Bilingualism and cognitive control:
does age of onset matter?
Student: Laura de Rooij Student ID: 11262451
MSc in Brain and Cognitive Sciences, University of Amsterdam Track: Cognitive Science
Supervisor: prof. dr. Enoch O. Aboh Co-assessor: dr. Judith Rispens Date: December 13, 2017
Abstract
Because bilinguals experience constant simultaneous activation of both languages, a cognitive control system is thought to develop to avoid interference between languages and guarantee selection of the target language. Given that bilinguals are assumed to resort to such control system constantly, it could be hypothesized that such an experience might result in a bilingual cognitive advantage in tasks that require similar control mechanisms. Much research has investigated such cognitive advantage in bilinguals, but since different factors of bilingualism may selectively contribute to the development of cognitive control, the present thesis focuses on the distinct influence of age of onset on the bilingual cognitive advantage. Drawing from both behavioural data and neuroimaging data, this thesis aims to shed more light on how findings of existing literature can guide future research in investigating the development of cognitive control in bilinguals. While neuroimaging data shows that early and late bilinguals have a different brain organization, this is not supported by behavioural studies, which show that differences in cognitive control between early and late bilinguals are more likely to result from the demand of the bilingual situation and how the languages are used. In essence, it seems that early and late bilinguals may be able to acquire similar cognitive control, and it thus remains an open question how a different neural substrate of language representation in late bilinguals may facilitate cognitive control. Possible directions for future research are also established.
1.
Introduction
Research has shown that, in bilinguals, both languages are simultaneously activated, even when the communicative setting requires the use of one language only (e.g. Martin, Dering, Thomas, & Thierry, 2009; Thierry & Wu, 2007). Given that two languages are activated while only one is used, it is claimed that bilinguals have a control mechanism that selects the appropriate language for specific contexts, thus avoiding intrusions from the competing language (e.g. Abutalebi & Green, 2007; Green, 1998). This language control and language selection process has been hypothesized to result from a domain-general process that is largely responsible for the control of information flow in our brain and for the selection of specific information. In bilinguals, such one over-trained cognitive control system results from the need to solve cross-language activation, and is thought to also transfer to related non-verbal activities for which similar control mechanisms are required.
Research suggests that, as a result, bilingualism leads to advantages in tasks that require conflict resolution or coping with competing stimuli (e.g. Bialystok & Martin, 2004; Carlson & Meltzoff, 2008; Costa, Hernández, Costa-Faidella, & Sebastián-Gallés, 2009; Costa, Hernández, & Sebastián-Gallés, 2008; Bialystok, Craik, Klein, & Viswanathan, 2004). However, it has been shown that several variables are at play (i.e. age of exposure to the second language, proficiency, how the languages are used) and the term “bilingualism”, therefore, comprises a variety of experiences. In order to find out how the aforementioned variables may differentially influence the cognitive benefits of bilingualism, all potentially confounding factors must be taken into account. Aiming to provide a clearer picture on how bilingualism influences cognitive abilities, the present thesis focuses on the question whether different ages of exposure to the second language matter to reap the cognitive control benefits of bilingualism.
In chapter two, I will first describe how one understands bilingualism from the perspective of cognitive control. Then, I will discuss how bilingualism influences cognitive control and what factors of cognitive control are thought to be most important in the debate on the bilingual advantage. However, because the literature does not provide consistent results on the proposed bilingual cognitive advantage, some researchers have posited that bilingualism does not confer positive effects on cognitive control. I will then shortly discuss this. The last part of chapter two comprises evidence from neuroimaging studies to argue that bilingualism does have cognitive benefits. In the third chapter, I will briefly touch upon the different variables of experiences of bilingualism that are seemingly important in determining the extent to which the bilingual develops enhanced cognitive control. I will also introduce the research question of the present thesis in this chapter: whether age of onset has a distinct influence on the development of
bilingual cognitive advantages. In the fourth chapter, I will give the literature review. This thesis draws from both behavioural data and neuroimaging studies, which are provided in separate sections in the literature review. The results of the thesis are discussed in the fifth chapter.
2.
Theoretical background
2.1 Bilingualism from the perspective of cognitive control
In order to understand the implications of bilingualism for cognitive control, it is important to describe some psycholinguistic models that explain how language control and language selection work in bilingual situations.
Consider a speaker that intends to name an object or to express an idea. Some psycholinguistic models (e.g. Caramazza, 1987; Levelt, Roelofs, & Meyer, 1999) have proposed that, before the speaker can do either, s/he has to select the appropriate conceptual representations that are associated with this idea or object, after which the grammatical and syntactic properties of the word are activated, and the correct phonology is accessed. Only then can the word be articulated. In a situation where the bilingual speaker only needs one language to convey an idea, one could imagine that the other language does not need to be activated. Yet, research has shown that both languages share the same semantic system (Costa, Miozzo, & Caramazza, 1999; Kroll & Stewart, 1994; Poulisse & Bongaerts, 1994), with the consequence that when one language is activated, the other language is automatically activated too. As a result, the bilingual speaker experiences more processing costs in comparison to a monolingual speaker: two lexical entries are activated, and the speaker has to decide which lexical entry is most appropriate. Evidence for the theory that both languages are jointly activated in bilinguals, even when only one language needs to be used, comes from psycholinguistic studies. Tasks that are often used are the lexical-decision task, in which participants must decide whether a string of letters is an existing word in either language, and the cross-language priming task, where a word in one language triggers retrieval of a semantically related word in the alternate language. Studies employing these tasks have shown that the non-target language influences the target language for both production and comprehension of speech1. Further evidence that the non-target language is
still involved when the bilingual performs a linguistic task in one language comes from neuroimaging studies (Martin et al., 2009; Thierry & Wu, 2007; Rodriguez-Fornells, Rotte, Heinze, Nösselt, & Münte, 2002; Kroll, Bobb, & Wodniecka, 2006). For example, in the study of Martin et al. (2009), Welsh/English bilinguals were presented with words in both their languages and were instructed to make word length decisions on words in one language while ignoring words in the other language. They used a low-level letter counting task to prevent participants from deliberately accessing the semantic content of the word. Martin et al. used event-related
Sebastián-potentials (ERPs) to get an indication of their neural response to the integration of the meanings of these words. This neural response is manifested in a negative waveform about 400 milliseconds after the stimulus has been presented, hence, the term N400. Earlier research has indeed shown that the N400 is associated with semantic integration and priming (Kutas & Hillyard, 1980; Kutas & Van Petten, 1988; Brown & Hagoort, 1993). Martin et al. found that the N400 amplitude was significantly lower when words were semantically related, independent of performance in the task, which was evidence for Martin et al. to conclude that accessing the meaning of a written word happens automatically in both languages. In another study by Thierry and Wu (2007), English monolinguals, Chinese-English bilinguals, and Chinese monolinguals were presented pairs of English words (for the English monolinguals and the Chinese-English bilinguals; the English words were translated for the Chinese monolinguals). The participants then had to indicate whether the words presented were semantically related. Half of the pairs contained a repeated character in the written Chinese forms, even though the orthographic feature was unrelated to the English meaning. For example, the words train and ham are semantically unrelated, but their translations in Chinese, respectively Huo Che and Huo Tui, share a character. They found that for English semantically related words, the N400 amplitude was significantly smaller in all groups; and intriguingly, for half of the pairs that contained a repeated character in the written Chinese forms, both Chinese groups also showed a smaller N400 response. Thierry and Wu concluded that the English-Chinese bilinguals were accessing the Chinese forms when judging the semantic relatedness between English words, even though it was unnecessary for successful performance in the task.
