| Abstract: | Reading comprehension is difficult to measure because it is a multifaceted construct influenced by a variety of cognitive, social and affective variables. There are also many distinct reasons for measuring reading comprehension such as the evaluation of instructional programs, the ordering of students by ability, and the diagnosis of reading difficulties. In this article we suggest that appropriate measures of reading comprehension depend on the fit between the purposes for testing and the properties of individual tests. Three test properties that we identify are statistical: stability of individual differences, consistency of an individual's scores across testing occasions, and sensitivity of the criterion variable to treatment or growth. Three other properties are conceptual in nature and pertain to test validity: nomothetic span, construct representation, and penetration. Each of these test properties and purposes is described, as well as the perils of mismatches among them. Comparative research on reading tests, construction of tests from theories and models, and use of testing portfolios can all improve the effective measurement of reading. |
