Cognitive and brain reserve in bilinguals: field overview and explanatory mechanisms

There is an ongoing debate on potential neuroprotective effects of bilingualism against cognitive decline during healthy aging. In this paper, we consider the neural and cognitive mechanisms through which these protective effects may operate. We review the evidence suggesting that bilingualism can act as a booster of neuroplasticity and/or as a brain protection mechanism providing effective compensation. Our main aim is to better define the linkage between reserve and lifetime bilingual experience and their effects on the mind and brain. We first illustrate the concept of reserve and contextualize existing results of bilingualism research within the reserve framework. Then, we discuss how bilingualism-induced enhancements of certain cognitive functions may constitute the basis for the neural underpinnings of reserve, i.e., brain reserve (BR) and cognitive reserve (CR). Finally, we discuss how the interplay between BR and CR fostered by multiple language use can provide protection to the aging brain.


Introduction
Human life expectancy has been growing steadily in recent decades and this increase is likely to continue in the forthcoming years (Raftery et al., 2013;Kontis et al., 2017). This is resulting not only into the obvious benefits of an extended life span but also in some adverse consequences including the increased risk of suffering from cognitive impairment and dementia, which presently constitute two of the major causes of decreased quality of life in the elderly (for a review, see Shearer et al., 2012). Along with a significant decrease in individuals' quality of life (patients' and caregivers'), increasing rates of dementia will result in healthcare resource demands that are projected to increasingly burden the governments of industrialized countries (Prince et al., 2016;Winblad et al., 2016;Livingston et al., 2017;Wimo et al., 2017). Concerning the purely economic consequences, recent studies show that senior citizens (aged 65 or above) in the United States constitute about 15% of the total population but that they also account for 34% of total health spending (Mitchell, 2016). A similar proportion of senior citizens in Canada accounts for an even larger healthcare spending share of 45% (Canadian Institute for Health Information, 2014). A study supported by the American National Institute of Health (Hurd et al., 2013) estimated a $215 billion expenditure for dementia for 2010 alone just in the US, highlighting the enormous care and treatment needs imposed by the aging population.
In such a dramatic scenario and considering the non-existence of any effective pharmacological treatment, a search for factors that may act as a buffer against neurocognitive decline constitutes one of the central issues for the community today. Ideally, one should aim at identifying ''ecological'' interventions in order to prevent diseases related to aging. The meaning here attributed to the term ''ecological'' is two-folded: on the one hand, governments should be able to implement such interventions through lowercost policies compared to current healthcare expenses; on the other hand, such interventions should be introduced gradually but precociously in individuals' everyday life in order to build a lifelong ''armour'' against age-related decline in later life stages.
Several of these protective lifestyle factors have been identified in the recent years, ranging from higher educational and occupational levels to mentally and physically stimulating leisure activities (Clare et al., 2017;Rouillard et al., 2017;Wang et al., 2017; for a review, see Cheng, 2016). The concept of reserve (see e.g. Stern, 2009) groups together all the aforementioned lifestyles and activities, which account for the interindividual differences observed in rates of cognitive decline. Reserve has been defined as the discrepancy between the severity of brain damage or agerelated deterioration and the resulting level of cognitive impairment (Stern, 2009). It may act as a mechanism of protection and compensation against cognitive decline, allowing individuals to better cope with age-related cognitive impairments (Stern, 2009;Bartrés-Faz & Arenaza-Urquijo 2011;Steffener et al., 2011). Hence, the protective factors mentioned above would support ''successful aging'' by fostering an individual's reserve. For a comprehensive overview on the concept of reserve, we refer the reader to Stern's work (e.g. Stern 2009. The present work will discuss bilingualism as a factor that may promote reserve and hence lead to healthy aging. Bilingualism and cognitive efficiency during healthy aging As mentioned above, widespread evidence suggests that second language use promotes the maintenance of cognitive and neural efficiency during aging (e.g. Bialystok et al., 2004;Gold et al., 2013b;Abutalebi et al., 2015b;Bialystok et al., 2016;Del Maschio et al., 2018). An initial indication supporting the role of bilingualism as a booster of reserve stems from studies associating multiple language use to enhancements in executive control (EC) across the lifespan (for a review, see Bialystok, 2017), despite an ongoing debate questioning such benefits Duñabeitia et al., 2014;Gathercole et al., 2014;Paap et al., 2015). In general, the effects of bilingualism upon executive functioning have been investigated by utilizing various EC tasks and analyzing differences between bilingual and monolingual individuals in resulting processing indices such as response accuracy and reaction time (RT). Early evidence indicated that bilinguals across different age groups outperform monolinguals in such tasks (for a review, see Bialystok, 2017). However, many recent studies failed to replicate bilingual effects in similar tasks (for a review, see Paap et al., 2015).
