The neuroanatomical consequences and pathological implications of bilingualism

Abstract

In recent years, there has been a rise in the number of people who are able to speak two or more languages. This has been paralleled by an increase in research related to bilingualism. Despite this, much of the neuroanatomical consequences and pathological implications of bilingualism are still subject to discussion. This review aims to evaluate the neuroanatomical structures related to language and to the acquisition of a second language as well as exploring how learning a second language can alter one's susceptibility to and the progression of certain cerebral pathologies. A literature search was conducted on the Medline, Embase, and Web of Science databases. A total of 137 articles regarding the neuroanatomical or pathological implications of bilingualism were included for review. Following analysis of the included papers, this review finds that bilingualism induces significant gray and white matter cerebral changes, particularly in the frontal lobes, anterior cingulate cortex, left inferior parietal lobule and subcortical areas, and that native language and acquired language largely recruit the same neuroanatomical structures with however, subtle functional and anatomical differences dependent on proficiency and age of language acquisition. There is adequate evidence to suggest that bilingualism offsets the symptoms and diagnosis of dementia, and that it is protective against both pathological and age‐related cognitive decline. While many of the neuroanatomical changes are known, more remains to be elucidated and the relationship between bilingualism and other neurological pathologies remains unclear.

Bilingualism induces significant grey and white matter cerebral changes, particularly in the frontal lobes, anterior cingulate cortex, left inferior parietal lobule and subcortical areas. An individual's native and acquired language largely recruit the same neuroanatomical structures however they display subtle functional and anatomical differences dependent on proficiency and age of acquisition. Bilingualism also offsets the symptoms and diagnosis of dementia and is protective against pathological and age‐related cognitive decline.

1. INTRODUCTION

In recent years, there has been a rise in the number of people who are able to speak two or more languages (Ansaldo et al., 2008; Bialystok et al., 2012). It is now estimated that over half of the world’s population are bilingual (Ansaldo et al., 2008; García‐Pentón et al., 2016; Wong et al., 2016). This rise in bilingualism has been paralleled by an increase in research into the neuroanatomical, cognitive, and pathological implications associated with the acquisition of a second language.

One of the contributing factors to this increased research effort is growing evidence that bilingualism elicits a cognitive advantage (Abutalebi et al., 2012; Petitto et al., 2012, Van Heuven and Coderre, 2015). Until the start of the 20th century, it was suggested that learning a second language at a young age had a detrimental effect on a child’s intelligence (Al‐Amri, 2013). This stance changed following the pivotal study by Peal and Lambert (1962) where English and French bilinguals showed superior performance on measures of both verbal and non‐verbal intelligence. These advantages of bilingualism have been attributed to increased cognitive and brain reserve as well as to structural changes associated with learning a second language (Abutalebi et al., 2012; Gold et al., 2013; Grant et al., 2004). Despite this change in understanding and the improved neuroimaging studies available throughout the 20^th and 21^st centuries, much of the neuroanatomical consequences and pathological implications of learning a second language are still subject to discussion.

Currently, it is largely agreed through structural and functional imaging studies that learning a second language elicits several gray and white matter neuroanatomical changes, and that the acquisition and use of a second language results in experience‐dependent structural plasticity and macro‐anatomical changes independent to those of monolinguals (Burgaleta et al., 2016). Traditionally, functional magnetic resonance imaging (fMRI) studies have shown that bilinguals have greater cortical area activation in areas related to language and cognitive control (Abutalebi, 2008; Indefrey, 2006).

It is also suggested within current literature that there are several factors which may affect the degree of structural and cytoarchitectural changes observed in bilinguals. Age of acquisition and level of proficiency have been suggested to influence these differing cortical representations, as has language similarity and educational level (Green, 2003; Wartenburger et al., 2003). Equally, some evidence exists to suggest that the process of learning a second language may prevent and ameliorate the impacts of pathological and age‐related cognitive decline (Smirnov et al., 2019; Weigmann, 2014).

However, where the evidence is lacking and where there are still several outstanding questions with regards to how each of these evidence bases is related. It is known that bilingualism elicits several neuroanatomical changes, but how these relate to those of first language (L1) acquisition and how second‐language (L2) representation differs to that of L1 is still subject to much controversy and confusion. Similarly, the factors involved in the development of these differing cortical representations would benefit from further clarity and decisive statements following thorough evaluation of the literature.

Finally, the clinical data available regarding the impacts of bilingualism on cognitive decline must now be assessed alongside the data regarding these neuroanatomical changes so that correlations and inferences can be made as to why an activity such as learning an additional language may act as a preventative measure against Alzheimer's and alike pathologies.

The recent rise in popularity of bilingual‐related research alongside new and enhanced imaging modalities means that there is now a literature base of sufficiently high quality to warrant a review which may address these questions. Such a review holds the potential to provide a first attempt at correlating traditional anatomy‐based research with real‐world clinical implications and may provide this historically heterogenous research area with the clarity required to enable progression into bilingual therapeutic interventional studies.

This review therefore aims to summarize the available literature on the neuroanatomical structures related to learning a second language and to determine how these areas contrast to those associated with first‐language acquisition. This review also aims to evaluate what factors are involved in the development of these representations.

Additionally, this review will provide a detailed synopsis regarding the structural and cytoarchitectural changes elicited by second‐language acquisition and to establish whether these changes are correlated with age of acquisition and/or level of proficiency. Finally, this review is designed to ascertain how the neuroanatomical consequences of bilingualism may offset the symptoms and diagnosis of dementia and to explore whether they are protective against both pathological and age‐related cognitive decline.

The questions that this review is designed to address are as follows:

1. What are the variations in the neuroanatomical representations between first and second languages and what factors are involved in the development of these representations?

2. What structural and cytoarchitectural changes are associated with second‐language acquisition and how are these correlated with age of acquisition and level of proficiency?

3. How do the neuroanatomical changes associated with bilingualism protect against both pathological and age‐related cognitive decline?

This review hypothesizes that the neuroanatomical areas associated with a second language will closely represent those of the first language and that the degree of similarity will be positively and negatively correlated with level of proficiency and age of acquisition respectively. It is also suggested that the process of learning a second language may result in subtle gray and white matter within the areas of the perisylvian language zone which may subsequently result in greater mental flexibility and connectivity which may be protective against pathological and age‐related cognitive decline.

2. METHOD

This research was undertaken in the form of a semi‐systematic meta‐narrative review according to the Hannah Snyder and RAMESES guidelines (Snyder, 2019; Wong et al., 2013). This methodological approach was adopted due to the anticipated heterogenicity within study designs, outcomes, and imaging modalities. The pluralistic approach also lends itself to historical research areas subject to conflicting findings and methodologies as well as to areas of research traditionally differently conceptualized and studied at different times by different groups (Wong et al., 2013).

A literature search was conducted on the Medline, Embase, and Web of Science databases for articles published until September 2020. The literature terms used for bilingualism were “bilingu*,” “trilingu*,” “plurilingu*,” “multilingu*,” “polylingu*,” “polyglot,” “diglot,” and “multilanguage.” These terms were combined with neuroanatomical and pathological terms as outlined by the full electronic search strategy used on the Ovid MEDLINE database in the Material S1.