Because the two languages are jointly activated in bilinguals, there are consequences of bilingualism for both linguistic and nonlinguistic processing. When two lexical entries are activated for one specific concept, the bilingual speaker encounters a problem that is not present for monolinguals: the correct language needs to be selected from two competing options. Despite the fact that joint activation of two languages creates possible language interference, intrusions from the alternative language rarely occur, which indicates that bilinguals possess a mechanism that is extremely accurate in selecting the target language. It is thought that bilinguals develop a general cognitive control system that is primarily responsible for driving attention toward the target language, facilitating fewer intrusions from the competing language. The question then arises how the desired language is selected over the other language. In the literature, the debate revolves around whether or not there is between-languages competition and inhibition of the unwanted language (e.g. Kroll & Gollan, 2014). One model that lies at the heart of this debate is
Green’s (1998) inhibitory control theory, which much work on the bilingual advantage in cognitive control is based on.
This theory was based on the assumption that a particular experience activates semantically related units in the bilingual’s languages. Before the speaker can retrieve the word, one such lexical entry (also referred to as “lemma”) must be actively suppressed. The supervisory attentional system (SAS; Norman & Shallice, 1986), or executive functioning system, is then responsible for the selection process, which is dependent on the extent to which the lexical entries from the target and non-target language compete with each other. Stronger inhibition is needed when both lexical entries are strongly activated, but if contextual cues clearly indicate which language needs to be produced or perceived, the SAS needs to exert less inhibitory control. Consistent with Green’s inhibitory control are the findings that switching from a dominant language (e.g. L1) to a nondominant language (e.g. L2) is more taxing than vice versa. For example, Meuter and Allport (1999) showed that bilinguals are slower in naming a digit in their L1 after having named a digit in L2, than the other way around. These findings were interpreted by Meuter and Allport as indicating that stronger inhibitory control is needed to suppress the dominant language, than that required to suppress their nondominant language. This inhibitory control theory is also consistent with the hypothesis that the SAS, which is responsible for resolving conflict between two jointly activated languages, may also mediate nonlinguistic (cognitive) control processes.
In sum, psycholinguistic models of language selection, in combination with evidence that both languages are simultaneously activated in bilinguals, explain why bilinguals must exert more linguistic control than monolinguals in daily interactions. Because nonlinguistic inhibitory control is needed to suppress the unwanted language, it has been hypothesized that bilingualism has several cognitive implications. These cognitive implications of bilingualism have been the subject of a great number of studies, which I will address in the next section. The main findings regarding the influence of bilingualism on cognitive control are elaborated on, including the most important factors of cognitive control that are thought to be affected by bilingualism.
2.2 What influence does bilingualism have on cognitive control?
It has been shown that bilinguals outperform monolinguals on a number of tasks that serve to evaluate individual differences in executive functioning: an umbrella term for executing, maintaining and adjusting goal-directed behaviour (e.g. Banich, 2009). Following the well-accepted tripartite model of executive control (Miyake & Friedman, 2012; Miyake et al., 2000; Shallice & Burgess, 1996), the term executive control refers to one’s capacity to inhibit specific
information, to be flexible in switching between tasks, and to update information. According to Miyake & Friedman (2012), inhibition refers to the deliberate overriding of dominant responses; shifting refers to the capacity to flexibly switch between mental sets or tasks; and updating entails constant monitoring and the rapid deletion or addition of working-memory contents. Following the hypothesis that these general executive control mechanisms are used in a more taxing manner in bilingual language processing relative to monolingual language processing, many studies have investigated how inhibition, shifting and monitoring develop differently in bilinguals.
Most research has employed attentional paradigms in which participants are presented with conflicting information that must be ignored to successfully perform in the task. Tasks that are often used are the Simon task and the flanker task (the latter is often inserted in the Attentional Network Task, ANT; Fan, McCandliss, Sommer, Saz, and Posner, 2002). These tasks consist of stimuli that are associated with a certain response, presented in either a congruent trial or in an incongruent trial. In congruent trials, there is no competing information irrelevant for the task; in other words, the stimulus information is compatible with the correct response. However, in incongruent trials, the stimulus information is incompatible with the correct response, i.e. there is information that the participant actively has to suppress to achieve successful performance. For example, in the flanker task (Eriksen & Eriksen, 1974), participants must indicate which direction the target arrow points at. This target arrow is surrounded by other arrows, which point in the same direction in the congruent trial, or in opposite direction in the incongruent trial. In the Simon task, participants must associate a response with a certain stimulus, i.e. they must press right if the stimulus is in one colour, and left if the stimulus is in another colour. In congruent trials, the response associated with the stimulus is compatible with the location where the stimulus appears, whereas in incongruent trials, there is no correspondence between the response they should make and the location where the stimulus appears (either on the left or the right side of the screen). Studies use difference in response times (RT) and accuracy levels between congruent and incongruent trials to indicate how bilingualism influences inhibitory control. Several studies have indeed reported that bilinguals from different age groups score higher than their monolingual peers in tasks where they are required to cope with competing stimuli (e.g. Bialystok & Martin, 2004; Bialystok & Majumder, 1998; Bialystok, Craik, Klein, & Viswanathan, 2004; Colzato et al., 2008; Costa, Hernández, & Sebastián-Gallés, 2008). Such studies support the hypothesis that bilingual language selection is to some extent mediated by inhibitory control, and findings like these have been interpreted as showing that bilinguals have enhanced cognitive control. However, it has also been shown that bilinguals
outperform monolinguals in tasks where there is no conflicting information, for which no inhibitory control is necessary.
