In today's interconnected world, English has evolved far beyond the boundaries of traditional "native" varieties like American or British English, gaining the status of an International Language. This means that in professional contexts, academic settings, and international communication, English speakers encounter a wide variety of Global English (GE) accents daily.
This raises an important question about the design of English tests: Should high-stakes English tests reflect this diversity by including different accent varieties?
I recently collaborated on a study led by Okim Kang and sponsored by the Duolingo English Test Competitive Grants published in Language Assessment Quarterly. The study explores exactly this question from the perspective that matters just as much as that of test developers and researchers—the test takers themselves!
What happens when test takers hear English in diverse accents?
The study examined how 160 test takers from four different language backgrounds (Chinese, Korean, Hindi, and Latin American Spanish) responded to different accent varieties in simulated listening tasks from the DET (the 2021 test version).
We were particularly interested in not just these test takers’ stated preferences, but also how their attitudes might relate to their actual test performance when hearing different accents. The research included both sStandard English accents (American and British English), and two kinds of “non-standard” GE accents: shared GE accents similar to their own (for example, English learners from Korea heard other Korean speakers), and non-shared GE accents—that is, accents test takers may not be as familiar with.
We specifically asked test takers whether they thought non-standard accents were acceptable and fair on high-stakes assessments, and what reservations they had about including them.
Preferences vary, but performance stays steady
What we found was a much more nuanced picture than we expected: there was no clear consensus among test takers about accent preferences. While our quantitative data showed American English was rated most acceptable overall followed by British English, the qualitative responses showed a more complex picture.
Some participants, especially those from India and Mexico, strongly advocated for accent diversity. As one Indian participant told us: "As we are living in a diversified world, we should be familiar with every other's accent." A Mexican test taker similarly noted: "I think both ways are trusted. However, the one that includes non-native English accents is more inclusive and realistic." Yet others, especially those in East Asia, felt equally strongly that standard varieties were more appropriate for testing contexts.
These mixed perspectives were particularly evident when we asked about test fairness. Over half of the responses stated that listening tests with non-standard varieties would be less fair than those with standard English accents only, with concerns about "lack of familiarity" and "status of American and British English varieties as the acceptable, target, and prestigious accents." Interestingly, Mexican participants were significantly more accepting of accent diversity, while Chinese participants were most resistant to non-standard accents on listening assessments.
However, and perhaps most importantly, despite these varied preferences, we found minimal correlation between attitudes toward different accents and actual test performance (r < 0.26). This suggests that test takers' concerns about non-standard accents affecting their scores may not be entirely justified by real-world performance.
Shaping the future of fair language testing
This research-first, test-taker focused approach is a big part of what makes the DET stand out in language assessment. We aim at making decisions about test design based on empirical evidence, understanding that these decisions must be made deliberately and with strong research backing. This cautious but forward-thinking approach ensures that the test remains both valid and reliable.
The DET currently uses standard English varieties to ensure test consistency and fairness to all test takers (e.g., Taylor & Geranpayeh, 2011). The findings in this study provide important insights into test taker attitudes towards including different non-standard accents on the test. However, the mixed preferences among test takers suggest that any shift toward including more accent diversity would need to be implemented carefully, with substantial supporting evidence.