Slide Slam R12
The language network and the executive control network are distinct in bilinguals
Saima Malik-Moraleda1,2, Theodor Cucu1, Benjamin Lipkin1, Evelina Fedorenko1,2; 1Massachusetts Institute of Technology, 2Harvard University
The brain networks that support language comprehension vs. executive control have been shown to be robustly dissociated in monolinguals (e.g., Fedorenko et al., 2012; Monti et al., 2012; Blank et al., 2014; Mineroff et al., 2018; Braga et al., 2020). However, the relationship between language processing mechanisms and domain-general executive control mechanisms in bilinguals remains debated. Some have explicitly argued that these mechanisms are more integrated/overlapping in bilinguals, with some parts of the language network, within the left inferior frontal cortex, supporting executive functions (e.g., Garbin et al, 2010; Coderre et al, 2016). However, empirical support for the latter claims comes from brain imaging studies that rely on averaging individual activation maps in the common space and drawing inferences from the group-averaged map. This approach can underestimate the separation between nearby functionally distinct areas given the inter-individual variability in the precise locations of these areas (Nieto-Castañón & Fedorenko, 2012). We investigated the relationship between the language network and the domain-general executive control (Multiple Demand (MD); Duncan, 2010) network using the individual-subject analytic approach. We used relatively large and carefully matched samples of early balanced bilinguals (n=80) and monolinguals (n=73) who completed a sentence comprehension task (Fedorenko et al., 2010) and an executive (spatial working memory) task, which has been previously established to be an effective ‘localizer’ for the MD network (Shashidara et al., 2019; Assem et al., 2020). We defined the language and the MD network in each individual, and then examined their responses to the two tasks using across-runs cross validation to ensure independence (Kriegeskorte et al, 2009). We found that the two networks are as robustly dissociated in the bilinguals as they are in the monolinguals: the language-responsive regions showed no response to the working memory task, and the MD regions showed a stronger response to the meaningless and unstructured control condition (nonword sequences) than to sentences. Interestingly, we observed overall stronger responses to the working memory task conditions, and a larger difference between the harder and the easier conditions, in the MD network of the bilinguals compared to the monolinguals. Given that stronger responses and a larger hard>easy effect have been linked with superior behavioral performance on executive tasks and a higher IQ (e.g., Assem et al., 2020), these results are compatible with the idea of superior executive abilities in bilinguals (e.g., Bialystok, Craik & Luk, 2012). That said, directly linking this difference to differences in linguistic experiences is challenging. Critically, even though differences in the MD network’s functionality in the two populations deserve further investigation, this network is robustly dissociated from the language network, even though the two lie side by side in the inferior frontal cortex (Fedorenko & Blank, 2020). More generally, this work highlights the importance of subject-specific analyses in the study of bilingualism, especially given the focus on the relationship between cognitive mechanisms that are supported by closely adjacent networks, and the well-established inter-individual variability in their precise locations in the frontal cortex (e.g., Frost & Goebel, 2011; Tahmasebi et al., 2012).