Overview: This paper examines the idea that variability in the bilingual experience both of L2 acquisition as well as current use is key to consider when thinking about neurobiological models of bilingualism and the link between bilingualism and cognitive control. The paper starts by talking about the Neurocognitive Language Control (NLC) model that they lay as the framework of cognitive control being recruited for managing multiple languages. They tie very specific regions to different aspects of control and describe left and right IFG connectivity varying with age of acquisition. The authors also point out a link between cognitive control abilities and bilingualism however they don’t point out the controversy surrounding this relation. They do point out that cognitive control advantages may differ between early and late bilinguals and importantly that it may differ by the context of language use, with a focus on social context. The paper then sets out to measure the social diversity of language use- whether it is fairly homogeneous or complex and involving a lot of code switching. They use entropy as a proxy for social complexity ie higher entropy is more complex and relate this measure to resting state and behavior.
Methods: Participants - 27 French-English bilingual adults. Language use - language entropy was computed from language use questionnaires across three settings, home, work, and social settings. A proportion of L1 to L2 was calculated with 0 meaning no complexity and one language and 1 being 50/50 complex or diverse language use cognitive control - AX-CPT task, participants were instructed to press a certain button if they saw an X after an A this taps prospective control which they suggest specifically varies with bilingualism as opposed to reactive control with the link here being prospective control is related to switching and managing two languages based on context. Imaging acquisition - data were acquired on a Trio 2.2 s TR 5 minutes of rest. Functional Connectivity analysis - FD threshold 1mm WOW outliers were modeled out at first GLM? physiological noise was estimated and regressed out. Global signal was NOT regressed out. Motion was regressed out, at Level 1. Five ROIs were used as seeds, two in the L IFG, on the anterior cingulate, one in left caudate, and one in left IPL. Correlations were computed between the seeds and all other voxels in the brain at first level analysis. Second level analysis, first level maps of connectivity were entered and measures of interest L2 acquisition and social entropy were regressed on the connectivity maps. Models for each seed were computed. p < .001 correction.
Results Behavior - There was a significant interaction with both L2 acquisition and social entropy and condition such that individuals with earlier L2 acquisition had less reliance on prospective memory and individuals with more complex language aka higher entropy had more reliance on prospective memory. FC results - connectivity between the L IFG and a cluster in the R IFG was inversely related to age of acquisition ie earlier acquisition meant stronger connectivity between the regions. Social entropy was positively related to connectivity between anterior cingulate and and the putamen. It was also positively related to connectivity between left caudate and and bilateral STG. Connectivity strength between L and R IFG interacted with condition of the AX CPT task such that greater connectivity was related to less reliance on proactive control and greater anterior cingulate putamen connectivity was related to greater reliance on proactive control.
Thoughts: First of all, the NLC model introduced in the beginning, have to look more into that. They set it up as a very regional breakdown of cognitive control in the brain which I don’t buy into. Slightly confused about the connectivity analysis and how they entered a seed connectivity map into a level 2 model and then regressed parameters on it. Motion threshold is wildly high for the rest frequency they are testing. The idea of testing individual differences in language use however is something important and I am glad they highlight and tackle it in two ways. The results seem fine? Hard to take away anything big from them.