Differences Across Levels in the Language of Agency and Ability in Rating Scales for Large-Scale Second Language Writing Assessments

Abstract

Abstract While large-scale language and writing assessments benefit from a wealth of literature on the reliability and validity of specific tests and rating procedures, there is comparatively less literature that explores the specific language of second language writing rubrics. This paper provides an analysis of the language of performance descriptors for the public versions of the TOEFL and IELTS writing assessment rubrics, with a focus on linguistic agency encoded by agentive verbs and language of ability encoded by modal verbs can and cannot. While the IELTS rubrics feature more agentive verbs than the TOEFL rubrics, both pairs of rubrics feature uneven syntax across the band or score descriptors with either more agentive verbs for the highest scores, more nominalization for the lowest scores, or language of ability exclusively in the lowest scores. These patterns mirror similar patterns in the language of college-level classroom-based writing rubrics, but they differ from patterns seen in performance descriptors for some large-scale admissions tests. It is argued that the lack of syntactic congruity across performance descriptors in the IELTS and TOEFL rubrics may reflect a bias in how actual student performances at different levels are characterized.