You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster A65, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

Variability in BOLD correlates of semantic judgment reduces with proficiency among L2 learners

Angela Grant1,2, Ping Li1;1The Pennsylvania State University, 2Concordia University

Research on second language (L2) acquisition has identified a common network of regions recruited during L2 processing: the language control network (Abutalebi & Green, 2007). Activity in this network, which includes prefrontal cortex, anterior cingulate cortex (ACC), basal ganglia and the inferior parietal lobule, is thought to be inversely correlated with proficiency, such that less proficient learners show more activity (Abutalebi, 2008). Yet, these regions are also frequently implicated in studies of proficient and even simultaneous bilinguals (e.g., Román et al., 2015). Our study uses a multivariate decoding technique in addition to traditional univariate analyses to better differentiate the effects of proficiency on the recruitment of regions in the language control network. Specifically, we used a support vector machine classifier as part of the Decoding Toolbox (Hebart, Görgen & Haynes, 2014) to categorize data from high and low proficiency L2 learners who completed a semantic judgment task. Our analyses find that the classifier was able to identify similar multi-voxel patterns in the caudate nucleus (part of the basal ganglia), ACC, and inferior frontal gyrus (IFG) among high proficiency, but not low proficiency learners. The reduced variability present in the BOLD signal was also accompanied by reductions in variability on a variety of behavioral tasks, including the semantic judgment task itself, participants’ self-ratings of L2 proficiency, and their L2 verbal fluency. Our multivariate decoding analyses complement our univariate analysis, which found significantly higher activation in the caudate for the low proficiency learners, while the ACC and IFG did not differ significantly in their gross activation level between the two groups. The combination of univariate and multivariate analyses of L2 BOLD data show that the effects of proficiency are not limited to overall activation level in the language control network. As learners increase in proficiency, the multi-voxel activation patterns within those regions also change to become more consistent, even across learners. Furthermore, the fact that the classifier identified regions in the language control network from a searchlight analysis of the whole brain provides novel support for the importance of the language control network as identified by Abutalebi and Green (2008; Green & Abutalebi, 2013).

Topic Area: Multilingualism

Back to Poster Schedule