For example, a study by Costa, Hernández, & Sebastián-Gallés (2008) showed that bilinguals outperformed monolinguals on the attentional network task (ANT). They found a bilingual advantage on incongruent trials, during which inhibitory control is needed for successful performance, but bilinguals also performed better on congruent trials. Similarly, Bialystok et al. (2004) found that older adults also showed a bilingual advantage for both incongruent trials and congruent trials on the Simon task. This is difficult to account for in terms of conflict resolution or inhibition, because it is unnecessary to exert any inhibitory control in congruent trials in both the flanker and the Simon task. The fact that bilinguals outperform monolinguals not only in conflict resolution tasks, but also in their overall RT measure has led researchers to hypothesize that enhanced inhibitory control cannot solely account for the results. It has been argued that the monitoring aspect of executive control might be more important in that bilinguals may be better at maintaining new information in their working memory, focusing on task-relevant information and ignoring task-irrelevant noise (e.g. Bialystok et al., 2004; Bialystok, Craik, Green, & Gollan, 2009).
However, this bilingual advantage in monitoring is not always apparent in young adults, complicating the hypothesis that bilinguals are overall better in monitoring. Other studies, for example, show faster global RTs for children and older adults in both the congruent and incongruent condition in the Simon task, but not for young adults (Bialystok et al., 2005; Salvatierra & Rosselli, 2010). It has been proposed that young adults are only at an advantage in monitoring processes when the task or condition is difficult. For example, a study by Costa, Hernández, & Costa-Faidella, & Sebastián-Gallés (2009) presented both bilingual and monolingual participants with a flanker task under different task versions: low-monitoring versions, where the congruent and incongruent trials were not evenly distributed (congruent trials appeared 92% of the time, and incongruent trials appeared 8% of the time), and high-monitoring versions, where the trials were more evenly distributed (both appeared 50% of the time). Costa et al. found that bilinguals performed faster in the high-monitoring versions of the flanker tasks, whereas bilinguals and monolinguals were comparable in terms of speed in the low-monitoring versions. They concluded that this general bilingual advantage in RTs indicates an enhanced ability of bilinguals to go back and forth between trials that require conflict resolution and trials that do not need inhibitory control because they have a more efficient monitoring system. This means that bilinguals can more easily determine whether misleading information can be ignored
or not. The activation of bilinguals’ monitoring system might depend on specific properties of the task, specifically the degree to which the experiment consists of different kinds of stimuli. Other studies have reported bilingual advantages in tasks that require the participant to shift between tasks, but without competing stimulus information. For example, a study by Prior & MacWhinney (2010) investigated the extent to which bilinguals have a more enhanced ability to shift between mental tasks. As Morales, Gómez-Ariza, and Bajo (2016) put it: “Given that
bilinguals often need to switch between languages, they may also engage control processes related to shifting the task goals, inhibiting the previous target, and redirecting attention to a new one.” The task that is usually
employed to evaluate individual differences in mental set switching is the task switching test. In the standard trials, participants must attend to one feature of the stimulus, i.e. shape or colour. In mixed blocks, participants are required to do the same, but the feature they must attend to may be different in every trial: they must follow instructions after each trial. This creates different kinds of trials in mixed blocks: participants either attend to the same feature in repetition trials, or they attend to different features in switching trials, where the task may change from trial to trial. Prior & MacWhinney found a bilingual advantage in switching costs in comparison to monolingual participants, as reflected by the differences between repetition trials and switching trials. However, there was no bilingual advantage in mixing costs, as reflected by the differences between mixed blocks and the standard blocks, indicating that bilinguals do not necessarily have enhanced attentional control. Later studies replicated these findings (e.g. Prior & Gollan, 2011; Soveri, Rodriguez-Fornells, & Laine, 2011). These researchers concluded that bilingual experience seems to influence the capacity to flexibly shift between task goals, but it does not affect monitoring. In the context of these inconsistent results on bilinguals’ monitoring and inhibitory abilities, Prior (2012) contrasted the two components of executive control in a 2-lag non-verbal switching task. Prior reported no difference between bilinguals and monolinguals for monitoring, but bilinguals did show enhanced inhibitory control.
Though these studies support the hypothesis that language selection and control influence non-linguistic inhibitory processes, bilingual advantage does not systematically appear. It has been suggested that different tasks could simply reflect different aspects of inhibitory control (Morales, Gómez-Ariza, & Bajo, 2016). Because the term “inhibition” comprises different processes and functions (Friedman & Miyake, 2004), different conflict resolution tasks might produce non-identical conflicts. For example, in the flanker task, the participant is required to suppress interference produced by stimulus information, whereas in the Simon task, participants must suppress a response. Because a bilingual is presented with language co-activation, bilingual advantages may be more likely to appear in tasks where similar mechanisms
of conflict resolution are required. Based on the language control mechanism bilinguals must have to cope with language co-activation, bilinguals may have a more enhanced ability to ignore competing information than to inhibit unwanted responses. Indeed, research that has tried to dissociate these two mechanisms has found more evidence that bilingual control is related to ignoring conflicting information than inhibition of unwanted responses (Bialystok & Martin, 2004; Bialystok & Viswanathan, 2009; Blumenfeld & Marian, 2013; Colzato et al., 2008; Esposito, Baker-Ward, & Mueller, 2013; Luk, Anderson, Craik, Grady, & Bialystok, 2010; Martin-Rhee & Bialystok, 2008). However, findings on the source of the bilingual advantage remain inconsistent, as evidence does not point towards one responsible single component of executive functioning. Instead, it seems that the influence of bilingualism on executive control mechanisms is dependent on the continuous monitoring of two languages, as well as the need to attend to the context, while inhibiting interference from the non-target language (Bialystok, Craik, & Luk, 2013). Another potentially responsible factor for inconsistent results is the different nature of bilingualism of participants. Studies are designed to contrast monolinguals and bilinguals, but these two groups cannot easily be compared because participants have different gradations in i.e. their proficiency, age of exposure of the second language, and use of their languages (how often they code-switch in everyday life).
By and large, studies have different findings on how executive functioning may develop in bilinguals. Though some researchers have argued that inhibitory control is the source of bilingual advantage, others have found evidence that both monitoring and task switching are components that bilinguals also outperform monolinguals on. Additionally, the claim that inhibitory control comprises different processes and functions has led to the hypothesis that research should distinguish between response suppression and interference suppression. Because studies have different findings on how executive functioning may develop in bilinguals, some researchers have questioned whether bilinguals even have superior cognitive control.
2.2.1 Does bilingualism confer any positive effects at all?
The inconsistency of research in establishing the source of the bilingual advantage and to what extent bilinguals have improved cognitive performance has led researchers to question bilingual influence on executive functioning processes.
