Effects of language proficiency on cognitive control: Evidence from resting-state functional connectivity

Highlights

• Language proficiency could be interpreted to cause changes on some but not other components of domain-general cognitive control.

• Significant changes of resting-state functional connectivity were observed in cognitive flexibility and inhibition between high- and low-proficiency bilinguals. However, no difference was found in working memory.

• The strength of resting-state functional connectivity showed a significantly negative correlation with behavioral performance.

Abstract

Cognitive studies suggest that bilingualism plays an additional role in the development of cognitive control, specifically in that bilingualism has been found to promote cognitive abilities in switching and inhibition. In recent years functional neuroimaging studies suggest that long-term experience of speaking two languages results in changes of neural activity in the cognitive control network. Here we explore the impacts of second language proficiency on intrinsic functional connectivity of the executive function network using resting-state functional MRI. Seed regions centering on different components of cognitive control were selected for the resting-state functional connectivity (rsFC) analysis based on previous studies. We performed a functional connectivity analysis of high- versus low-proficiency bilinguals and found that language proficiency affected distinct components of the cognitive control system. Specifically, for switching, the rsFC of high-proficiency bilinguals was weaker than that of the low-proficiency peers in the left anterior cingulated cortex and for inhibition, in the right middle frontal gyrus. For working memory, however, the rsFC showed no difference as a result of proficiency. Finally, the strength of rsFC showed a significant negative correlation with behavioral performance in both bilingual groups. These findings were interpreted within the current debates on bilingualism and cognitive control.

Introduction

Bilinguals with two working languages in the brain need to select the correct language for language comprehension and production according to different contexts. In this process, cognitive control plays an important role (Abutalebi and Green, 2007; Bialystok, 2009). Because of their extensive experience in using two languages for communication, bilinguals have demonstrated better executive function, although this assumption has recently become contentious (De Baene et al., 2015; Luk et al., 2012; Bialystok, 2017; Paap et al., 2015). A large number of studies have devoted to the effects of bilingualism on domain-general cognitive control, but it remains unclear whether this effect could also be observed in the cerebral networks of domain-general cognitive control (Li et al., 2014a, Li et al., 2014b).

Cognitive control has been traditionally considered an integral mechanism, which usually functions in various cognitive tasks. But recent studies have begun to examine cognitive control in a componential perspective for bilingual effects (e.g., Dong and Li, 2015; Kroll and Bialystok, 2013). It is now regarded as a complex set of distinct cognitive domains, including, at least, inhibition (Jahanshahi et al., 1998), cognitive flexibility (Smith and Jonides, 1999), working memory (Goldman-Rakic, 1996), and planning (Morris et al., 1997). The current study adopts a 3-component definition of cognitive control according to Miyake et al.‘s framework, which contains inhibition, shifting, and working memory (2000).

The effect of bilingualism on cognitive control has been a topic of intensive study and debate in the last decade. There has been a large body of literature showing that bilingualism could confer cognitive benefits, for example, in terms of increasing accuracy and decreasing reaction time in a variety of paradigms such as Stroop task, flanker task, Simon task, and cued-task switching paradigm (Bialystok et al., 2004, 2005; Martin-Rhee and Bialystok, 2008; Costa et al., 2008, 2009). These studies have tended to focus on the bilingual effects in different domains of cognitive control, especially in inhibition and cognitive flexibility.

As a key component of cognitive control, inhibition has been highlighted in many studies. Inhibition refers to the ability to resolve conflicts by directly suppressing interfering stimuli and promoting attention to the target stimuli (Colzato et al., 2008). With the recruitment of inhibition, non-target stimulus is excluded and the relevant information is processed (Green and Abutalebi, 2013). Several studies have shown that bilinguals demonstrate a higher performance over monolinguals in suppressing interfering stimulus and in goal maintenance, and such stronger inhibition ability has been attributed to the acquisition and use of two languages (Green, 1998; Kroll et al., 2008; Bialystok, 2009). Other studies have shown bilingual effects due to inhibition as well (Bialystok et al., 2005; Bialystok, 2006; Misra et al., 2012; Barac et al., 2014).

