Investigating phonological and semantic cues in sentence processing and prediction among Mandarin Chinese EFL learners of different proficiency levels : a visual world eye-tracking study

Thesis

Predictive mechanisms are central to real-time language comprehension, yet it remains debated how L2 proficiency shapes the use of semantic and phonological cues, and whether L1 phonology is co-activated during L2 processing. This study investigated the time course of prediction in Mandarin Chinese learners of English at two proficiency levels (advanced, intermediate) compared with native English speakers, using a visual world eye-tracking paradigm with cloze sentences omitting the target noun. Sixty-one participants (advanced L2: n = 22; intermediate L2: n = 18; L1 English: n = 21) viewed four-picture arrays that always included one of the following: a target, an English semantic competitor, an English phonological competitor, or a Chinese phonological competitor, with the remaining pictures serving as distractors, while hearing each sentence once. Webcam-based eye-tracking (Gorilla) incorporated automated polling, calibration, and quality control. Across groups, anticipatory looks to the target emerged well before the cloze position, confirming the prediction. Native listeners showed robust, sustained semantic activation; advanced L2 listeners also activated semantic alternatives, but with delayed and more variable timing. English phonological competitors generally attracted low early fixations, though subtle and delayed activation was observed in the L1 and advanced L2 groups. Notably, advanced L2 participants showed a late increase in fixations to the Chinese phonological competitor, surpassing intermediate L2 listeners in the predictive window, which provides evidence for the transient co-activation of L1 phonology during L2 processing among highly proficient bilinguals. Intermediate L2 listeners focused almost exclusively on the target, with no reliable engagement of semantic or phonological competitors. These findings support a graded, proficiency-dependent account of prediction, clarifying when (late vs. early) and for whom (advanced vs. intermediate learners) L1 phonology intrudes during L2 prediction. The results also have implications for bilingual processing models that aim to promote anticipatory, meaning-based comprehension in advanced learners.

Authors: Tang, Q

• DOI: 10.5287/ora-xqqbqebpn
• Type of award: MSc taught course
• Level of award: Masters
• Awarding institution: University of Oxford
• Language: English
• Keywords: psycholinguistics; eye-tracking; ALSLA2025; quantitative research; ALSLA; gorilla experimental builder; language predictive mechanism; visual world eye-tracking study; sentence processing; applied linguistics
• Subjects: Quantitative research; Second language acquisition; English language--Conversation and phrase books--Mandarin; English language--Study and teaching--Mandarin speakers; Phonological awareness; English language--Study and teaching--Foreign speakers; Semantics; Eye tracking; Psycholinguistics; Phonology; English language--Sentences; Applied linguistics