An example of empirical and model based methods for performance descriptors: English proficiency test

Abstract

Great emphasis is given to the development of high-stake tests all around the world and in Turkey. However, limited emphasis is given to adequate score reporting. Too much emphasis on rankings and almost no emphasis on performance level descriptors (meaning of the scores) have leaded a “ranking culture” in Turkey. There is an immense need to raise awareness about score reporting and performance level descriptions in Turkey. This study aims to raise awareness about the use of performance level descriptors in a high-stake exam in Turkey, an English proficiency exam. The study sample is consisted of 630 undergraduate students who took the 2016-2017 English proficiency exam of a public university in the southwest of the Turkey. In order to identify the potential exemplars, two types of item mapping methods (i.e. experimental based method and model-based method) were used in the present study. Item grouping for performance level descriptors provided hierarchical and interpretable structure. Using these performance level descriptors, it is possible to give criterion referenced feedback to each student about his/her reading abilities.