Refining the processing dynamics of English compound words in L2 learners: a psycholinguistic modeling approach

Understanding how EFL learners process compound words offers important insight into the structure and dynamics of the mental lexicon. We investigated how different linguistic features—frequency, semantic transparency, and morphological structure—influence the time course of compound noun processing in Chinese learners of English. Eye-tracking data were collected from 40 advanced L2 learners while reading sentences containing compound nouns. Three temporal measures (first fixation duration, gaze duration, total fixation duration) were used to index early, middle, and late processing stages. Using supervised machine learning models (decision tree, random forest, neural network, SVR), we identified the optimal predictors for each stage. Results showed frequency measures exerted dominant effects across all stages, while semantic features increasingly contributed at later stages. Morphological features were comparatively weak. These findings supported a dynamic, data-driven perspective on compound word processing and align with the theoretical principle of opportunity maximization. Implications for psycholinguistic theory, second language acquisition, and hybrid computational modeling were discussed.