Bibliographic databases were searched with the independently validated pre‐specified search strategy which was reviewed by two reviewers and performed in blind duplicate. To ensure full coverage of the literature, spinal journals at Southampton University were hand‐searched and the gray literature source Opengrey was also searched. No date nor place of publication or journal restrictions was applied.

Exclusion of studies was performed in blind duplicate by authors CT and SH. Disagreement was resolved via discussion and consultation with third author SB. The online software Rayyan (Ouzzani et al., 2016) was used for title and abstract and full‐text screening.

All human neuroimaging studies regarding the neuroanatomical effects of second‐ or third‐language acquisition were included. Studies were not excluded based on study population. Therefore, no age, gender, or other demographic exclusions were made. To ensure a comprehensive review covering several forms of bilingualism, no language‐based restrictions were made within the search, and all forms of bilingualism were considered, including bimodal bilinguals. However, only articles written in English were included.

A range of study outcomes and imaging modalities were considered. These included, but were not limited to, cortical stimulation, functional magnetic resonance imaging, structural and event‐related magnetic resonance imaging, diffusion tensor imaging, and cortical thickness cross‐sectional analysis.

This review will therefore measure whether the binary outcome of learning an additional language elicits significant neuroanatomical changes relative to those of a first language and whether these are related to level of proficiency and age of acquisition. This review will subsequently test whether these changes affect the diagnosis and progression of age‐related and pathological cognitive decline. To enable these aims to be met, all types of study design were considered and full‐text review deemed an article relevant for inclusion if it provided information on bilingualism relating either to neuroanatomy or pathological implications.

Studies were excluded during title and abstract and full‐text screening if their sample size was less than 5, with the exception of twin studies, or if the outcomes of the study could not solely be attributed to the bilingual nature of the study population. Studies were also excluded if they did not provide original outcome analysis relating to the neuroanatomical structural and cytoarchitectural changes elicited by second‐language acquisition or if they did not provide primary analysis regarding the effect of bilingualism on cognitive decline and pathology. Studies were further excluded due to low quality, which was informally assessed in duplicate by the lead authors using the criteria of the ROBINS‐I and Cochrane risk of bias tools as guidelines. However, due to not being a systematic review, no empirical assessment of bias was performed and this should be noted as a potential limitation when interpreting the discussion.

The focus points for this review were determined by scoping searches of the available literature. Discussion amongst the authors led to subsequent identification of the areas of research that were thought to benefit most from inclusion in this review. These areas were broadly categorized as: the neuroanatomy associated with language, the neuroanatomy of bilingualism, the impact of proficiency and age of acquisition and pathologies associated with bilingualism.

A primary limitation of this method lies within its principle of pragmatism in which the lack of a truly objective bias/quality assessment gives rise to the potential for publication bias and does not account for study quality or risk of bias. Furthermore, a degree of subjectivity is introduced when determining the quality of studies for inclusion. However, by performing study screening in duplicate and reporting only a narrative synthesis, the implications of these limitations are minimized.

3. RESULTS

Primary search generated 771 papers from the Ovid MEDLINE database, 1694 from EMBASE and 1235 from Web of Science. Following duplicate removal 995 papers were identified. These papers first underwent screening by title and abstract and were then screened in full text. Seven hundred and fifty‐nine were excluded following title and abstract screening, leaving 236 papers for full text screening, 107 of which were excluded. Following the addition of 8 papers from forwards and backwards citation searching, 137 papers were included in the final review (Figure 1).

FIGURE 1.

PRISMA flow diagram [Colour figure can be viewed at wileyonlinelibrary.com]

4. DISCUSSION

4.1. Neuroanatomy associated with language

4.1.1. Introduction

Language is commonly divided by linguists into the following components: phonology, morphology, syntax, semantics, and pragmatics (McNally & Kennedy, 2008). It is conventionally accepted that the cerebral structures involved in these language abilities reside around the Sylvian fissure of the language dominant hemisphere, as taught by the Broca‐Wernicke‐Lichtheim‐Geschwind model (Nasios et al., 2019) or so‐called “House of Speech.” However, the more recent Hickok and Poeppel model of dual‐stream processing describes two functional pathways both stemming from the Sylvian Parietal‐temporal (SPt) area in the dominant posterior planum temporale (Hickok et al., 2003, 2009). These two pathways are termed the ventral and dorsal streams (Hickok et al., 2003, 2009). This reflects a transition from the Broca‐Wernicke anatomical model to a more functional understanding of and approach to language processing (Hickok & Poeppel, 2000, 2004).

In addition, in recent years there has been increased attention devoted to areas of the brain beyond the perisylvian language zone (Hickok, 2009; Sierpowska et al., 2018) (Figure 2). For example, the superior temporal gyrus (STG) is thought to be involved in the construction of phrases and intelligibility. The middle and inferior temporal gyri are also thought to contribute to word‐level processing with the basal ganglia and cerebellum contributing to linguistic computation (Poeppel & Hickok, 2004). New imaging modalities and recent research have improved our understanding of the neuroanatomical representations of language. However, remains unclear. This review therefore attempts to provide the first comprehensive evaluation of the neuroanatomical structures and changes related to learning a second language and to determine what factors are involved in the development of these representations and how these changes may influence pathological and age‐related cognitive decline.

FIGURE 2.

The roles of neuroanatomical language areas beyond the perisylvian language zone. ITG, inferior temporal gyrus; MTG, middle temporal gyrus; STG, superior temporal gyrus (Servier, 2021) [Colour figure can be viewed at wileyonlinelibrary.com]

4.1.2. Language recognition/processing

The primary auditory cortex (A1) is located in the STG, directly under the lateral fissure and bounded anteriorly by Heschl's gyrus and posteriorly by the most posterior aspect of the Sylvian fissure (Hickok & Poeppel, 2007; Scott et al., 2000). It is responsible for conscious sound recognition and is the first higher level cortical processing component of language recognition (Davies & Sugano, 2018; Webster & Fay, 2013). The primary auditory cortex is organized such that low frequencies are represented laterally while high frequencies are represented medially around Heschl's gyrus (Langers & van Dijk, 2012; Webster & Fay, 2013). However, a detailed tonotopic map is yet to be fully established (Langers & van Dijk, 2012). During the recognition and perception of speech, a core role of the auditory cortex is the analysis of complex acoustic patterns which consequently allow the detection of words that formulate a linguistic message (Brodbeck et al., 2018a).

