From coarse to fine: A cognitive diagnosis of EFL learners’ inferential ability in EFL reading

Abstract

This study applied a multi-CDM approach to diagnose EFL learners’ inferential ability via an online diagnostic reading test covering six inferential subskills. By integrating data-driven approach with linguistically informed modifications, the multi-CDM was validated using responses from 886 Chinese university students. Results demonstrated the superiority of multi-CDM over six single CDMs: it achieved a decent model-data fit, the highest test-level classification accuracy (72.1%) and subskill-level reliability (87%-95.8%), showing robust diagnostic performance. Subsequent diagnosis revealed pronounced disparities: high-proficiency learners excelled in literal comprehension and local inferences, yet struggled with temporal inference. Low-proficiency learners exhibited systemic deficiencies, particularly in global inferences and literal comprehension, underscoring cognitive overload challenges. Instructional implications were provided to different groups. Finally, limitations and future directions were discussed.