Late Second Language Acquisition
The reasons why language acquisition is more difficult in adulthood than childhood, and often results in limited language attainment, are of major interest not only to researchers, but also to language learners, applied linguists, and stakeholders in education policy, particularly in multilingual countries such as Canada. The 'critical period hypothesis' (Lenneberg, 1967) maintains that brain-maturational constraints irreversibly prevent late learners (e.g., after puberty) from relying on the same neuro-cognitive mechanisms used by early learners (e.g., native speakers). Alternative accounts question the existence of a critical period in SLA, refer to reported cases of native-like proficiency in late L2, and hold factors such as first language (L1) background, socialization, and motivation levels responsible for differences in final attainment (Birdsong, 2006). With respect to brain mechanisms, my colleagues and I have argued that proficiency rather than age of L2 acquisition predicts the brain activation, such that even late learners can elicit 'native-like' activation patterns, once they reach high L2 proficiency levels (Friederici et al., 2002; Steinhauer et al., in press). In the second term of my CRC, my interdisciplinary research program will have an additional focus on transfer and competition effects between languages and will also employ new experimental paradigms.
A current PhD student in the program, Ms Erin White, uses ERPs to investigate proficiency-dependent changes of brain activation profiles in French, Chinese and Korean late learners of English, and will now also examine the downside of brain plasticity: first language attrition in adopted children. The postdoctoral fellow in my lab, Dr Drury, and several other graduate students will continue to be involved in conducting ERP studies on rule learning and phoneme discrimination in early and late bilinguals. A new PhD student will be recruited in 2009 to develop artificial language paradigms for ERP studies. Our findings can be expected to have strong impact on theories of second language learning, and are integrated in the neurophysiological model of second language acquisition that I began to develop during the first CRC term (Steinhauer et al., in press). At a larger scale, this work will contribute to Canada's role as a leading nation in research on language learning and bilingualism.
(a) Proficiency stages and language transfer. My data-based model of SLA motivates a specific sequence of five ERP profiles corresponding to increasing levels of L2 proficiency in morpho-syntax. Proficiency levels are specific to linguistic structures, such that a given L2 learner will display higher levels of proficiency for simpler structures, where relative difficulty also depends on L1 background (transfer). A given structure in English may be difficult for French learners of English (e.g., adjective placement: the red house vs. la maison rouge), but not for Chinese learners, and vice versa. To further test the model, our current large-scale study on late SLA will be extended to systematically contrast simple and difficult syntactic violations in the same subjects and cross difficulty with L1 background.
(b) Artificial Language (AL). As my previous student conducting artificial language research at Georgetown is now setting up her own EEG lab in Chicago, I plan to continue my innovative ERP work at McGill. These paradigms serve as highly controlled miniature models of natural language learning and allow studying rapid increases in proficiency levels within weeks. One such project will test the validity and replicability of an important recent study (Bahlman et al., 2006) whose findings of distinct ERP patterns for different grammars (phrase structure versus finite state) are likely to be influenced by methodological flaws. The second experiment will test cross-domain transfer effects in grammar rule learning. (c) Cross-modal transfer. One of our recent studies suggests that syntactic rule learning can take place in two different neurocognitive modes (Morgan-Short, Steinhauer et al., 2007), which may involve different memory circuits. It has been claimed that implicit rule learning in language and other cognitive domains relies on the same procedural brain circuits (Ullman, 2001), but this has not yet been tested empirically. I plan to conduct a study in which subjects acquire the exact same grammar rules twice: (a) in a linguistic context using AL and (b) in a non-linguistic context using the serial reaction time (SRT) task. In SRT tasks, rules are learned implicitly by pressing buttons that correspond to lights that flash in a sequence following the (grammar) rules (Seger, 1997). In case of shared circuits, learning in one task should profit from previous rule experience in the other task (transfer). As SRT tasks involve both motor and cognitive learning, which can be teased apart (Goschke et al., 2001), transfer effects between language and SRT will be revealed in both of these dimensions. This study would provide a strong test for Ullman's theory and clarify the domain specificity of linguistic rule learning.