These conflicting findings may be explained if one clearly identifies the individual cognitive components that make up 'executive control' or 'executive functions' (Costa et al., 2009). For instance, Costa et al., (2009) reported the presence or absence of bilingual effects depending on the precise percentage of congruent versus incongruent trials in a given experimental task. The authors found that bilingual effects were present when the experiment involved highly demanding trials with mixed congruency (e.g., 50% congruent and 50% incongruent trials), but when the experiment had only or mostly (e.g., [ 90%) trials of one type, bilingual advantages would disappear. This seminal work pointed to the significance of understanding EC from the perspective of task demands linking the specific effects to cognitive 'monitoring'. An alternative account suggests that younger generations in the present digital era are exposed to a multitude of cognitively challenging activities inevitably boosting their EC. Hence, potential bilingual effects in younger participants may be largely masked by their generally stronger executive functioning. Older individuals, particularly those in retirement, are not exposed to so many cognitively challenging activities anymore; therefore, the effects of bilingualism upon their cognitive performance may be stronger (Valian, 2015). Importantly, the latter is only true if the speaker is still exposed to her second language and actively uses it (Abutalebi et al., 2015a).
Overall, while there is no agreement on the effects of bilingualism on boosting EC in younger subjects, the picture is much more conclusive when it comes to senior individuals. Indeed, there is now plenty of evidence reporting that older bilinguals do maintain better cognitive efficiency, outperforming their monolingual peers on a number of different cognitive measures not strictly limited to EC tasks (Bialystok et al., 2004(Bialystok et al., , 2008Gold et al., 2013b;Abutalebi et al., 2015b;Estanga et al., 2017;Del Maschio et al., 2018;Incera & McLennan, 2018;Rosselli et al., 2019;Zunini et al., 2019), but extending to executive-related memory recall tasks (Wodniecka et al., 2010;Ljundberg et al., 2013;Rosselli et al., 2019), semantic memory (Arce Rentería et al., 2019) and general intelligence (Bak et al., 2014). This evidence is usually put forward to explain why bilingualism could promote reserve (Perani & Abutalebi, 2015;Bialystok et al., 2016).

