Sentence processing in first language attrition: the interplay of language, experience and cognitive load
Item statusRestricted Access
Embargo end date22/12/2023
In a bilingual mind, two languages frequently interact with each other during language comprehension and production. Both languages stay active, and bilingual speakers must resolve interferences from the unwanted language, to understand and produce speech in the target language (Linck, Hoshino, & Kroll, 2008; Linck, Kroll, & Sunderman, 2009). Inhibiting a language is cognitively demanding, and this kind of mental practice is carried out by bilingual speakers daily (Green, 1998). The bilingual experience not only interacts with speakers’ general cognitive functions, such as attention control and working memory, it also affected their linguistic performance in both of their languages (Bialystok, 2009; Kroll & Bialystok, 2013). Immersion in the second language (L2) environment with limited contact with the first language (L1) can gradually lead to change in the L1 system in terms of knowledge, processing and use; for example, slower and less efficient access to the L1 compared to the L2, even after the L1 has been fully acquired (Levy, McVeigh, Marful, & Anderson, 2007). This non-pathological, post-puberty change in the L1 due to the interaction between two languages is what we called L1 attrition in adulthood (Köpke & Schmid, 2004). Given the state of the literature at the moment, bilingual processing at the lexical level has been investigated more often than bilingual processing at the sentence level (Kroll, Dussias, Bice, & Perrotti, 2015; Schmid & Köpke, 2017a). Granted that bilingual lexical retrieval can bring us valuable insights into bilingual processing in general, but sentence-level processing involves not only word recognition and lexical retrieval, but it also involves semantic processing, syntactic processing, as well as discourse processing. During this process, the interplay between language and cognition is quite different from what we have found on the lexical level. The current study aims to fill this gap by investigating sentence processing in L1 attrition, among Mandarin-English late bilingual speakers with different bilingual profiles (passive vs. active bilinguals) and varying lengths of residence (LoR) in the L2 environment. I investigated two structures in this dissertation, wh-topicalization and reflexive ziji (‘self’) in Mandarin. Processing said structures requires integrating information from different domains, including syntax, semantics and discourse. In addition, Mandarin and English differ in terms of syntactic, semantic and discourse constraints to those structures, making them the ideal testing ground for cross-language influence. The first study (Chapter 4) examined the acceptability judgments of sentences with wh-topicalization in a speeded paradigm. Different from wh-movement in English, wh-phrases in Mandarin wh-questions usually stay in situ, or they can be topicalized in specific contexts. Native mandarin speakers must use both syntactic and discourse information to decide whether it is grammatical to topicalize the wh-phrase in the current context. Using a Speeded Acceptability Judgment Task, I found that sensitivity to the discourse information decreased as bilingual speakers spent more time in the L2 environment: as LoR increased, their judgments became more neutral; and it became increasingly difficult for them to reject ungrammatical or less-preferred wh-topicalizations. Also, their linguistic performance was associated with their performance in cognitive tasks, suggesting certain aspects of general cognitive functions affected the attrition process. The second study (Chapter 5) used a Speeded Comprehension Task with two-alternative forced-choice to investigate how bilinguals interpret the Mandarin reflexive pronoun ziji under time pressure. Different from its English counterpart self, ziji allows both long-distance and local binding; besides, the pronoun resolution is highly context-dependent: native speakers use either semantics or discourse information to resolve ambiguities when there is more than one potential antecedent in the context. Results show that despite long-term immersion in the L2 environment, active bilingual speakers outperformed their passive bilingual peers in both reaction time and response accuracy. All participants, however, showed the same preference for local binding and were more accurate when they used the semantic rather than the discourse information to determine the antecedent. A significant correlation between cognitive tasks and the speeded comprehension task suggests that general cognitive abilities, such as working memory and attention control, play a role in pronoun resolution in L1 attrition. Through these studies, this thesis contributes to the understanding of bilingual processing and L1 attrition at the sentence level. Although much more work is required to fully understand the interplay of language, experience and cognitive factors, this thesis provides evidence that the effect of L1 attrition manifests itself differently with different linguistic structures, and the degree of L1 attrition is influenced by both bilingual experience and cognitive demands of the task.