An Application of Item Response Theory to Language Testing: Inadequacy of the Rasch Model

Abstract

Even though the application of IRT to language testing has recently attracted much attention, no model- data fit research has been conducted to explore the appropriateness of IRT modeling in language testing. No study as examined the adequacy of the Rasch model which has been predominant n language testing. Employing fit statistics, and the residual analyses, the current paper coducted model-data fit studies on two EFL reading tests with the 1-, 2-, and 3- IRT models to investigate the adequacy of the Rasch model. The results clearly indicate that the Rasch model fails to provide an adequate fit for these data, thus suggesting that the prevailing use of the Rasch model in language testing needs to be re- evaluated. The comparison of model-fit between the 3- parameter model and the 2- parameter model also suggests that for language tests, the discrimination parameter is more significant than is the guessing parameter.