Shahram Ghahraki - English Language and Literature Department, University of Isfahan, Isfahan, Iran
Manssor Tavakoli - Applied Linguistics Department, University of Isfahan, Isfahan, Iran
Saeed Ketabi - Applied Linguistics Department, University of Isfahan, Isfahan, Iran

Perhaps the degree of test difficulty is one of the most significant characteristics of a test. However, no empirical research on the difficulty of the MSRT test has been carried out. The current study attempts to fill the gap by utilizing a two-parameter item response model to investigate the psychometric properties (item difficulty and item discrimination) of the MSRT test. The Test Information Function (TIF) was also figured out to estimate how well the test at what range of ability distinguishes respondents. To this end, ۳۲۸ graduate students (۳۹.۹% men and ۶۰.۱% women) were selected randomly from three universities in Isfahan. A version of MSRT English proficiency test was administered to the participants. The results supported the unidimensionality of the components of MSRT test. Analysis of difficulty and discrimination indices of the total test revealed that ۱۴% of the test items were either easy / very easy, ۳۸% were medium, and ۴۸% were either difficult or very difficult. In addition, ۱۴% of the total items were classified as nonfunctioning. They discriminated negatively or did not discriminate at all. ۷% of the total items discriminated poorly, ۱۷% discriminated moderately, and ۶۲% discriminated either highly or perfectly, however they differentiated between high-ability and higher-ability test takers. Thus, ۳۸% of the items displayed satisfactory difficulty. Too easy (۱۴%) and too difficult (۴۸%) items could be one potential reason why some items have low discriminating power. An auxiliary inspection of items by the MSRT test developers is indispensable.

IRT, MSRT, high-stakes, item analysis, item difficulty, item discrimination, Accountability