Firstly, it seems that bilinguals do not only score better in tasks that require executive control mechanisms, but also in other tasks. For example, Morales, Gómez-Ariza & Bajo (2016) summarize that bilinguals are more creative (Hommel & Colzato, 2009; Ricciardelli, 1992), have better episodic memory (Ljungberg, Hansson, Andrés, Josefsson, & Nilsson, 2013), and they
outperform monolinguals in cognitive flexibility tasks (Bialystok, Craik, & Ruocco, 2006; Bialystok & Martin, 2004). Secondly, recent reviews have argued that bilingual cognitive advantages are not systematically found (e.g. Blumenfeld & Marian, 2007; Hilchey & Klein, 2011; Paap & Greenberg, 2013; Paap, Sawi, Dalibar, Darrow, & Johnson, 2014; Tao, Marzecová, Taft, Asanowicz, & Wodniecka, 2011).
An important paper was published by Paap & Greenberg (2013), in which they raised critical questions about the existence of any bilingual cognitive effects. They conducted three studies in which they compared bilinguals to monolinguals on different indicators of executive control, aiming to see whether individual differences in performance on one indicator of executive processing in one task (i.e. inhibitory control in the Simon task), correlated with performance of the same indicator of executive control in another task (i.e. inhibitory control in the flanker task). However, they found no main effect of group and a significant main effect of condition. This study confirmed that performance on inhibitory control in one task does not necessarily predict performance on that same indicator in a different task, leading Paap & Greenberg to conclude that studies testing participants’ performance in only one task do not provide coherent support that bilinguals are at a cognitive advantage. Paap & Greenberg undermined the interpretation that inhibitory control, shfiting and monitoring are valid indicators of executive processing, and generated a lively discussion in the field whether bilingualism confers cognitive advantages. Other work was undertaken to further elucidate the possible failures in bilingual research.
For example, Hilchey & Klein (2011) reviewed a number of studies testing the hypothesis that bilinguals’ continuous use of their executive control system results in advantages in their non-linguistic inhibitory control. They reviewed thirteen studies that employed either the Simon task, flanker task or flanker interference with a Simon-like component.2 Though at the core of
these studies, they might investigate the same component of executive functioning, namely inhibitory control, different designs were used. From the data gathered from the reviewed studies, Hilchey & Klein concluded that few experiments reported demonstrably large bilingual advantages. Specifically the magnitudes of the interference effects between monolinguals and bilinguals were similar for both children and young adults, inconsistent with the view that bilingualism has benefits for inhibitory control processes. However, Hilchey & Klein argued that there seems to be a consistent bilingual advantage in global RTs. While Paap, Sawi, Dalibar,
2 Simon task: Bialystok et al. (2004); Bialystok et al. (2005); Bialystok, Martin, & Viswanathan (2005); Bialystok (2006); Morton & Harper (2007); Martin-Rhee & Bialystok (2008); Bialystok et al. (2008); Bialystok & DePape (2009). Flanker task: Costa, Hérnandez, & Sebastián-Gallés (2008); Carlson & Meltzoff (2008), Costa, Hérnandez, Costa-Faidella, & Sebastián-Gallés (2009); Luk, Anderson, Craik, Grady & Bialystok (2010). Flanker interference with a Simon-like component: Emmorey, Luk, Pyers, & Bialystok (2009).
Darrow, & Johnson (2014) in later work refer to studies incompatible with these findings (e.g. Antón et al., 2014; Dunabeitia et al., 2014), such broad empirical findings of a bilingual benefit in RTs have to be accounted for in terms of executive functioning control.
Other criticism of Paap & Greenberg (2013) of other studies’ results on a bilingual cognitive advantage was that the evidence in support for the bilingual cognitive advantage hypothesis is less obvious in larger samples. Studies with a smaller sample size may thus be misleading in generating false positive rates if the bilingual advantage seems to diminish while the power of the studies improves. However, as Button et al. (2013) mention: reducing sample sizes can even be more informative than larger sample sizes, because it allows researchers to control for potentially confounding factors, whereby, in turn, more valid conclusions can be inferred from the study conducted.
It is clear that, although the literature does not provide consistent results on a possible bilingual advantage in executive functioning control, empirical findings support the hypothesis that language control and selection influences domain-general executive control processes to some extent. Studies employing neurophysiological measures further provide evidence that bilinguals use their brains differently from monolinguals, which I will discuss in the next section.
2.3 What influence does bilingualism have on brain structure?
In the endeavour of understanding the cognitive benefits of bilingualism and the brain structure that underlies this process, it is important to look at studies employing neurophysiological measures. Several studies have investigated neural changes in connection with the hypothesis that bilinguals have cognitive advantages. Such neural changes can be achieved in different ways: through a change in white matter (WM) connectivity (i.e. connectivity of the network), through a change in grey matter (GM) density (i.e. capacity), or through a change in the responsiveness of neurons (i.e. regional efficiency). Due to space limitations, this thesis cannot include an exhaustive literature review. I will therefore focus on the most relevant findings for the current discussion.
As Bialystok, Craik & Luk (2013) mention, most neurophysiological research on bilingualism has used functional magnetic resonance imaging (fMRI) to study bilinguals performing a linguistic task in their two languages. Participants are usually required to either name pictures or pronounce words in response to a cue which language is required, after which performance for single and mixed language is compared. Hernandez, Martinez & Kohnert (2000) conducted one of the first fMRI studies, in which they found that the dorsolateral prefrontal cortex (DLPFC) was activated when participants switched between languages. This finding was
surprising, because the DLPFC is known to be an important area for the general executive control system. Subsequent research has shown that more regions are involved in language switching: bilateral frontal regions, bilateral precentral areas, bilateral caudate, bilateral pre-supplementary areas (pre-SMA), and bilateral temporal regions (Abutalebi, Anonni, Zimine et al., 2008; Garbin, Gosta, Sanjuan, 2011; Guo, Liu, Misra, & Kroll, 2011; Hernandez, 2009). Moreover, the anterior cingulate cortex (ACC) is activated in both language switching and non-verbal switching tasks and is thought to be especially important for conflict monitoring and attention (Abutalebi, Della Rosa, Green et al., 2012; Li, Yang, Scherf, & Li, 2013). In other words, much important research has suggested that, in bilinguals, frontal regions important for executive control are activated in attending to both languages. Indeed, Hedden and Gabrieli (2010), and Toro, Fox, & Paus (2008) argued that there is an overlap in brain regions activated in both cognitive control and language switching in bilinguals.
Using both fMRI and transcranial magnetic stimulation (TMS), Nakamura et al. (2010) showed strong connectivity between the left inferior frontal gyrus (IFG) and the left middle temporal gyrus (MTG), which was stronger in the frontal-temporal direction than in the temporal-frontal direction. Nakamura et al. interpreted their findings as showing top-down control from left ITG to left MTG in bilinguals. Further research has provided evidence that specifically the left ITG seems to be an important brain region for bilinguals in their need to control both language switching and non-linguistic processes. Garbin et al. (2010) presented monolingual and bilingual young adults with a colour-shape switching task. The participants saw a stimulus, i.e. a red circle, after which they were cued to respond to either the “colour” or the “shape”. The behavioural data showed that bilingual participants were at an advantage for both RT of switch costs and accuracy, and the neurophysiological data showed different patterns of activation for both groups of participants. Whereas the right IFG was more activated in monolinguals, in bilinguals this was the case for the left IFG. Most importantly, they found that higher levels of activation in the right IFG were associated with larger switch costs for monolinguals, while higher levels of activation in the left IFG were associated with smaller switch costs for the bilinguals.