Neuroprotection during pathological aging
If bilingualism indeed fosters reserve in healthy aging individuals, then its effects should be arguably strongest when these individuals have to face the burdens of cognitive decline more acutely, for instance, in dementia. Many cross-sectional studies have compared clinical differences between bilingual and monolingual seniors retrospectively. The landmark work by Bialystok and colleagues (2007) showed a striking 4-year delay in the onset of dementia for bilinguals as compared to monolingual peers matched for gender, years of education, and socioeconomic status-all factors known to have an impact on the development of reserve. While these results were replicated in a separate study , findings from two subsequent investigations raised a concern that the bilingualism-induced protective effects may be limited to low-education bilinguals (Gollan et al., 2011) or to immigrant groups (Chertkow et al., 2010). It is in principle possible that the real key to the development of reserve in bilinguals would be the extra cognitive challenges linked to immigration, such as learning a second language (L2) later in life, living in an L2-dominant environment, as well as facing a multitude of other cognitive, cultural and economic adjustments. Moreover, immigration status could constitute a confounding factor per se, as differences in lifestyle, diet, ethnic background, education, and attitude to health may implicitly differ between non-immigrant monolingual and immigrant bilingual populations. In such cases, differences in dementia's age of onset attributed to bilingualism may actually arise from differences in these and other confounding variables. A critical contribution to this debate came from a study by Alladi et al. (2013) conducted in India on a large sample of 648 individuals. The authors reported, in striking resemblance with Bialystok's group, a 4/5-year delay of dementia symptoms in native bilinguals compared to native monolinguals. The issue of education level was also addressed by conducting a separate analysis on illiterate participants showing an even larger delay of dementia onset for bilinguals (i.e., 6-years), confirming the assumption that the beneficial effects are stronger in low-educated individuals.
A study conducted in Belgium provided further corroborating evidence (Woumans et al., 2015) by comparing bilingual and monolingual Alzheimer's dementia (AD) patients matched for gender, education, occupational complexity, and cognitive impairment severity. Consistently with previous literature, clinical manifestation of AD appeared to be delayed by 4.6 years in bilingual patients. More recent studies confirmed bilingualism's effect on the delay of AD onset: Zheng et al. (2018) reported a 7-year delay in Cantonese/Mandarin bilinguals as compared to both Cantonese and Mandarin monolinguals, while Mendez et al. (2020) reported a 4-year delay in a sample of bilinguals with different L1s and English as an L2. Moreover, Perquin et al. (2013) examined 232 senior multilinguals in Luxembourg who were either classified as healthy or affected by mild cognitive impairment (MCI). The retrospective analysis showed that MCI incidence was inversely correlated to the number of languages spoken. These results were replicated in a follow-up study by Wilson et al. (2014) who followed a cohort of 964 individuals over a 6-year period, observing the incidence of MCI with time. Bilinguals developed MCI to a significantly lesser extent than monolinguals. In a retrospective study conducted in India on a cohort of 115 MCI individuals, bilingualism has also been shown to delay the onset of MCI by 7.4 years (Ramakrishnan et al., 2017). Finally, bilinguals have been reported to develop Parkinson's Disease-related cognitive decline 3 years later than their monolingual peers (Saidi, 2019).
Even in the face of increasing evidence that bilingualism may act as a booster of reserve during healthy and pathological aging, the exact cognitive and neural mechanisms behind the phenomenon remain uncertain. In the next section we will review some evidence on the neural basis of reserve in bilinguals and an attempt will be made to understand the potential neural mechanisms that may boost reserve in speakers of multiple languages.
Building up the reserve: effects of bilingualism on mind and brain Based on the evidence above, one may argue that speaking multiple languages leads to the development of reserve. At the same time, we know relatively little about how reserve is built up and what its underlying cognitive and neural mechanisms look like. It has been postulated (see Abutalebi & Green, 2016) that one of the key mechanisms is to be found in the increased cognitive challenge that bilinguals have to face on a daily basis. This challenge is linked to the fact that bilinguals, differently from their monolingual peers, have to control their two languages in order to avoid unwanted interferences (Green, 1998). In order to achieve this, the bilingual person uses a cognitive device labelled as ''language control'' that inhibits the language not in use and activates the one to be used. This language control system is neurally tightly linked to the more domain-general EC network (Abutalebi & Green, 2007). This control process comes inevitably with a cost; for this reason, bilingual subjects are typically slower in picture naming and various other language related tasks (e.g., Ivanova & Costa, 2008). Following this logic, by using the control device regularly, bilinguals develop the neural structures underlying domain-general EC in terms of either increasing grey and/or white matter related to the executive network or setting up more efficient brain connections. This, in turn, could eventually compensate for the loss of brain structure and function seen in neurodegenerative disorders, as we will also illustrate below. As to its neural underpinnings, as mentioned, bilingual language control is governed by a network comprising cortical and subcortical brain regions largely overlapping with the general-domain EC network, including the anterior cingulate cortex (ACC), the prefrontal cortex (PFC), the caudate nucleus, the inferior parietal lobule (IPL) bilaterally and the cerebellum (Green & Abutalebi, 2013), each with a peculiar role in controlling multiple language usage (see Calabria et al., 2018 for details).