Diffusion tensor imaging (DTI), event‐related fMRI and studies alongside event‐related potentials electroencephalogram (ERG) based studies have all illustrated that phonology, the first component of language, can be considered as the recognition of an auditory input as speech rather than a non‐descript sound (Carr, 2019; Lass, 1984; Yallop & Fletcher, 2007). This occurs in an asymmetrical bilateral fashion in the STG (Hickok, 2009; Rodriguez‐Fornells et al., 2009), and potentially can be localized even further to the posterior superior temporal sulcus (pSTS) (Hickok et al., 2009). It is termed asymmetrical as evidence suggests hemispheric differences in the processing of acoustic/speech information primarily regarding selectivity for temporal (left hemisphere) versus spectral (right hemisphere) resolution (Hickok, 2009; Zatorre et al., 2002). The two hemispheres also differ in their sampling rates: the left hemisphere functions at 17‐46Hz greater than the right (Hickok, 2009; Poeppel, 2003). Faster sampling rates allow detection of segmental level phonemes (single sounds), whereas slower sampling in the non‐dominant hemisphere allows for detection of prosody and intonation which are supra‐segmental functions of speech (Hickok, 2009; Poeppel, 2003). There is also a hierarchical representation of speech length; simple phonemes are found posteriorly in the STG progressing through to syllable and phrases more anteriorly (DeWitt & Rauschecker, 2012). The bilateral and asymmetrical representation associated with the ventral pathway of speech recognition enables parallel processing pathways (Hickok, 2009; Hickok & Poeppel, 2000, 2004), which is evidenced by the absence of significant speech recognition deficits on unilateral temporal lobe damage (Hickok, 2009; Hickok & Poeppel, 2004, 2007). Recent research has identified neural signals from the successful recognition of individual words in continuous speech (Brodbeck et al., 2018b; Broderick et al., 2018; Ding et al., 2016); however, no clear neurophysiological signals have been identified when demonstrating the transition of acoustic and/or phonetic representations to symbolic lexical representations (Brodbeck et al., 2018a).

Following speech recognition, auditory input is processed at the temporal parietal junction (TPJ) (Hickok & Poeppel, 2000, 2007; Jones et al., 2014). Language processing is then subdivided between information that concerns comprehension and that concerned with the motor system for an articulatory response (Hickok & Poeppel, 2000, 2004; Poeppel & Hickok, 2004). This requires both the ventral route for comprehension (temporal lobes bilaterally) and the dorsal route for sensory‐motor integration (left/dominant frontal lobe) (Hickok, 2009; Hickok & Poeppel, 2000, 2004, 2007). The TPJ is also responsible for attributing meaning to non‐auditory language inputs, such as sign language (visual) or braille (somatosensory) and will receive its input from the primary visual cortex rather than the ventral auditory stream in such instances (Hickok, 2009; Hickok & Poeppel, 2007; Jones et al., 2014).

Meanwhile, the lateral and posterior regions of the temporal lobe (middle and inferior temporal gyri) likely house the conceptual access mechanisms for speech and are thus responsible for the transformation of an acoustic signal into a conceptual representation and the following translation of said conceptual representation into any number of spoken utterances (Hickok, 2009; Jackendoff, 1988; Sierpowska et al., 2018).

Although the neuroanatomical pathways involved in processing phonological information are broadly understood, those relating to more complex aspects of language such as grammatical comprehension and syntax construction remain unclear (Caplan et al., 2000; Dodel et al., 2005; Green, 2003; Grodzinsky & Friederici, 2006; Novick et al., 2005). So far, current evidence indicates involvement of several cognitive networks located in the anterior temporal lobe, Broca's area and fronto‐basal ganglia circuits; however these are beyond the scope of this review (Caplan et al., 2000; Novick et al., 2005).

4.1.3. Language comprehension

Once language has been received, the posterior lateral and inferior temporal cortical areas (the middle and inferior temporal gyri) play an important role in the coupling of meaning to sound (Hickok, 2009; Hickok et al., 2000; Hickok & Poeppel, 2004, 2007). This is supported by the fact that damage to the left posterior temporal lobe is often associated with deficits in auditory comprehension (Bates et al., 2003). Cortical stimulation (Miglioretti & Boatman, 2003) and fMRI (Binder et al., 1997; Rissman et al., 2003) have shown involvement of the middle temporal gyrus in auditory comprehension, and the superior temporal, inferior frontal, and posterior middle temporal gyri in lexical‐semantic processing. The anterior temporal lobe may also be involved in sentence and lexical‐semantic processing although the evidence is less conclusive (Hickok, 2009).

A common approach to understanding the neuroanatomical representation of language and its comprehension is via conceptual semantics; a framework for semantic analysis (Jackendoff, 1988; Kuznetsov, 2015). The primary goal of conceptual semantics is to describe how humans express their understanding of the world by means of linguistic utterances (Jackendoff, 1988). Conceptual‐semantics can be viewed through three main concepts (Caramazza & Mahon, 2003; Jackendoff, 1988). Most widely supported is the notion that conceptual information has a diffuse representation throughout the cortex involving both sensory and motor cortical regions that originally represent that information (Martin & Chao, 2001). Secondly, some believe that conceptual information has a more focally organized semantic loci located in the anterior temporal region (Caramazza & Mahon, 2003). Finally and subject to the least support is the idea that conceptual information and semantic knowledge are held in functionally specialized systems that are devoted to the processing of information from evolutionary relevant conceptual categories (Caramazza & Mahon, 2003). While it is evident that the posterior, lateral, and inferior temporal cortical areas are fundamental to language comprehension and semantics, much is still to be understood about the way auditory inputs are comprehended and are then ascribed a meaning.

4.1.4. Language output

When articulating a single word, the production of language involves two computational stages as a minimum (Dell et al., 1995, 1997). The first stage is selection of the appropriate lexical term (Dell et al., 1997) and the second stage is interfacing with the phonological form (Dell et al., 1995).

The role of the left‐hemispheric posterior sensory‐related cortex in speech production is evidenced by speech production deficits caused by damage to areas such as the left dorsal posterior superior temporal gyrus (pSTG) (Hickok & Poeppel, 2007). fMRI studies in patients with conduction aphasia provide further evidence for involvement of the left pSTG as an auditory cerebral region in speech phonology (Hickok et al., 2000; Indefrey & Levelt, 2004). These studies demonstrate the role of the left posterior superior temporal regions in phonological speech production.

Most of the data discussed so far have been with respect to phonemics and single‐word processing. By contrast, the functional representation of syntax comprehension is less well understood. Broca's area has long been considered a cornerstone of syntactic processing, production, and comprehension (Elmer, 2018; Hickok, 2009). This is supported by Broca's aphasics, individuals who struggle to understand complex syntax (Hickok & Avrutin, 1996). Recent evidence has suggested that the anterior temporal lobe bilaterally plays a role in syntactic processing as it has a syntax preferential response: damage to this area produces difficulties in understanding complex or syntactically ambiguous sentences (Humphries et al., 2006; Zaidel et al., 1995). Despite the relative paucity of data surrounding more complex syntax recognition and comprehension, it can be concluded that recognition of speech is primarily conducted in the superior temporal lobe circuits bilaterally and conversely that speech production is carried out in the fronto‐parietal and temporal circuits of the dominant hemisphere. Table 1 summarises the main neuroanatomical areas mentioned in this review along with their linguistic roles.

TABLE 1.