Another area that has been implicated as an essential area for bilingualism is the left inferior parietal lobule (IPL), which plays a role in lexical learning, semantic integration, and phonological working memory (Baddeley, 2003; Della Rosa et al., 2013; Mechelli et al., 2004). Intriguingly, grey matter (GM) density in the IPL can expand as the learner in the second language (L2) becomes more proficient; however, it is negatively correlated with the learner’s age of L2 acquisition. Della Rosa et al. (2013) assessed bilinguals’ language competence and executive
control abilities to establish the relationship between neural changes and behavioural data. They found that there was a strong positive correlation between language competence and cognitive control as a function of development due to bilingual language experience. Thus, it was argued that as a result from the cognitive demands of the continuous co-activation of two languages, GM density can increase in the IPL, an area thought to be important for both multilingual talent and the executive control ability (Mechelli et al., 2004; Richardson & Price, 2009). Grogan et al. (2012) have also replicated this finding and confirmed that the IPL and other regions in the temporo-parietal cortex are important for bilingualism.
Changes in cortical GM density have been the subject of a number of other studies, which have shown that bilinguals have more GM volume in the bilateral cerebellum (CB; Pliatsikas, Johnstone, & Marinis, 2013), and the caudate nucleus (CN; Grogan, Green, Ali, Crinion, & Price, 2009; Zou, Ding, Abutalebi, Shu, & Peng, 2012). For example, Grogan et al. (2009) showed that larger GM density in the CN correlated with enhanced performance of high-proficient bilinguals in a phonemic fluency task administered in their L2. Because the CN is hypothesized to be important for recognizing phonological anomalies and language switching, among other things, Grogan et al. interpreted their results as indicating that the need to inhibit the first language (L1) while performing L2 production results in higher activation of the CN. One recent study by Luk, Bialystok, Craik, & Grady (2011) has investigated bilingual influence on the plasticity of WM. They measured resting-state functional connectivity in monolingual and bilingual older adults by using diffusion tensor imaging (DTI). They showed that bilingual older adults had primarily higher WM integrity in the corpus callosum, which connects the two hemispheres. Additionally, it was found that bilinguals had increased anterior-posterior connectivity, which was claimed to suggest that WM structures are better maintained throughout the course of healthy cognitive aging in bilinguals (Davis, Dennis, Buchler, White, Madden, & Cabeza, 2009).
In sum, it seems that the nature of bilingual advantage lies in the fact that the need to manage two co-activated languages seems to enhance frontal-posterior attentional control mechanisms, which, in turn, leads to more efficient other types of cognitive control. We now have enough evidence to conclude that bilingualism confers benefits on executive functioning control. However, as mentioned earlier, the term “bilingualism” does not refer to a homogeneous group of speakers; therefore, we cannot evaluate the extent to which speaking more than one language affects cognitive control if we do not take into account the different factors that come into play in bilingualism.
3.
Important factors in the bilingual experience
The studies reviewed up until this point focused on a small set of bilinguals, i.e. bilinguals who had an early age of exposure to the second language, who were highly proficient in speaking their second language, and who often switched between their languages on a daily basis. However, each of these factors (age of onset, proficiency and use of the languages) may differentially influence the effect of bilingualism on cognitive benefits, indicating that bilingualism should be treated as a continuous experience rather than a categorical one (Bialystok, Luk, & Craik, 2013). Kroll and Bialystok (2013) argued for taking a multivariate approach to reveal how different types of bilingual experiences may have different consequences for cognitive control processes, and to account for the inconsistencies in the literature so far on bilingual cognitive advantage. Green and Abutalebi (2013) recently provided one such framework that specifies how different bilingual situations probably demand different control mechanisms because the language control processes the bilingual needs to exert in a situation depend on the demands of the language context. Green and Abutalebi considered three interactional contexts to establish demands of different conversational exchanges (single-language context, dual-language context, and dense code-switching contexts), which probably differ in how interference is resolved. For example, in the dual-language context, where both languages are used but with different speakers, the bilingual must control goal maintenance, interference suppression and conflict monitoring. However, in the dense code-switching context, where speakers can use both languages in the same utterance and adapt words from one language in the context of the alternative language, demand on opportunistic planning is higher. In short, because each of the contexts may require different control mechanisms to be involved, cognitive control dynamics are influenced differently in different bilingual situations.
Similarly, the more proficient the bilingual gets in both languages, the more inhibitory control s/he may need to exert in order to prevent language errors, and the better s/he thus might be at suppressing irrelevant information. This was indeed reported by Blumenfeld & Marian (2013) who recently looked at the relationship between performance on a spoken word-recognition language task, and a cognitive Stroop task. They found that bilinguals who were more proficient in their L2 had increased activation of both languages in the word-recognition task, and smaller Stroop effects as compared to bilinguals who were less proficient in their L2. They included a time-course analysis that revealed an enhanced ability of the more proficient bilinguals to resolve cross-language inhibition. These findings not only confirm that the same control network is activated during both the language task and the cognitive task, but also that individual
differences in inhibitory control are related to language control. The level of proficiency may thus differentially influence the dynamics of language control and its consequence for domain-general cognitive processes.
Of specific interest in the present thesis is the factor age of onset and its influence on the cognitive benefits of bilingualism. Under the hypothesis that the extent to which cognitive control must be exerted depends on the demand of the bilingual situation, and that practice in controlling language processes is important for developing better cognitive control, early acquisition of two languages might correspond with better executive functioning. Because early bilinguals have had more exposure and opportunity to practice with attending to two languages, the more proficient the early bilingual may be, the more inhibitory control he or she needs to exert, and the more enhanced his or her executive functioning abilities may be. However, if lifelong or early bilinguals are compared to late bilinguals and have equal proficiency and use their languages equally, the development of executive functioning may well be comparable. The present thesis tests this hypothesis by looking at the findings of existing literature. In order to establish distinct effects of the age of onset, it is necessary to critically look at studies that have investigated a possible difference in underlying mechanisms in early and late bilinguals. The following chapter reports on such studies in two different sections: behavioural studies on the one hand and neuroimaging studies on the other.
