIFFICULTY AND DISCRIMINATION

Traditionally, test item difficulty is a statistical concept, defined in terms of the performance of examinees rather than in terms of intrinsic properties of the item itself. A clearer understanding of the association between item properties and examinee performance, however, would result in numerous benefits, including better prediction and control of item difficulty in the test development process and enhanced construct validity of the test. As a first step toward the goal of achieving such understanding, a theoretical framework is delineated, drawing on both measurement concepts and concepts drawn from cognitive psychology and personality theory. In this formulation, the difficulty of an item is seen as a function of the demands set by the item tasks and the abilities and attributes which the examinee may find necessary or useful in responding correctly to the item. In addition, interactions of examinee abilities and item characteristics may occur where solutions to an item may be reached by using different strategies or abilities and the difficulty of meeting the item demand using these different approaches is not equivalent. Examples from the measurement and psychological literature provide suggestions of a number of examinee characteristics and item properties which might be expected to affect item difficulty. The formulation is then extended to item discrimination. Finally, the literature on verbal analogies is reviewed within the theoretical framework to suggest sources of variation in the difficulty and discrimination of this item type.