Primary neuroanatomical areas involved in language

Neuroanatomical location	Role in language
Primary auditory cortex	Conscious sound recognition Higher level cortical processing for language recognition Low frequencies are represented laterally around Heschl's gyrus. High frequencies are represented medially around Heschl's gyrus
Superior temporal gyrus and the posterior superior temporal sulcus	Asymmetrical bilateral recognition of auditory input as speech Simple phonemes are represented posteriorly in the STG More complex syllable and phrases are represented more anteriorly in the STG
Temporal parietal junction	Auditory, visual (sign language) and somatosensory (brail) input processing and meaning attribution
Middle and inferior temporal gyri	Coupling of meaning to sound House the conceptual access mechanisms for speech Transform acoustic signals into conceptual representations Translation of the conceptual representations into spoken utterances
Middle temporal gyrus	Auditory comprehension processing
Superior temporal, inferior frontal, and posterior middle temporal gyri	Lexical‐semantic processing
Anterior temporal lobe	Sentence and lexical‐semantic processing
Temporal lobes bilaterally	Comprehension language processing
Left/language dominant frontal lobe	Sensory‐motor integration
Anterior temporal lobe, Broca's area and fronto‐basal ganglia circuits	Grammatical comprehension and syntax construction
Left‐hemispheric posterior sensory‐related cortex	Role in phonological speech production
Broca's area	Syntactic processing, production and comprehension

Abbreviations: Brodbeck et al., 2018a; Hickok & Poeppel, 2000, 2004; Langers & van Dijk, 2012; Poeppel & Hickok, 2004; STG, superior temporal gyrus (Dell et al., 1995, 1997; Webster & Fay, 2013.

4.2. Neuroanatomy and bilingualism

4.2.1. Introduction and theoretical concepts

One of the major questions concerning the neuroanatomy of bilingualism is whether L2 occupies the same anatomical location and utilizes the same cognitive pathways as the native language L1, or if L2 and L1 are anatomically distinct. Indeed, theoretical models of bilingualism often assumed to shared semantics across languages, such as the revised hierarchical model (Kroll & Stewart, 1994) and the BIA+ (bilingual interactive activation) model (Dijkstra & Van Heuven, 2002a, 2002b). However, this does not imply that the semantic representation of every concept should completely overlap across languages. Other models, like the distributed feature model (Van Hell & De Groot, 1998a, 1998b) or the model of Duyck and Brysbaert (2004), assume partially overlapping semantic representations between translation equivalents across languages depending on specific characteristics of the concepts.

Much of the evidence suggesting anatomically distinct locations stems from the Critical Period Hypothesis (CPH) proposed by Penfield and Roberts in 1959 and from the declarative/procedural model proposed by Ullman (2001a, 2001b, 2004). The CPH states that in the first few years of life the ability to learn a language readily develops if adequate stimuli are present; thereafter acquisition of a language becomes more challenging and consequently the individual will not achieve comparable proficiency (Claussenius‐Kalman et al., 2019; Grimshaw et al., 1998; Hernandez & Li, 2007; Johnson & Newport, 1989). It follows that if L2 is acquired outside this critical period then L1 and L2 are learnt, processed, and interfaced differently and they may therefore have different cognitive/cerebral representations (Abutalebi, 2008). The classification of early vs late bilinguals can vary in current literature; however, it is generally considered that early/simultaneous bilinguals are exposed to a second language before the age of seven while those exposed to a second language after the age seven are classified as late/sequential bilinguals (Sabourin et al., 2014). Others propose a much younger threshold of 3 years for early bilinguals (Kovelman et al., 2008).

The declarative/procedural model states that cognitive procedures regarding the lexicon of vocabulary rely on declarative memory associated with temporal lobe circuits (Ullman, 2001a, 2001b, 2004), whereas learning and applying grammatical rules, as well as the morpho‐syntactical elements of language, are served by the procedural memory system involving frontal/basal‐ganglia circuits at the nonconscious level (Abutalebi, 2008; Caplan, 2006; Grodzinsky & Friederici, 2006; Ullman, 2001a, 2001b, 2004). The declarative/procedural model states that by their nature, lexicon and grammatical rules are distinct and are therefore processed and accessed differently and reside in distinct cerebral structures. Just as these aspects of language are suggested to be acquired and therefore represented differently, the same notion can be applied to the acquisition of L2. L1 is learnt largely at the subconscious implicit innate level, whereas L2 is acquired to a greater degree via a conscious repetitive learning process via declarative explicit memory (Ullman, 2001a, 2004). Therefore, it can be hypothesized that L1 and L2 would have different anatomical localities based on their method of acquisition (Abutalebi, 2008).

However, the CPH hypothesis and the declarative/procedural model were challenged by Singleton and Lengyl who argued that there is no critical learning period for vocabulary as it is always learnt at a conscious declarative level (Singleton, 2005; Singleton & Lengyel, 1995). Furthermore, the CPH hypothesis as applied to L2 acquisition is subject to debate as it is now generally agreed that there is no critical time‐period in which L2 must be learnt (Rothman, 2008). The evidence supporting separate cerebral loci for both L1 and L2 is therefore largely theoretical and has been replaced by more recent functional neuroanatomical evidence.

4.2.2. L1 and L2 functional localization

Following a multi‐voxel pattern analysis study the semantic representations of both L1 and L2 are considered to partly overlap in all modalities of language in highly proficient individuals (Van de Putte et al., 2018). For word reading and word listening tasks, similarity has been observed in the left middle occipital gyrus extending into the ipsilateral inferior occipital gyrus, the banks of the right calcarine fissure, the left Inferior frontal gyrus (IFG), the left superior frontal gyrus and the right precuneus (Van de Putte et al., 2018). Furthermore, across‐language overlap in picture naming tasks involves the middle occipital and fusiform regions bilaterally and extends into the inferior temporal regions (Van de Putte et al., 2018).

From this study and others (Dodel et al., 2005, Indefrey et al., 2005), it would appear that syntax processing areas of L1 (left precentral gyrus, left basal ganglia, left IFG and supplementary motor area) are closely related to those of L2. An additional fMRI syntax judgment study found that brain activity for L1 and L2 are overlapped and was independent of L2 proficiency (Luke et al., 2002). Another fMRI study investigating acquisition of language rules in an artificial language concluded that the process of learning L2 is done via the existing language network of L1 (Opitz & Friederici, 2004).

The concept that L1 and L2 share cerebral loci, namely the left IFG, is further supported by an fMRI adolescent twin study, showing that L2 followed L1 specialization pathways (Sakai et al., 2004). However, it was also found that these areas were functionally more active in L2 sentence production, specifically when making/identifying past tense forms (Dodel et al., 2005; Indefrey et al., 2005). It was concluded that activation levels in the left dorsal IFG represents the acquisition of past tense knowledge and that cortical plasticity for L2 acquisition is guided toward the L1 specialization pathways of the left dorsal IFG (Sakai et al., 2004). This suggests that despite anatomical similarities, the areas of the brain concerned with syntax construction may be functionally interlinked more so during L2 sentence production than in L1.

Other studies demonstrate differences in functional activity between L1 and L2. An evaluation of several meta‐analyses concluded that L2 is represented by similar cortical regions as L1 with additional brain regions usually involved in cognition recruited when performing language tasks (Cargnelutti et al., 2019). This suggests that in bilinguals an increased cognitive load is required to maintain the two languages and to continually inhibit the language that is not in current use. It was also observed that late bilinguals consistently recruit wider cortical areas than early bilinguals typically around the left IFG at the border between Brodmann area 45 and the dorsolateral prefrontal cortex (DLPFC) (Cargnelutti et al., 2019).

