Exploring linguistic characteristics across CEFR writing levels and their relationships with student backgrounds

This study attempted to identify significant linguistic features that distinguish across CEFR writing levels. The sample included essays from 959 German-speaking EFL students on two tasks: an Independent argumentative task and an Integrated summarizing task. Using natural language processing tools, 19 linguistic features encompassing fluency, lexical, syntactic, and cohesion dimensions were extracted and analyzed through ANOVAs and post-hoc tests. Additionally, regression analyses were conducted to investigate significant predictors of CEFR writing levels with both linguistic features and students’ background variables included in the models. The results indicated that fluency features significantly differentiated CEFR levels in both tasks. However, the patterns of other significant features varied depending on the task type. The students’ self-assessment factor was also found to be a strong predictor of CEFR writing levels. These findings offer practical implications, including the importance of incorporating various genres in writing assessment and instruction.