4. Literature review
Most bilingual research has tested either lifelong or early bilinguals, such that the bilingual had a very early onset of age when s/he first came in contact with the L2. There seems to be no uniform distinction at what age of onset speakers should be classified as being a late bilingual. Studies that have investigated late bilinguals have used different and somewhat arbitrary cut-off ages: most early bilinguals are individuals exposed to the L2 until ages three to seven, but other research has classified individuals who came in contact with the L2 at age ten as early bilinguals. Because the aim of the present thesis is to investigate the role of age of onset in the cognitive benefits of bilingualism, this thesis does not select studies based on the cut-off age that was used.
4.1 Behavioural studies
An important study was done by Pelham & Abrams (2014), who found that the habitual use of two languages may be essential for executive control development in bilinguals. They compared performance of English monolinguals (age = 19.3), early Spanish-Bilinguals (mean age = 20.3, and an age of onset before 7 years old), and late Spanish-English bilinguals (mean age = 22.0, and an age of onset after 13 years old) in the Attentional Network Task. The early and late bilingual group spent an equal amount of time speaking both languages and had comparable proficiency. It was found that both early and late bilinguals showed executive control benefits, specifically inhibiting interference from distractors, in comparison to the monolingual group. Pelham and Abrams concluded that the habitual use of two languages can result in bilingual benefits, whereas age of onset and thus the length of time being proficient in both languages is not of influence. As long as both early and late bilinguals have equivalent proficiency and usage of both languages, age of onset was claimed not to be an important factor in investigating executive control in bilinguals. However, it is also possible that the cognitive differences might not be very big because all speakers were relatively young and equally proficient in both languages.
Yow & Li (2015) answered the question raised above how proficiency might influence performance of young speakers. They found that proficiency and the age of onset predicted individual performance in different tasks involving either inhibitory control, task switching, or updating information and monitoring. They presented 72 English-Mandarin early bilinguals (mean age = 20.93) with different executive functioning tasks. All bilinguals learned their L1 and L2 simultaneously before the age of seven. Instead of controlling for proficiency, age of onset, and use of languages, they included these factors as continuous variables. It was found that age of acquisition correlated positively with interference costs in the response inhibition (Stroop) task. In other words, the later the bilinguals acquired their L2, the worse they were at inhibiting
unwanted responses. Furthermore, the more balanced the bilingual was (in terms of proficiency in both languages and the usage of the languages), the better he or she was at task switching. They claimed that it is habitual control of two language systems that facilitates the development of some executive functioning skills in bilinguals. Yow and Li claimed that some components of executive functioning, such as inhibiting unwanted responses and task switching, are more enhanced abilities in bilinguals, whereas other components are not, such as updating information and monitoring. However, since it has been claimed that a bilingual advantage is more difficult to find in young adults because they are at the peak age for cognitive control (e.g. Bialystok, Luk & Craik, 2013), Yow and Li’s conclusion that monitoring and updating information are not part of the bilingual cognitive advantage might turn out to be unsupported.
In fact, another study by Tao, Marzecová, Taft, Asanowicz, and Wodniecka (2011) found that both early and late bilinguals outperform monolinguals on conflict resolution, but it is age of onset that might be responsible in the emergence of efficient monitoring processes. They looked at three groups of young adults in Australia: monolingual English speakers (mean age = 20.4), early bilingual English-Chinese speakers (mean age = 18.9), and late bilingual English-Chinese speakers (mean age = 20.8). The early bilinguals learned English at or before the age of six, and reported a higher level of proficiency in English than Chinese, and a higher percentage of speaking English than Chinese on a daily basis. The late bilinguals learned English at or after the age of twelve, and reported a higher level of proficiency in Chinese than English, and a higher percentage of speaking Chinese than English on a daily basis. Self-ratings of use of both languages and proficiency indicated that the late bilinguals were more balanced than the early bilinguals. All participants were presented with an Attentional Network Task, in which both the early and late bilinguals showed to have more efficient executive functioning control in comparison to monolinguals. However, there were differences between the early and late bilinguals: whereas the early bilinguals outperformed late bilinguals on the monitoring component, the late bilinguals were better at conflict resolution. When proficiency and use of the languages were controlled for, the differences between the early and late bilinguals were greatly reduced. Because the late bilinguals were also more balanced than the early bilinguals, this is further evidence that the more proficient the bilingual is in his or her L1 and L2, and the more balanced s/he is in using both languages, the more executive functioning benefits may be derived.
As previously mentioned, much research has shown that bilingual advantage in conflict resolution seems be found in the overall RT measures and not just in trials that require resolution of conflict (see Chapter 2.2). Costa et al. (2009) proposed two hypotheses that account for these
findings: bilingualism may either independently influence both conflict resolution and monitoring systems, or, in contrast, the monitoring mechanism accounts for bilingual advantage in both overall RT and conflict resolution. While most research has shown that conflict resolution generally occurs with faster global RTs, the study by Tao et al. provides support for the latter hypothesis that the monitoring and conflict resolution mechanisms may develop independently in bilinguals. Tao et al. argued that late bilinguals may have to train their interference control ability to a larger extent than early bilinguals because their L2 is more entrenched, thus they are better at conflict resolution tasks. Likewise, as a result of early bilinguals having had lifelong exposure to both languages and opportunity to practice with the languages, early bilinguals are better at monitoring.
In order to be successful at monitoring, i.e. to upate novel information to the task goal, and to eliminate task-irrelevant noise, one must constantly focus on working memory representations (Miyake, Friedman, Emerson, Witzki, Howerter, & Wager, 2000; Osak, Nishizaki, Komori, & Osaka, 2002). Therefore, some studies have looked at WM in early and late bilinguals. Delcenserie & Genessee (2016) compared monolinguals, lifelong bilinguals, early sequential bilinguals (with an age of onset between 4 and 6 years old), and late sequential bilinguals (with an age of onset between 7 and 13 years old). All bilinguals were selected based on their levels of proficiency and general cognition to ensure all bilinguals were comparable. All participants were undergraduate university students. They found that early and late sequential bilinguals outperformed monolinguals on tests of both verbal and non-verbal WM, but they scored lower than the simultaneous bilinguals. Delcenserie and Genessee interpreted their findings as implying that, all other factors being equal, simultaneous (or lifelong) bilinguals reap the most WM advantages: because all groups were comparable in terms of proficiency levels, age of onset is responsible for the between-group differences. However, all other factors in the study of Delcenserie and Genessee were not equal. They took length of exposure into account, which turned out not to be statistically important for their analysis, but they did not control for how often those languages are used on a daily basis and in what interactional context. Because the simultaneous bilinguals learned two languages from birth, perhaps because their parents speak in different languages to them at home, it is possible that they were more exposed to both languages on a daily basis at the time of testing. If their situation requires them to constantly control both languages to avoid intrusions from the unwanted language, more so than early and late sequential bilinguals, this might result in lifelong bilinguals maintaining their language selection mechanism to a larger extent, which facilitates the development of working memory.