This finding of anatomical dissimilarities between L1 and L2 is supported by a study of patients undergoing awake dominant‐hemisphere craniotomies: 95% of subjects displayed some degree of distinction between L1 and L2 specific locations (Lucas et al., 2004). These L1 specific sites were found to be focally discrete and mostly located in the inferior frontal and posterior inferior parietal areas, whereas L2 sites were found to be more widely distributed across the exposed cortical surface of the temporal lobe (Lucas et al., 2004). This study confirmed the presence of language‐specific cortical centers in bilinguals and provides strong evidence supporting the notion that both L1 and L2 differ in their anatomical distribution. However, other studies have found no overall difference in the cortical extent for L1 and L2, suggesting that L2 may not always require greater cortical representation (Lucas et al., 2004). Table 2 provides a summary of the neuroanatomical language representation of L2 relative to L1.

TABLE 2.

Summary table of the neuroanatomical language representation of L2 relative to L1.

Neuroanatomical location	L1 and L2 relationship
Left middle occipital gyrus	L1 and L2 similarity in word reading and listening tasks
Ipsilateral inferior occipital gyrus
Banks of the right calcarine fissure
Left Inferior frontal gyrus
Left superior frontal gyrus
Right precuneus
Bilateral middle occipital region	L1 and L2 across‐language overlap in picture naming tasks
Bilateral fusiform region
Bilateral inferior temporal regions
Left precentral gyrus	L1 syntax processing areas closely related to those of L2
Left basal ganglia
Left Inferior frontal gyrus
Supplementary Motor Area
Left Inferior frontal gyrus	The wider areas in late bilinguals consistently recruited more than early bilinguals
Broadmann area 45 and the dorsolateral prefrontal cortex border

Abbreviations: First language. L2; L1; Second language. (Dodel et al., 2005; Indefrey et al., 2005; Luke et al., 2002; Sakai et al., 2004; Van de Putte et al., 2018).

It is clear that the issue of cortical and network utilization between L1 and L2 remains under debate. Given that language is a complex function formed of vocabulary, syntax, and grammar, it may be that different rules of representation apply to different elements of language. This review encourages further research into this area.

4.2.3. Gray matter changes

The process of acquiring a second language elicits structural and cytoarchitectural changes and increases gray matter density, particularly in the inferior parietal lobule (IPL) (Abutalebi et al., 2012; Friederici, 2009; Luo et al., 2019; Olsen et al., 2015). Two studies regarding gray matter changes in bilingualism (Abutalebi et al., 2015a; Mechelli et al., 2004), both utilizing Voxel‐Based Morphometry (VBM), found that bilinguals had significantly greater volume of gray matter in their left and right IPL. In addition to the IPL, further studies using structural and event‐related magnetic resonance imaging (MRI), VBM, and T1 structural MRI scanning have demonstrated increased gray matter density in the left anterior inferior temporal gyrus (Abutalebi et al., 2014), the left putamen (Abutalebi et al., 2013a, 2013b) and in the cerebellum (Pliatsikas et al., 2014).

In addition, cortical thickness cross‐sectional analysis measured using MRI datasets showed greater thickness of the pars triangularis and pars orbitalis, and a reduced cortical thickness in the right orbital gyrus for late bilinguals compared with monolinguals (Klein et al., 2014). Interestingly, it has also been shown via structural MRI data that bilingual individuals had a significantly greater gray matter volume on both sides of Heschl's gyri compared with monolinguals (Ressel et al., 2012). Here we can see several cortical regions that have been evidenced to undergo structural changes due to L2 acquisition.

There is also an observed greater gray matter density in the anterior cingulate cortex (ACC) of bilinguals (Abutalebi et al., 2012, 2015b; Olsen et al., 2015). The ACC is thought to be a primary area responsible for switching between and within languages as well as in verbal and non‐verbal conflict management as evidenced by a fMRI linguistic switching task study (Abutalebi & Green, 2007; De Baene et al., 2015; Hernandez, 2009; Proverbio et al., 2004; Sierpowska et al., 2013; Wang et al., 2007). The increased ACC activation and gray matter density seen in bilinguals is thought to induce a greater degree of neuronal plasticity and therefore be, at least partially, responsible for their superior performance in certain cognitive tasks (Abutalebi et al., 2013a, 2013b, 2014).

Bimodal bilinguals also have an increased gray matter volume in the head of the left caudate nucleus and greater functional activation of this region during language switching tasks (Zou et al., 2012). A positive correlation is seen between gray matter volume and the size of the switching effect in the head of the left caudate nucleus when using VBM (Zou et al., 2012). However, there are limited data available regarding bimodal bilinguals.

The proposed explanation for deep gray matter changes in bilinguals is that their basal ganglia strengthen already formed cerebral connections so that they may over‐ride the most active connection which is essential for language switching (Stocco et al., 2010). Hence the basal ganglia in a bilingual brain experiences structural and functional changes per the conditional routing model (Stocco et al., 2010). Current literature is mostly in agreement that the process of learning a second language elicits several distinct gray matter changes. However, there is less agreement regarding the implications of these gray matter changes and how these may affect an individual's susceptibility to certain cerebral pathologies.

4.2.4. White matter changes

In addition to these gray matter changes, bilinguals also demonstrate increased myelination of certain white matter tracts such as the corpus callosum. Frontal lobe white matter volume is also significantly higher in adult bilinguals than in adult monolinguals (Gold et al., 2013; Olsen et al., 2015; Pliatsikas et al., 2015). Studies using MRI imaging, DTI, and functional resting‐state data found that bilinguals and multilinguals have greater white matter density in the corpus callosum, the superior and inferior longitudinal fasciculi bilaterally, the frontal–occipital fasciculus, and the uncinate fasciculus (Della Rosa et al., 2013; Luk et al., 2011a, 2011). This is thought to be due to experience‐related structural reorganization as a direct consequence of the increased cognitive load associated with the acquisition and maintenance of L2.

Longitudinal studies using MRI to observe the process of second language acquisition found that structural changes are most evident in the IFG, middle frontal gyrus (MFG), STG, caudate nucleus and their connecting white matter tracts, particularly the corpus callosum (Martensson et al., 2012; Schlegel et al., 2012; Stein et al., 2012). These effects are associated with greater anterior to posterior connectivity extending posteriorly to the superior longitudinal fasciculi, and anteriorly to the right inferior frontal–occipital fasciculus and uncinate fasciculus (Luk et al., 2011a, 2011).

Other changes that are worth considering but have less supporting evidence are those related to histological and neuronal plasticity in bilinguals. Activity and experience are known to dynamically regulate adult neurogenesis (Song et al., 2016) and following fMRI, structural MRI (sMRI), DTI, and fluorodeoxyglucose‐positron‐emission tomography (FDG‐PET) studies, it is thought that bilingualism could play an important role in the neurogenesis changes induced by cognitive activity (Kim et al., 2019). Vascular endothelial growth factor (VEGF) is an important promoter of adult neurogenesis (Licht et al., 2011) and VEGF levels increase as a result of activities such as bilingualism as these activities require high intensity hippocampal‐dependent cognitive activity (Oh et al., 2012). VEGF indirectly increases neurogenesis via increased release of brain‐derived neurotropic factor (BDNF) and directly via a fetal liver kinase 1 (Flk‐1)‐dependent mechanism (Nowacka & Obuchowicz, 2013). It is therefore suggested that bilingualism increases neurogenesis. However these suggestions are still largely theoretical and the supporting evidence is weak therefore further studies are required for confirmation (Kim et al., 2019).

