Predicting the difficulty of EFL reading comprehension tests based on linguistic indices

Abstract

Abstract Estimating the difficulty of reading tests is critical in second language education and assessment. This study was aimed at examining various text features that might influence the difficulty level of a high-stakes reading comprehension test and predict test takers’ scores. To this end, the responses provided by 17,900 test takers on the reading comprehension subsection of a major high-stakes test, the Iranian National University Entrance Exam for the Master’s Program were examined. Overall, 63 reading passages in different versions of the test from 2017 to 2019 were studied with a focus on 16 indices that might help explain the reading difficulty and test takers’ scores. The results showed that the content word overlap index and the Flesch-Kincaid Reading Ease formula had significant correlations with the observed difficulty and could therefore be considered better predictors of test difficulty compared to other variables. The findings suggest the use of various indices to estimate the reading difficulty before administering tests to ensure the equivalency and validity of tests.

Keywords

• Reading comprehension
• Corpus linguistics
• Reading difficulty
• Readability
• Computational linguistics