Topic and Accent Coverage in a Commercialized L2 Listening Test: Implications for Test-takers’ Identity

Abstract

Abstract I analyzed a corpus of the international English language testing system (IELTS) comprising 256 listening sections (1996–2021). The primary objective of the study was to gain insights into the assumptions made by test designers regarding the real-life contexts that test-takers will encounter. Overall, 15 superordinate topic areas and 300 subtopics were identified in the corpus. There was relatively more diversity in topic coverage before 2000. However, the test did not incorporate texts that would address sociocultural matters related to local or international contexts. Additionally, English-as-L2 accents were virtually unrepresented with only three samples from 514 speakers, potentially suggesting a racialized perspective on the listening construct. I argue that it is possible that this way of testing promotes and normalizes test designers’ ideologies, while overlooking the importance of the diversity of domains that test-takers will encounter in daily life. I discuss the potential construction of test-takers’ identity through exposure to these topics and suggest that test designers may consider reevaluating topic and accent coverage in the test to improve fairness and equity in the test. Finally, I provide ideas on how the use of generative artificial intelligence (AI) can enhance quality of language assessments.