In summary, it is clear that bilingualism induces both gray and white matter changes (Figures 3 and 4). These changes are particularly notable in the frontal lobes, ACC, left IPL, and subcortical structures such as the left caudate, putamen, and the supramarginal gyrus (SMG) as these are the anatomical areas involved in the processing, interpretation and production of language. These changes result in enhanced executive functions, greater mental flexibility, and an increased capacity for conflict resolution beyond that relating to language (Kroll & Bialystok, 2013). It is proposed that this advantage is achieved via fine‐tuning of the executive control system which overlaps with language networks in the fronto‐parietal network, frontal cortex, and ACC (De Baene et al., 2015).

FIGURE 3.

Common areas of increased gray matter within the bilingual brain. ACC, anterior cingulate cortex; AITG, anterior inferior temporal gyrus; HG, heschl's gyrus; HoCN, head of the caudate nucleus; IPL, inferior parietal lobule; P, putamen (Servier, 2021) [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 4.

Common areas of increased white matter within the bilingual brain. CC, corpus callosum; FL, frontal lobe; FOF, frontal‐occipital fasciculus; ILF, inferior longitudinal fasciculus; SLF, superior longitudinal fasciculus; UF, uncinate fasciculus (Servier, 2021) [Colour figure can be viewed at wileyonlinelibrary.com]

4.3. The impact of proficiency and age of acquisition

4.3.1. Introduction

Bilingual language representation results from the combination of many factors, and different activation patterns are often attributed to age of acquisition and level of proficiency as two of the more influential determinants (Luk et al., 2011, 2011b; Perani et al., 1998). It has previously been suggested that level of proficiency carries greater cortical representation influence than age of acquisition; however, the extent of their effect and their structural, functional, and clinical neuroanatomical implications are still unclear. Two such factors with potentially significant neuroanatomical and clinical implications require further clarity and evaluation.

At the turn of the 20^th century, discussion regarding second language education was primarily focused on pedagogical techniques (Stern, 1976). Schooling patterns were largely ignored which resulted in the age of pupils at which languages could most effectively be learned being overlooked (Stern, 1976). Since research on bilingualism first emerged, children have been exposed to second and third languages at far younger ages and begin formal language education earlier than they would have done previously (Pinter, 2011). Evidence now consistently promotes that the learning of a foreign language should begin before one's teenage years and perhaps even before primary school (Kovelman et al., 2008; Stern, 1976). This may be justified using both the critical period hypothesis (Rothman, 2008; Singleton, 2005) and Jean Piaget's theory and stages of cognitive development (Piaget, 1971; Pinter, 2011). The anatomical representations and structural changes associated with L1 and L2 are influenced by level of exposure, degree of proficiency, and in particular by the age at which L2 is acquired (Archila‐Suerte et al., 2015; Bryll et al., 2009; Claussenius‐Kalman et al., 2019; Kuzmina et al., 2019; Perani et al., 1998; Singleton & Lengyel, 1995). White and gray matter changes have both been positively correlated with proficiency and negatively correlated with age of acquisition (AOA) (García‐Pentón et al., 2016; Gold et al., 2013; Luk et al., 2011; Mechelli et al., 2004; Perani et al., 1998; Pliatsikas et al., 2016; Sugiura et al., 2015; Wei et al., 2015).

4.3.2. Proficiency

Gray matter volume density increases in the IFG, caudate, STG, SMG, putamen, and the pars opercularis as proficiency in both L2 and third languages (L3) increases (Abutalebi & Green, 2007; Abutalebi et al., 2013a, 2013b; García‐Pentón et al., 2016; Stein et al., 2012). However, proficiency not only influences the anatomical changes associated with bilingualism but also influences the areas of the brain recruited (Abutalebi et al., 2008; Marian et al., 2014). In less proficient bilinguals, there is increased recruitment of additional cognitive areas such as Brodmann areas 9 and 47 of the prefrontal cortex (Abutalebi et al., 2008; Marian et al., 2014), suggesting that there are wider differences between L1 and L2 during the earlier phases of L2 acquisition.

Initially, bilinguals have a greater degree of left IFG activation (Indefrey, 2006) which can be explained via the efficiency of neuronal organization principle. Namely, several fMRI and hemodynamic studies have provided strong reproducible evidence suggesting that in the early stages of L2 acquisition, a greater neuronal circuit, with a greater number of disparate anatomical regions as well as more neurons within existing circuits, is required to perform a language task than would be required for L1 or a more proficient L2 speaker (Green, 2003; Indefrey, 2006; Wartenburger et al., 2003). It can also be concluded with equal confidence that as proficiency in L2 increases, the greater prefrontal activations observed in L2 are reduced in accordance with Green's convergence hypothesis (Green, 2003). Following achievement of a native like‐proficiency in L2, the prefrontal activations observed in L2 cease to exist and the neural representation of L2 resembles that of L1 (Perani & Abutalebi, 2005). These findings are true for both semantic and grammatical functions of language and are widely replicated throughout the literature (Wartenburger et al., 2003).

The suggestion that increased left IFG activation in lesser proficient individuals is a compensatory mechanism is contrary to the finding that less proficient bilinguals have been shown to activate fewer anterior cortical areas during sentence and discourse processing (Hernandez et al., 2007; Perani et al., 1998, 2003). However, this has been attributed to the less sophisticated linguistic comprehension and less elaborate sentence construction generally associated with L2 compared with L1. It could thus be considered as a reflection on the differences between semantic vocabulary and understanding a wider context of language (Hernandez et al., 2007; Perani et al., 1998, 2003).

4.3.3. Age of Acquisition

Age of acquisition and proficiency are unlikely to be mutually exclusive for early learners and therefore both factors must be considered. A later AOA is negatively associated with gray matter density and L2 proficiency (Mechelli et al., 2004). Within the frontal‐lobe language‐sensitive region of Broca's area, an fMRI study of the superior edge of the anterior commissure and the inferior edge of the posterior commissure showed that second languages in late bilinguals are spatially separated from L1 (Kim et al., 1997). However, fMRI also illustrated that in early bilinguals, L1 and L2 tend to be represented in common frontal cortical areas (Kim et al., 1997). Interestingly, in both late and early bilinguals, the temporal‐lobe language‐sensitive regions of Wernicke's area show little or no separation of activity when imaged with fMRI (Kim et al., 1997).

It is suggested that AOA mainly affects cortical representation of grammatical processes (Wartenburger et al., 2003) and has no major role in the lexicosemantic domain, suggesting that lexicosemantic processing of L2 can be processed in L1 regions regardless of AOA (Indefrey, 2006; Perani & Abutalebi, 2005). This is consistent with a study (Kim et al., 1997) demonstrating that temporal lobe comprehension tasks may be less influenced by L2 AOA than speech tasks. fMRI and connectivity analysis studies have also shown that there is greater left prefrontal cortex activity in late bilinguals compared with early bilinguals, which further suggests that AOA contributes to the cerebral representation of language (Dodel et al., 2005; Indefrey et al., 2005; Sakai et al., 2004).