Perhaps the most convincing evidence of the influence of age of onset on bilingual cognitive advantage comes from Yang (2017), who presented participants with auditory and visual digit span tasks to measure participants’ WM. Three groups were compared: a Korean near-monolingual group, a Korean-English bilingual group with intermediate proficiency in L2, and a Korean-English bilingual group with high L2 proficiency. All participants were late sequential bilinguals who learned their L2 after the age of 10. The study controlled for the amount of language use, age of onset, level of L2 proficiency, and IQ. It was found that the intermediate group was better at remembering the order of the digits presented both in visual and auditory numeric forms, and they performed better at recalling the reversed order of digits in the auditory form. The high bilingual group did not differ in performance from the monolingual group, showing that only the intermediate bilingual group had a WM advantage. Based on interview data of the participants, Yang claimed that the highly proficient bilinguals had a lower demand of using strategies related to holding L2 information than the intermediate bilinguals had. Bilingual cognitive advantage may thus be primarily derived from the cognitive training that the bilingual is engaged in when using both languages.
Surprisingly, most, if not all, research on early versus late bilingualism has focused on young adults whereas studies with early and late older bilingual participants are relatively few and far between. One example, however, is the study by Keijzer & Schmid (2015), who investigated executive functioning advantages in eldery, long-term, immersed bilinguals (age 71+). Native Dutch speakers who emigrated to Australia as adults were compared to Dutch and English monolingual controls on a number of cognitive and linguistic measures. The late bilinguals did not outperform the other participant groups (neither the Dutch nor the English monolinguals), indicating that the late bilinguals had no bilingual cognitive advantage. Interestingly, Keijzer and Schmid did find that language use patterns were associated with the scores on the Simon task: participants who still spoke Dutch at home were better at conflict resolution. Some subjects used Dutch in the home, and spoke English in all other situations, which was hypothesized to have an effect on the extent to which they had a cognitive advantage. A drawback of this study then is the fact that some late bilinguals did not speak their language at home at all. It is commonly the case that when adults migrate to a new country, and do not use their L1 so often anymore, they may lose proficiency in their native language. If use of the languages is important for the bilingual cognitive advantage, we would need to investigate adults who use their two languages actively and for whom it can be assumed that their level of competence in L1 has not changed.
In sum, some studies showed that early and late bilinguals do have different executive control abilities, i.e. that early bilinguals have a better development of the monitoring component,
while late bilinguals are better at resolution of conflict. Other studies did not find evidence that early and late bilinguals have different control mechanisms, and instead posited that the demand of the bilingual situation, and the use of the two languages, is responsible for whether early and late bilinguals have similar cognitive control abilities. This is especially supported by the two latter studies with intermediate late bilingual vs. highly proficient late bilinguals, and elderly long-term immersed Dutch-English bilinguals. Their findings imply that not age of onset is the source of a difference in performance on components of executive control, but how these languages are used in specific contexts is more important.
4.2
Neuroimaging studiesIn order to find out how age of onset might have an influence on language representation in the brain, a number of studies have employed neurophysiological measures to elucidate a possible difference in brain structure. Abutalebi & Green (2007), among others, have posited continued neuroplasticity for language learning in the adult brain, and that both behavioural and neural changes can result from L2 learning later in adulthood. Research has shown that late bilingualism indeed generates neuroanatomical changes, though it leads to a slightly different neural substrate than lifelong or early bilingualism.
Mechelli et al. (2004), for example, showed that L2 learning increases GM density in the left IPL, but both level of proficiency and age of onset modulate the degree of structural reorganization of this brain area. They used voxel-based morphometry (VBM) to investigate neuroplasticity in English and Italian bilinguals. In the first experiment, they investigated the differences in density of GM and WM between monolinguals and bilinguals. Three groups were compared: English monolinguals (who had no or little previous exposure to another language), English early bilinguals (who learned a European L2 before the age of 5), and English late bilinguals (who learned a European L2 between the ages of 10 and 15 years). The late bilinguals were required to have been practicing their L2 regularly for at least 5 years at the time of testing. In the second experiment, they investigated the relationship between proficiency and brain structure, and between age of onset and brain structure. A total of 22 Italian speakers were investigated, all of whom had learned English as their L2 between the ages of 2 and 34 years old. It was found that L2 proficiency positively correlated with GM density in the left IPL, while age of onset negatively correlated with GM density in this same area. In other words, the more proficient the speaker is in the L2, the higher GM density is in the IPL, but, this GM density decreases as age of onset increases. However, ages of the participants and daily exposure to both languages remain unknown. Though it is interesting to hypothesize that structural plasticity is correlated with L2 performance and age of onset, one cannot conclude that level of proficiency and time of onset only are important, if all factored are not controlled for.
Klein, Mok, Chen, & Watkins (2014) provided more convincing evidence that age of onset is important for structural reorganization in the brain: they found that cortical thickness of the left and right IFG differ in early and late bilinguals. They tested for cortical thickness in four groups: monolinguals (mean age = 25, and no or little exposure to an L2), simultaneous bilinguals (mean age = 23, and acquired their L2 between 0 and 3 years of age), early sequential bilinguals (mean age = 26, and acquired their L2 between 4 and 7 years old), and late sequential bilinguals (mean age = 28, and acquired their L2 between 8 and 13 years old). Though all subjects
used both languages on a daily basis, they differed in proficiency and the degree to which they used their languages. Magnetic Resonance Imaging (MRI) scans were obtained to test for cortical thickness differences between participant groups. It was found that the pars triangularis and pars orbitalis of the left IFG were thicker in early and late sequential bilinguals compared to the monolinguals, whereas in these regions in the right IFG the opposite was found: the pars orbitalis was thicker in the monolinguals and simultaneous bilinguals in comparison to the early sequential bilinguals and late sequential bilinguals. In subsequent testing for cortical thickness in relation to age of onset in the bilingual groups, they controlled for age of onset, years of L2 exposure, and L2 proficiency. They found that cortical thickness and age of onset were positively correlated in the left IFG, whereas age of onset was negatively correlated with cortical thickness of the right IFG. In other words, the later in life the L2 is learned, the thinner the cortex of the right IFG is, but the thicker the cortex of the left IFG.
These results on the relationship between cortical thickness of the left and right IFG and age of onset are consistent with the results of Hull & Vaid (2007) on different functional lateralization patterns in different bilingual groups. They investigated early bilinguals (who had acquired both languages by age six), and late bilinguals (who acquired their L2 after six years of age). It was found that the early and late bilinguals had different patterns of functional lateralization for both languages: whereas both hemispheres were involved in early bilinguals, late bilinguals showed dominance of the left hemisphere. Moreover, a correlation between proficiency and functional lateralization was found: the less proficient a late bilingual was in the L2, the more left lateralization s/he showed. However, this is not consistent with the study of Berken, Chai, Chen, Gracco, & Klein (2016), who found greater lateralization of the left IFG in late highly proficient bilinguals in comparison to early highly proficient bilinguals.
