A Critique of Classical Test Theory

Abstract

Examination of the methodological literature of the behavioral and social sciences indicates that measurement terms are used differently than in the natural sciences. The rationale for these departures is usually ascribed to classical test theory, a measurement model claimed to be more applicable to psychological and social data than the traditional measurement model of the natural sciences, which requires the development of standard instruments defining, by consensus, parametric values. Classical test theory seemingly avoids this necessity, but only by inviting validation by fiat, resulting in instrument evaluations which are trivial, misleading, or invalid. The development of measurement in the behavioral and social sciences might be encouraged by the abandonment of classical test theory and a return to natural-science measurement theory.