When Does Attachment Vary? Relative Clause Position and Individual Differences in Chinese Learners of English

This study investigated attachment resolution in second language (L2) processing by examining whether Chinese learners of English adopt fixed or variable attachment when reading English relative clauses (RCs), and how these preferences are shaped by syntactic position and individual differences. Using eye‐tracking, participants read sentences containing relative clauses that were either globally ambiguous or disambiguated toward high or low attachment through reflexive‐gender cues, with RC‐modifying position (subject vs. object) manipulated. An offline comprehension task assessed attachment interpretations. To examine individual differences in attachment processing, L2 proficiency and working memory were measured independently. The eye‐tracking data revealed a robust ambiguity advantage, with globally ambiguous clauses processed faster than disambiguated ones, particularly in later eye‐movement measures such as regression path time and total reading time. Offline responses showed different attachment patterns across RC‐modifying positions, with more high‐attachment responses when the RC modified the object. Online reading was also modulated by L2 proficiency. Higher‐proficiency learners exhibited a greater decrease in reading times for globally ambiguous sentences compared to disambiguated ones. Furthermore, a three‐way interaction showed that higher proficiency amplified the differences between attachment conditions in object‐modifying RCs but not in subject‐modifying RCs. Additionally, although the study found no effects of working memory on attachment preferences, higher working memory was associated with longer reading times across conditions. Overall, the findings support a variable account of L2 attachment resolution and suggest that both structural factors and learners’ proficiency contribute to variation in L2 sentence processing. The study contributes by distinguishing between Fixed‐Attachment and Variable‐Attachment Accounts and showing that L2 attachment preferences vary across structures and across learners.