In addition to cognitive performances, brain imaging studies have also provided rich information about the neurocognitive effects of bilingualism on inhibition. It is assumed that bilinguals need more mental effort for language inhibition, which engages a stronger activation of the left inferior frontal gyrus (IFG) in semantic inhibition (Van Heuven et al., 2008) and the dorsolateral and inferior frontal lobe in phonological inhibition (Rodríguez-Fornells et al., 2002). In a Magnetoencephalography (MEG) study, Bialystok et al. (2005) asked the participants to perform a Simon task (a classic inhibition task) when MEG was recorded. The bilingual group showed faster reaction times than the monolingual group when performing the Simon task. Moreover, greater brain activation was observed in the left anterior cingulate cortex and the middle frontal gyrus for the monolinguals, areas that are classically implicated in cognitive control. However, the left IFG, which is more associated with language, shows stronger activation in bilinguals compared to monolinguals when they perform a Stroop task (Coderre and van Heuven, 2014; Grundy et al., 2017). Similarly, in a functional magnetic resonance imaging (fMRI) study, Luk et al. (2010) tested monolingual and bilingual young adults with the flanker task, another classic task that involved conflict monitoring and inhibition. Different neural patterns were observed for the monolinguals versus bilinguals: the bilinguals relied more on inferior frontal and subcortical regions, while the monolinguals more on middle frontal areas and temporal-parietal regions. This pattern is consistent with the MEG findings by Bialystok et al. (2005), suggesting that bilingualism might influence the neural recruitment during inhibition.

The notion of cognitive flexibility refers to the ability of switching between different cognitive tasks (Crone et al., 2004; Davidson et al., 2006). It is engaged in the control of attention and selection of information and in the distribution of cognitive resources when two or more cognitive tasks are shifting in one task session (Branzi et al., 2015). Cognitive flexibility, assessed by switching task paradigms, occurs when one task is switched to another. Typically, the task is characterized by two types of trials: switching or alternating, in which the response in a current trial is different from the following trial, as compared to the response in a repetition trial which is the same as the next trial (Monsell, 2003). Many studies have shown that people with more bilingual experience could switch between tasks by allocating cognitive resource between tasks more efficiently to select the correct response (Abutalebi and Green, 2007; Luk et al., 2012; Hernández et al., 2013; De Baene et al., 2015).

Neuroimaging studies have also used nonverbal switching tasks to investigate the effect of bilingualism. For example, Kovelman et al. (2008) showed that the activation of left IFG in bilinguals was stronger than monolinguals when performing sentence judgments in the switching context. The differential activation for bilinguals and monolinguals prompted researchers to ask whether there were domain-general cognitive control differences between bilinguals and monolinguals. For example, the study by Garbin et al. (2010) revealed that in a non-verbal switching task, bilinguals showed a smaller switching cost but more activation for switch trials in the left inferior frontal cortex, a region recruited in semantic processing. By contrast, another study with a similar non-verbal switching task showed that the monolinguals demonstrated larger switch costs and stronger activation in the ACC, the main region for domain-general cognitive control including monitoring and inhibition (Abutalebi et al., 2013). Likewise, consistent findings were reported in Rodríguez-Pujadas et al. (2013): in the non-linguistic switching tasks, bilinguals had more activation of the left inferior frontal gyrus, regions responsible for language processing, whereas the monolinguals showed larger activation of ACC in cognitive processing. These results suggested that the accumulation of bilingual experience could alter bilinguals’ neural activity during switch tasks.