The effects of AOA and proficiency can be hard to separate as they are not often controlled in studies. However, an fMRI study in early Spanish–English bilinguals and late English‐Spanish bilinguals matched for proficiency found increased neural activity in the left IFG in the late bilinguals compared with early bilinguals (Hernandez et al., 2007). This suggests AOA influences cortical language representation independent of proficiency and that late bilinguals require greater cognitive recruitment to perform at a similar level of proficiency.

To conclude, both AOA and proficiency influence the structural changes and the cortical areas that are activated in bilinguals. It is observed through fMRI studies that both AOA and level of proficiency impact the neural substrate of L2, but that they have differential effects on semantics and grammar and that level of proficiency is a major determinant of brain pattern activity for semantic judgment (Wartenburger et al., 2003). Noteworthy areas include the left IFG, putamen, prefrontal cortex, and Broca's and Wernicke's area. Current evidence indicates that proficiency may have a greater impact on neural representations of L2 than AOA.

4.4. Pathology associated with bilingualism

4.4.1. Introduction

The cortical regions responsible for language and language switching (ACC, left prefrontal cortex, left caudate and SMG) are also responsible for cognitive control and attention (Abutalebi et al., 2012; Luk et al., 2011). This might imply that bilingual recruitment of these networks enhances these pathways and may therefore contribute to the greater cognitive task performance and lesser cognitive decline observed in bilinguals when compared with monolinguals (Abutalebi et al., 2012; Bak et al., 2014; Bialystok et al., 2012; Stepan‐Buksakowska et al., 2012; Wong et al., 2016). The neural networks associated with bilingualism overlap with those that tend to deteriorate with age or with certain cerebral pathologies (Grant et al., 2004; Mechelli et al., 2004; Pliatsikas, 2020). Therefore, there is evidence that bilingualism may have protective cognitive implications (Grant et al., 2004).

4.4.2. General cognitive health

Current literature largely agrees that bilingualism elicits beneficial cognitive health effects. Bilingualism may have a protective effect on intelligence against age‐related decline regardless of AOA, previous intelligence, gender, and socioeconomic status (Bak et al., 2014) with a positive effect of bilingualism on later‐life cognitive reserve.

A review by Grant A et al provided an explanation using the posterior‐to‐anterior shift in aging (PASA) hypothesis (Dobbins & Davachi, 2006; Grant et al., 2004). This hypothesis suggests that as individuals’ age, greater DLPFC activity, and lesser left visual cortex activity are displayed during memory tasks, while the converse is observed in younger adults. PASA has been described as a compensatory mechanism for the reduced posterior brain region activation associated with age (Grant et al., 2004). It is known that bilingualism enhances the frontal cognitive systems (Ossher et al., 2013) as well as preventing decline in posterior cortical areas and enhancing posterior to frontal cortical connections (Ossher et al., 2013). This suggests that due to the lesser posterior decline experienced in bilinguals, they may experience a lesser degree of PASA (Dobbins & Davachi, 2006). In addition, the degree of PASA that they do experience will be shifted to an enhanced superior frontal system and any fronto‐posterior communication will be better preserved (Luk et al., 2011a, 2011). In summary, bilinguals have greater preserved gray matter posteriorly, enhanced frontal cerebral networks, and greater fronto‐posterior communication (Luk et al., 2011a, 2011). This therefore results in greater brain and cognitive reserve and consequently a lessened PASA, greater executive and memory functioning and superior generalized cognitive health (Dobbins & Davachi, 2006; Grant et al., 2004).

4.4.3. Aphasia

Aphasia has proven useful in understanding language processing in bilinguals. For instance, it has previously been concluded in a 65‐study meta‐analytic review of 130 patients that the severity of acquired aphasia in bilingual individuals depends on several factors. These include AOA, proficiency, level of exposure, number of languages known, and the sequence in which they were learnt (Kuzmina et al., 2019). The included studies from which these seemingly robust conclusions are drawn used several methodological approaches to directly measure language performance so that they could assess how the extent of the bilingual aphasia was affected by these factors. These included picture naming, written, and auditory syntactic comprehension tasks in patients aged 17–91 of varying levels of proficiency, age of acquisition, language of origin, and language similarity. Other studies within the literature further support these findings (Clarke & Bonakdarpour, 2019; Dickerson, 2011; Marshall et al., 2018).

There are two major theories regarding aphasia in bi/multilinguals; the localiZationist approach and the dynamic/shared representation (Kuzmina et al., 2019). Combining these two hypotheses forms the amalgamated hypothesis which states that languages share cortical regions, but that they may also occupy separate, independent areas (Ansaldo et al., 2008; Conner et al., 2018; Kuzmina et al., 2019; Lorenzen & Murray, 2008). This closely follows the results from previously mentioned studies that show that L2 and L1 are cerebrally represented in a very similar manner, but that factors such as AOA and level of proficiency can result in activation and recruitment of different areas.

The pattern of recovery from aphasia is described by two opposing laws: Pitre's rule and Ribot's Law. Pitre's rule states that initial recovery is of the most intensively used language and Ribot's law states that initial recovery is of the earliest acquired/native language (Pearce, 2005). Both Pitre's rule and Ribot's law are supported by experimental evidence: however, Ribot's law has since been discovered to only be true in patients who are not truly fluent in L2 (Pearce, 2005). Additionally, many conflicting observations within current studies have led to no universal rule within bilingual aphasics being established. Consequently, these rules must be interpreted with caution and the recovery of bilingual language within aphasic patients must be further defined.

One of the greater challenges regarding aphasic bilinguals is rehabilitation, as it is still not completely understood if both L1 and L2 should be rehabilitated or if greater emphasis should be placed on one language over the other. Ansaldo and Saidi (2014) used cross‐linguistic therapy (CLT) with potential advantageous effects; these were dependent on several factors including pre‐ and post‐morbid proficiency. They also found that semantic approaches result in better cross‐linguistic transfer of therapy effects (CLTE) than phonological approaches and that cognates have better CLT potential than non‐cognates. However, the cognate advantage disappears if cognitive control circuits (the prefrontal cortex, ACC, the basal ganglia, and IPL in the left hemisphere) are damaged.

It is most likely that L1 and L2 recover at the same rate, but that discrepancies are a result of the relative differences in AOA and proficiency (Fabbro, 2001). Despite this, many of the questions regarding the management and rehabilitation of bilingual aphasics are still without adequate answers and considerably more research into this area is required to develop our understanding of and approach to rehabilitation.

4.4.4. Alzheimer's disease

Current evidence indicates that acquiring and utiliZing more than one language affects the neuroplasticity of bilingual individuals (Mechelli et al., 2004; Pliatsikas, 2020). These changes include: increased and strengthened synaptogenesis (Calvo et al., 2016; Kim et al., 2019), increased gray and white matter density (Li et al., 2014), increased neurogenesis (Kim et al., 2019), and enhanced functional connectivity (Kim et al., 2019). These structural changes have been demonstrated in children, adolescents, and in the elderly and are sensitive to AOA, proficiency, and language‐specific characteristics (Li et al., 2014). It is hypothesized that the bilingual brain undergoes these changes due to increased levels of cognitive stimulation and executive control and that this is likely to protect against cognitive decline (Bialystok et al., 2007; Kim et al., 2019).

