Unpacking the cognitive complexity of EFL listening through CDM-based DIF and subgroup profiling

Listening poses significant challenges for both English as a Foreign Language (EFL) learners and educators, and the cognitive complexity involved in listening test performance remains a relatively unexplored area. Using data from 4612 Chinese EFL college students, this study compared five cognitive diagnostic models (CDMs) and applied CDM-based differential item functioning (DIF) detection followed by attribute mastery profiling to ensure item-level fairness and examine subgroup differences. Results showed that the G-DINA model provided the best fit, supporting its applicability to large-scale listening assessment. Six listening subskills were identified, confirming that listening comprehension integrates both linguistic knowledge and cognitive processing, with lower-order skills generally more accessible than higher-order ones. Although only a small number of items exhibited DIF, clear differences emerged in mastery probabilities across gender and academic disciplines, revealing distinct cognitive profiles rather than measurement bias. These findings demonstrate the value of cognitive diagnostic assessment for refining construct representation, improving instructional targeting, and strengthening fairness in EFL listening assessment.