Bilingualism, feedback, cognitive capacity, and learning strategies in L3 development

99學年度新聘教師論文著作[[abstract]]Part of the Latin Project1 , the current study investigates the relationship between individual differences (Level of Bilingualism) and pedagogical conditions (Types of Feedback) as it affects L3 development. The design includes working memory (WM) capacity and learning strategies as moderator variables to explain the effects identified. This laboratory study looks at 90 L1 Mandarin/L2 English bilinguals as they interact with a computer lesson on assignment of semantic functions in L3, Latin. Overall, the current study supports previous studies (e.g., Alanen, 1995; Carroll & Swain, 1993; de Graaff, 1997; DeKeyser, 1995; Ellis, 1993; Nagata, 1993; Nagata & Swisher, 1995) showing that more explicit feedback is more effective than less explicit, "right" or "wrong" feedback. The results support Cummins' Threshold Hypothesis (1976) as they identify a threshold between the intermediate and beginning L2 levels at which the benefits of the bilingual experience appear. However, the results also suggest that the appearance of bilingual advantages depend on the complexity of the tasks performed (Bialystok, 1986, 1988, 1997, 1999, 2001, 2004, 2008). Specifically, the results support Lado (2008) and Sanz, Anfruns, Lado, Lin, and Medina (2005a) showing that participants at a higher L2 level outperformed their counterparts only in the less explicit condition, and that more explicit feedback may level out differences among learners at different L2 levels. Different results identified by repeated measures ANOVAs and ANCOVAs revealed a key role for WM capacity and Learning Strategies in L3 development. The current study supports the more the better hypothesis (Miyake & Friedman, 1998) according to which, having higher WM capacity helps learning and domain-generality hypothesis (e.g., Turner & Engle, 1989) , as WM capacity does not seem to be language dependent. The results also suggest that sentence span, computational span, and speed of processing tests are more suitable WM measures than non-word recall and digit backward tests. Lastly, different results identified by repeated measures ANOVAs and ANCOVAs in the current study revealed a relationship between Learning Strategies and L3 development: Specifically, between Compensation Strategies and L3 development, and between Metacognitive Strategies and L3 development. Importantly, the results support previous studies on strategies showing that participants at a higher proficiency level use strategies more frequently than their counterparts, and extend the findings of Wharton (2000) and Kemp (2007) to the initial L3 development in a laboratory environment with participants of different L2 proficiency levels in an EFL context