In Alzheimer's disease, gray matter volume often declines in the anterior hippocampus, the parahippocampus, and the precuneus (Kim et al., 2019; Raji et al., 2009). The neuronal changes induced by bilingualism promote diffuse brain area integration which may attenuate this pathological cognitive decline (Kim et al., 2019; Palop et al., 2006). Bilingualism‐induced plasticity is thought to enable the recruitment of alternative cerebral networks as well as the strengthening of existing networks as compensatory mechanisms (Marian & Shook, 2012).

The available literature therefore appears to unanimously indicate that the process of learning a second language offsets and ameliorates both pathological and age‐related cognitive decline though associated structural and functional changes. Despite further evidence being required before definitive conclusions can be drawn, of the numerous studies currently available, several high‐quality papers have found that multilinguals are diagnosed between 4 and 5 years later than monolinguals with Alzheimer's disease (Bialystok et al., 2007; Chertkow et al., 2010; Craik et al., 2010; Gollan et al., 2011; Schweizer et al., 2012). The absence of contrary findings within the current evidence base appears to further support these conclusions. The underlying principles providing an explanation for these conclusions are those of cognitive and brain reserve as described above (Gold et al., 2013; Grant et al., 2004; Stern, 2013).

A further study comparing white matter integrity and gray matter volumetric patterns using high‐resolution structural imaging and DTI (Gold et al., 2013) found that the parts of the brain that were identified as displaying greater white and gray matter density in bilinguals are associated with the memory circuit. This included the hippocampus, entorhinal cortex, and parahippocampal white matter (Gold et al., 2013). These are often areas of the brain that experience degeneration in Alzheimer's disease (Stepan‐Buksakowska et al., 2012), thereby supporting the idea that bilingualism may offset or delay the symptoms of dementia.

Based on the available literature, this review suggests that there is a significant level of high quality clinical research to conclude that learning a second language ameliorates pathological and age‐related cognitive decline and that the homogenous positive findings of current research serve to further support this conclusion. While plasticity as a factor relating to age‐mediated cognitive decline and Alzheimer's disease are relatively well understood, there is a relative paucity of data relating to the histological changes and epigenetic influences of bilingualism. This review encourages further exploration of these areas.

4.4.5. Other diseases

The relationship between bilingualism and other neurological pathologies such as Parkinson's disease remains unclear. A cross‐sectional study of Welsh‐English bilinguals explored the potential of a bilingual advantage in executive function in the context of Parkinson's patients (Hindle et al., 2015). The study found no significant difference between monolinguals and bilinguals in the executive function tests. Further studies are required to elucidate the effects of bilingualism on Parkinson's disease. There has also been very little research into the effects of bilingualism on other cerebral pathologies such as epilepsy, stroke, and amnesia.

5. CONCLUSION

This review has brought together the current evidence surrounding the acquisition, comprehension, storage and articulation of language to offer what we hope is a clear and coherent outline of our present understanding. It summarizes relevant research regarding the neuroanatomical changes experienced by bilinguals and relates this to the neuroanatomy of language with the intention of consolidating the current literature base.

It is largely accepted that the cerebral structures of language reside around the Sylvian fissure in the language dominant hemisphere. However, in recent years increased attention has been devoted to areas of the brain outside the perisylvian language zone such as the superior, middle and inferior temporal gyri, the basal ganglia, and the cerebellum. It appears that recognition of speech is primarily conducted in the superior temporal lobe circuits bilaterally while speech production is carried out in the fronto‐parietal and temporal circuits of the left hemisphere. Although the neuroanatomical pathways involved in the processing of language and phonological information are broadly understood, those relating to complex aspects of language such as grammatical comprehension and syntax construction remain unclear.

In the fields of bilingualism and neuroanatomy, an enduring question is whether a second language occupies the same anatomical location and cognitive pathways as a native language. The evidence supporting separate cerebral loci is largely theoretical and has been superseded by more recent functional neuroanatomical data. Following analysis of the included studies, it is possible to conclude that L2 mostly utilizes the same neuroanatomical pathways as L1 via a complex interplay and that L1 and L2 recruit the same anatomical structures but that these demonstrate different functional activity. Second language processing ultimately utilizes the same neuronal networks as those used for a first language. However, the presence of language‐specific cortical centers in bilinguals has been evidenced. Therefore, the issue of cortical and network utilization between L1 and L2 remains under some debate.

Learning a second language induces structural and cognitive changes that include greater cerebral activity and executive control as well as increased synaptogenesis and neurogenesis. The process of learning a second language increases both gray and white matter. These changes are particularly notable in the frontal lobes, corpus callosum, ACC, left IPL and subcortical structures including the left caudate, putamen, and SMG. White and gray matter changes have both been positively correlated with proficiency and negatively correlated with age of acquisition. There are subtle neuroanatomical differences observed between L1 and L2 and these can be attributed to the differing degrees of proficiency and AOA.

The neuroanatomical changes in bilinguals result in enhanced executive functions, greater mental flexibility and an increased capacity for conflict resolution beyond that relating to language. Bilingualism has been evidenced to improve frontal cognitive systems as well as preventing decline in posterior cortical areas and enhancing posterior to frontal cortex connections. This results in greater brain and cognitive reserve and consequently a lessened PASA.

Bilingualism may offset and ameliorate both pathological and age‐related cognitive decline though the associated structural and functional changes. However, there is a relative paucity of data relating to the histological changes and epigenetic influences of bilingualism. There is also little research into the effects of bilingualism on other cerebral pathologies such as epilepsy, stroke, and amnesia. This review encourages further research into these areas.

Author contributions

Charles Taylor contributed equally to the conceptualization and design of this research and to the acquisition of data and data analysis/interpretation as well as to the critical revision of the manuscript and approval of the article. Charles Taylor led the drafting of the manuscript. Mr Samuel Hall led the conceptualization and design of this research and contributed equally to the acquisition of data, data analysis/interpretation drafting of the manuscript, critical revision of the manuscript, and approval of the article. Mr Susruta Manivannan contributed equally to the conceptualization and design of this research as well as to the acquisition of data, data analysis/interpretation, drafting and critical revision of the manuscript, and approval of the article. Mr Nilesh Mundil contributed equally to the conceptualization and design of this research as well as to the acquisition of data, data analysis/interpretation, drafting and critical revision of the manuscript, and approval of the article. Dr Scott Border contributed equally to the conceptualization and design of this research as well as to the acquisition of data, data analysis/interpretation, drafting and critical revision of the manuscript, and approval of the article.

Supporting information

Material S1

Taylor, C. , Hall, S. , Manivannan, S. , Mundil, N. & Border, S. (2022) The neuroanatomical consequences and pathological implications of bilingualism. Journal of Anatomy, 240, 410–427. 10.1111/joa.13542