This presentation examines the construct implications of accessibility requests in high-stakes testing, highlighting significant tensions when accommodating diverse test-taker needs within standardized environments. Common requests can conflict with security measures (e.g., lockdown browsers vs Bluetooth cochlear implants) or the language construct itself (e.g., screen readers in Reading tests), potentially disadvantaging certain test-taker groups.

Using 2020-2025 reasonable adjustments requests data from a high-stakes computer-adaptive language test, this case study investigates:
1. What are common request types and associated disabilities, and are there notable patterns?
2. [How] do specific accessibility needs conflict with test security and the operationalized language construct?
3. How can the challenges identified be addressed?
4. What are the perspectives and experiences of test takers, test centre managers, and accessibility advocacy groups regarding the implementation and effectiveness of accessibility accommodations in high-stakes language testing?

To answer these questions, a mixed-method analysis was employed on anonymized reasonable adjustments request forms and communication with test takers, particularly in Europe and South America. The quantitative analysis reveals the frequency of disabilities and identifies instances where conflicts between the construct and testing approach occur. Qualitative insights from interviews with test takers, test centre managers, and an accessibility advocacy group shed light on practical challenges, perceived fairness, and suggestions for improving accessibility in high-stakes language testing environments.

This study provides empirical grounding for the debate about the future of language assessment constructs, questioning whether existing frameworks can simply ‘evolve’ through procedural adjustments, if accessibility demands necessitate a ‘revolution’ in conceptualization and measurement, or if alternative pathways like subskill exemption are adequate for equitable and valid assessments.