According to previous studies, working memory is responsible for updating information, interacting to produce complex cognition while encoding incoming information and monitoring it for relevance to the task at hand (Miyake et al., 2000; Shipstead et al., 2015). Compared to the large number of studies that investigated inhibition and cognitive flexibility, only a few studies have identified the effects of bilingualism on working memory. Bialystok et al. (2014) examined the working memory function of monolingual and bilingual young adults, and found that there was no significant difference in the reaction time on an N-back task (a classic working memory task) for the two language groups. The result implied that bilingual experience has no impact on working memory. Similar results were reported in other studies (Ratiu and Azuma, 2015; Paap et al., 2015). Reaction times of bilingual children showed no difference in working memory span task, compared with monolingual children. In addition, researchers also found no significant difference in accuracy between bilingual and monolingual children. These studies indicated bilingual effect might be limited to only certain components of cognitive control, not including working memory (Bialystok, 2017).

Taken together, empirical work has been done to identify the impact that bilingual experience has on specific components of cognitive control. Recently, Paap and Greenberg (2013) cast doubts on the effect of bilingualism, showing that no overall language group contrasts were significant across various executive function tasks. It is important to note that while the findings on inhibition and cognitive flexibility are more consistent, the findings on working memory are very mixed and weak. This is likely due to the fact that working memory is not a reliable or stable factor underlying bilingual language experience. For example, some behavioral studies have indicated no working memory difference between bilinguals and monolinguals, especially in healthy adult population (Bialystok, 2010; Bonifacci et al., 2011; Duñabeitia and Carreiras, 2015; Paap et al., 2016). It is likely that working memory ability reaches a peak level at a time when second language learning occurs during adolescent and college years (as are the participants in our study). This suggests that in adults, knowing two languages did not guarantee an advantage on the capacity of working memory over monolingual peers. These findings lead us to assume that bilingualism does not convey their advantage in working memory abilities.

Most brain imaging studies that have examined bilingualism effects on cognitive control have relied on task-dependent fMRI. In recent years researchers have also begun to examine spontaneous neural activities during language and cognitive processing, with the aim of exploring the intrinsic correlations of relevant regions that are often confounded by different tasks in different studies that rely on task-based MRI methods.

There has been increasing evidence that coherent intrinsic brain correlation is an important feature of healthy brain functioning (Fox and Raichle, 2007; van den Heuvel et al., 2009). Resting-state fMRI (rs-fMRI) has been an indispensable tool to capture the coherent activity of related brain areas (Van Dijk et al., 2010). Resting-state functional connectivity (rsFC) provides a way to investigate the changes on functional connectivity, particularly in terms of functional integration and correlation in a brain network (Marrelec et al., 2008; Fjell et al., 2015; Shinkareva et al., 2010). Moreover, with the comparison of different language groups, it is possible to examine the influences of bilingualism in the cognitive control networks and how these influences are correlated with behavioral performance.

Luk et al. (2010) used both resting-state and task-related MRI to assess the effects of bilingualism on domain-general cognitive control. Results of resting-state MRI illustrated that bilinguals showed increased functional connectivity between parietal, temporal and subcortical regions, regions less related to executive functions. By contrast, monolinguals demonstrated increased connectivity between frontal and precentral regions, which were crucial in cognitive control. Their task-related MRI patterns revealed that bilinguals recruited frontal, temporal and subcortical regions in non-verbal cognitive tasks but monolinguals activated frontoparietal regions. But in a further study, Grady et al. (2015) focused on three main networks related to cognitive control: default mode network (DMN) (Dang et al., 2012), frontoparietal control network (FPN) (Spreng et al., 2013) and salience network (SLN) (Seeley et al., 2007), in order to test whether bilingualism alters the brain's functional connectivity in resting state. Results indicated a greater functional connectivity in both the DMN and the FPN for bilinguals, but no significant effect in the salience network between the bilinguals and monolinguals. Similar results were reported by Berken et al. (2016) that bilinguals had significantly higher connectivity in the DMN and the FPN than the monolinguals, with no differences in other networks not involved in executive function.

