Investigating criterial discourse features across second language development: lexical bundles in rated learner essays, CEFR B1, B2 and C1

In this study, we investigated criterial discourse features in L2 writing through the use of recurrent word combinations, a.k.a. lexical bundles, taking a corpus-driven and expert-judged approach by examining L2 English data across various proficiency levels from L1 Chinese learners. Proficiency was determined by a robust rating procedure which is often used in high-stakes tests, instead of the traditional approach of utilizing extra-linguistic judgement such as program levels. Expository and argumentative essays produced by learners were rated by experienced raters and then subjected to post-rating statistical analysis. Three sizeable subcorpora, representing the Common European Framework of Reference B1, B2, and C1 levels, were then selected for investigation. After lexical bundles were retrieved and refined, structures and discourse functions were manually annotated. The findings suggest that learner writing at lower levels tends to share more features with conversation, whereas the discourse of more proficient writing is closer to that of academic prose. The implications and limitations of the study will also be discussed