By means of seed-based resting-state fMRI (rs-FMRI), Berken et al. (2016) measured functional connectivity between the left IFG and the right IFG in simultaneous French-English bilinguals (mean age = 23.1, and acquired two languages from birth) and English-French or French-English sequential bilinguals (mean age = 25.7, and learned an L2 after 5 years of age). rs-fMRI has been shown to effectively examine the influence of language experience on functional organization of the brain by deriving functional connectivity between brain areas. This is done through identification of low-frequency BOLD signal fluctuations in the brain that occur when the subject does not show task-driven behaviour (e.g. Biswal, Van Kylen, & Hyde, 1997; Biswal, Yetkin, & Haughton, & Hyde, 1995). All bilinguals were highly proficient, and the simultaneous and sequential bilinguals were comparable in daily use of their languages. It was found that simultaneous bilingual had greater functional connectivity between the left IFG and right IFG,
and the degree of connectivity negatively correlated with age of onset in sequential bilinguals. In sequential bilinguals, they found greater lateralization of the left IFG. Additionally, it was found that connectivity between the left IFG and right IFG was associated with reduced neural activation in the left IFG during speech production. These combined findings indicate that early bilinguals have increased neural efficiency in the left IFG, which is involved in cognitive control and language production. Though it is claimed that the neural substrate of language learning might still change in late L2 learners, it was hypothesized that optimal functional connectivity is only achieved by simultaneous bilingual acquisition. Berken et al. concluded that their findings might reflect an enhanced ability of early bilinguals to handle two simultaneously activated languages in comparison to late bilinguals, which would explain a possible early bilingual advantage in language processing.
Following the hypothesis that functional connectivity between the left and right IFG is greater in simultaneous bilinguals, and that age of onset negatively correlates with the degree of connectivity, it may be expected that early and late bilinguals differ in terms of performance on components of cognitive control these areas are responsible for. Previous studies have argued that the right IFG is primarily responsible for attentional control and response inhibition (Dove, Pollmann, Schubert, Wiggins, & von Cramon, 2000; Hampshire and Owen, 2006; Hampshire, Chamberlain, Monti, Duncan, & Owen, 2010). The right IFG is then hypothesized to play a role in slowing down, pausing, or completely suppressing an unwanted response. Indeed, other research on lesions in the right IFG has shown that damage to the pars triangularis region in the right IFG results in disrupted response selection (Aron, Fletcher, Bullmore, Sahakian, & Robbins, 2003). As for the left IFG, research has not implicated this area as central to a specific component of executive functioning control, but this region is thought to play a role in both linguistic and nonlinguistic executive control, and language processing (Coderre, Smith, van Heuven, & Horwitz, 2015). This suggests that the left IFG is important for interdependence between language control and cognitive control, and it was subsequently claimed by Berken et al. (2016) that greater functional connectivity between the left and right IFG allows for a more efficient handling of two active languages. These findings on different connectivity between certain brain areas in early and late bilinguals show that structural brain plasticity is likely influenced by the time of L2 learning.
Other neuroanatomical research has also argued for different control mechanisms in early and late bilinguals. For example, Martin, Strijkers, Santesteban, Escera, Hartsuiker, & Costa (2013) investigated whether early bilinguals control a new language acquired later in life the same way as late bilinguals do. By means of comparing ERPs when switching between a dominant L1
and a late acquired L3 in a picture-naming language-switching task, their aim was to see how different demands were placed on early and late bilinguals’ language control mechanisms. They looked at two ERP components: the anterior N2 component, which in previous research has been observed as being important for response selection processes and cognitive control (Folstein & Van Petten, 2008), and the Late Positive Component (LPC), which seems to reflect the process that links the input of a picture to the correct lexical item in the target language (e.g. Liotti, Woldorff, Perez, & Mayberg, 2000). Two groups were compared: Spanish-Catalan early bilinguals who learned English as their late L3 (mean age = 21), and Spanish-English late bilinguals who learned Catalan as their late L3 (mean age = 24). Both groups of participants were comparable in level of proficiency in L1, and in L3 (4.0 and 2.2, respectively, on a four point scale questionnaire on self-based proficiency); however, the late bilinguals were not highly proficient in their L2 whereas the early bilinguals were. The early bilinguals had an age of onset of 2 years for their L2, and 10 years for their L3. The late bilinguals had an age of onset of 10 years for their L2, and 22 years for their L3. It was found that the anterior N2 component was larger for late bilinguals compared to early bilinguals, indicating that the L3 is not controlled in the same way in the early and late bilingual groups. The LPC component was not modulated, implying that early and late bilinguals do not differ in terms of stimulus-response mapping. In other words, age of onset arguably affects how bilinguals control their language output, despite the similarity of proficiency in the L1 and L3.
On the basis of different neural substrates of early and late bilinguals, it has been claimed that the continuous use of two languages in early bilinguals and thus the recurrent need to avoid intrusions from the unwanted language might cause a specific development of brain structure. Martin et al. used their results (the different amplitudes of the N2 component between early and late bilinguals) as support for the hypothesis that early bilinguals benefit from the control mechanism they have developed from birth, which is used to to learn and control a new late language. However, if a different development of control mechanisms in the brain stems from an extensive and long-lasting need to control both languages and select the appropriate language, it is entirely possible that the late bilinguals in their study will eventually develop the same cognitive control mechanism as early bilinguals. Seeing that the late bilinguals had been exposed to their L3 for approximately two years at the time of testing, and they still had comparable proficiency with the early bilinguals (who had 11 years of previous exposure to their L3), it is possible that different amplitudes of the anterior N2 component are the result of different contexts in which the L3 in both bilingual groups were used. These finding thus raise more questions as to whether
age of onset is an important factor for cognitive control development in bilinguals, or whether the context of learning and use influences cognitive control to a greater extent.
In sum, whereas behavioural data seems to imply that early and late bilinguals do not develop different executive control mechanisms due to different ages of onset per se, neuroimaging studies have more evidence that the neural substrate of language representation is influenced by age of onset. Different neurophysiological measures indicate that i.e. GM density, functional connectivity, and cortical thickness differ between early and late bilinguals, where early bilinguals seem to reap a more optimal cognitive control mechanism. However, not all neuroimaging studies have controlled for potentially confounding factors, complicating the hypothesis of some studies that age of onset is primarily responsible for a different neuroanatomical organization.
Now, let us turn to the final chapter where the results from both behavioural studies and neuroimaging studies are discussed.