Converging results from both resting-state and task-related MRI suggest that bilinguals engaged different cerebral regions and networks from monolinguals when performing executive function tasks. Importantly, these differences were closely associated with domain-general cognitive control that appear to be affected by bilingualism. However, these results are not always consistent with each other (e.g., the role of the frontoparietal network in Luk et al. (2010) vs. Grady et al. (2015)). In addition, in some studies, neuroimaging differences between bilinguals and monolinguals are reported without accompanying behavioral data. Some researchers have argued that brain results would not be interpretable in the absence of behavioral data (Paap et al., 2015). García-Pentón et al. (2016) also highlighted the importance to connect with behavioral data more closely, and argued that behavioral data are essential in the interpretation of neuroimaging data, especially for understanding the effects of bilingualism in an integrated view.

Although many empirical studies have identified the impacts of bilingualism on cognitive control, it is only recently that researchers start to pay attention to the variables that can account for individual differences, especially on the question whether bilingualism affect some people but not the others (Luk and Bialystok, 2013; Bialystok, 2017; Li et al., 2014a); One approach is to use the bilingual's L2 proficiency as a variable (e.g., high proficiency vs. low proficiency) to examine the impact of degree of bilingualism rather than bilingualism vs. monolingualism categorically for identifying the presence or absence of bilingual effects (Singh and Mishra, 2012, 2013; Weber et al., 2016). There has been ample evidence on L2 proficiency in modulating language activation patterns (Abutalebi, 2008; Abutalebi et al., 2014; Abutalebi and Green, 2007). In addition, L2 proficiency in language learners is often correlated with the performance level on linguistic tasks (Wong et al., 2007; Li et al., 2014a, Li et al., 2014b). Different proficiency levels are also associated with different fMRI patterns in brain networks, modulating the neural circuits of semantic access and cognitive control (Grant et al., 2015). These results might lead to a deeper understanding of whether L2 proficiency could affect the mechanism of cognitive control in a similar way.

Recent studies have also examined the influences of language proficiency on executive functions. Bialystok and Viswanathan (2009) tested high vs low L2 proficiency child bilinguals, and found that high-proficiency bilinguals performed better than low-proficiency bilinguals in both inhibition and non-verbal switching tasks. Similarly, Wang et al. (2016) examined the impact of language proficiency on adult bilinguals in cognitive control, reporting that high-proficiency adult bilinguals outperformed low-proficiency peers in both inhibition and cognitive flexibility. Goral et al. (2015) tried to examine the effect of L2 proficiency by comparing performances of old and adult bilinguals in the Stroop task. Their findings included age-related decline in inhibition but little or no age-related change in working memory and attention. These studies indicated the significance of using language proficiency to assess the effect on domain-general cognitive control.

Given that it remains unknown about the bilingual effects on cognitive control, especially the role of language proficiency, our study will take the following approaches to examine the intrinsic effect of second language proficiency on domain-general cognitive control. First, because bilingual researchers should take a more componential perspective to study bilingual effects (Bialystok et al., 2005, 2017; Dong and Li, 2015; Paap and Greenberg, 2013; Costa et al., 2009), the current study will focus on the three most studied components of cognitive control: inhibition, cognitive flexibility, and working memory. Second, our study will apply rsFC to explore the functional connectivity patterns that result from bilingual experience through brain connectivity analysis. Based on the meta-analysis of activation likelihood estimation (ALE) in Niendam et al. (2012), our study focuses on a set of key regions of interest (ROIs) in medial and middle frontal areas and cingulate cortex for cognitive control. We hypothesize that rsFC patterns between groups with different levels of second language proficiency would show distinctions in some components of domain-general cognitive control. In particular, Gold et al. (2013) previously showed that the degree of activation in the cognitive control regions was negatively correlated with task performance, i.e., less activation was associated with better performance, we suggest that a similar correlation could exist between task performance and rsFC.