Development of Cognitive Processes in L2 Reading Comprehension: A Longitudinal Bayesian Knowledge Tracing Study of Chinese EFL University Students

Understanding how the cognitive processes underlying second language (L2) reading comprehension change over time is crucial for characterizing reading development among English as a Foreign Language (EFL) learners. Guided by Kintsch’s Construction–Integration (C-I) model and Complex Dynamic Systems Theory (CDST), this study examines the development of four cognitive processes in reading comprehension in terms of longitudinal changes in their estimated mastery: extracting explicit information, interpreting and rephrasing information, making inferences, and synthesizing information across text. Data from 303 undergraduate EFL learners across three semesters were analyzed using Bayesian Knowledge Tracing (BKT), with item difficulty incorporated to improve the estimation of process mastery. The results showed differentiated trajectories of estimated mastery across the four processes. Extracting explicit information and making inferences showed increases in estimated mastery, whereas interpreting and rephrasing information and synthesizing information across text changed more slowly. Cluster analysis further identified four learner profiles based on estimated mastery changes (Early Decliners, Moderate Maintainers, Imbalanced Performers, and Advanced Achievers), highlighting substantial variability in EFL reading development. These findings advance understanding of L2 reading comprehension by showing that component cognitive processes follow differentiated and heterogeneous trajectories over time, and also demonstrating the utility of BKT for tracing process-level changes from longitudinal item-response data.