On the Predictive Validity (Generalizability) of IELTS Academic Tests

Authors

1 Department of Humanities, College of English language, Ahvaz Azad University Branch, Islamic Azad University, Ahvaz, Iran

2 Farhangian Teacher Education University, Iran

Abstract

Predictive validity of high-stake tests has attracted the attention of applied linguists since a couple of decades ago. Despite the existence of a plethora of studies on reliability and validity issues of high stake tests, it seems that predictive validity of International English Language Testing System (IELTS), as a high stake test should be measured appropriately. This study set out to investigate the predictive validity of IELTS academic tests practiced in Iran. The participants included 70Iranian IELTS test takers who took IELTS academic test in 2014 to 2015. To carry out the study, Construct and Validation Self-Assessment Inventory for Learners of English for Academic Purposes was used to collect data. The questionnaire included 48items on a five-point Likert scale. Coronbach's alpha coefficient of this questionnaire indicated the reliability of the instrument. The test-takers' scores on the four skills of academic IELTS were also used as another portion of data in this study. The collected data were subjected to a series of statistical analyses in relation to the proposed questions in this study. The data were analyzed using Regression and Analysis of Variance. The results of the study revealed that IELTS four modules significantly predict the test takers' performance in academic settings. The findings can be used by IELTS centers to make a change in test format, rubrics, as well as uses.

Keywords