Alternatively, and not strictly related to the EC network, one has also to consider that bilinguals, unlike monolinguals, have two lexicons linked to a somewhat common semantic system (Kroll & Stewart, 1994). Thus, another source of enhancement in bilinguals' reserve may reside in L2 learning per se. Bilingual users learn new L2 words by forming new connections with the existing L1 vocabulary and general semantic knowledge (Kroll & Stewart, 1994) and this learning process could potentially drive neuroplastic changes in the bilingual brain, as observed in imaging studies (e.g. Mechelli et al., 2004;Grogan et al., 2012;Abutalebi et al., 2014;Olsen et al., 2015). At the same time, areas responsible for the storage of lexical and semantic conceptual knowledge (such as the temporal lobes) are heavily affected in both healthy and pathological aging, showing marked signs of neurodegeneration (Fjell et al, 2009). Thanks to the enhanced neuroplastic changes induced by the formation and strengthening of lexico-semantic connections needed to encode items in multiple lexicons, these areas could potentially resist neural decline more efficiently. An example (see below for a more detailed discussion) is the temporal pole, a region suggested to store modalitynonspecific conceptual properties of objects (Lambon Ralph et al., 2008). This area is known to be strongly affected by neurodegeneration in AD (Fjell et al., 2009) and fronto-temporal dementia (FTD, also known as semantic dementia, SD; Mummery et al., 2000). Reduced brain atrophy in this region has been demonstrated in bilingual seniors (Abutalebi et al., 2014) who appeared to maintain higher gray matter volumes (GMV) in the left temporal pole than their monolingual peers. Further confirming a role of new vocabulary learning in fostering neuroplastic changes in the bilingual brain, GMV levels were found to be positively associated with L2 naming scores (Abutalebi et al., 2014).
Brain reserve, neural reserve and neural compensation If we take for granted that bilingualism can provide protection against age-related cognitive impairment by means of increased reserve, it should also be possible to identify its direct effects on neural structure and functioning. In general, two mechanisms have been advocated to describe the structural and functional underpinnings of the reserve phenomenon: brain reserve (BR) and cognitive reserve (CR; see Fig. 1).
BR has been defined as an individual's capacity for resilience to deterioration of cerebral tissue associated with healthy or pathological brain aging. It may result from the interplay between different factors, some biological (e.g., brain size or neuronal and synaptic counts), others environmental (e.g., neuroplastic structural changes promoted by life experiences). BR could then act as a passive buffer against cognitive decline, and is thus often included in what is labelled as ''passive reserve models'' (Katzman, 1993;Stern, 2009).
CR, instead, may be seen as an active mechanism of compensation against healthy and pathological agerelated decline. In particular, it would allow individuals to better cope with age-related cognitive impairment in two separate, yet interrelated ways (see Fig. 1). On the one hand, the key to neuroprotection would then be neural reserve (NR), defined as the variable degree of flexibility, capacity, and efficiency of an individual's brain networks (Stern, 2009;Bartrés-Faz & Arenaza-Urquijo, 2011;Steffener et al., 2011). Individuals with more efficient brain networks would need less effort to cope with increases in task demands: in the face of the same increment in task difficulty, they would face the need to deploy their neural resources to a lesser extent to achieve optimal results (Steffener et al., 2011). Thus, in case of further rises in task-related cognitive load, they would maintain a ''fuller warehouse'' to draw from, and might therefore resist brain pathology or age-related decline to a greater extent. On the other hand, the protective effect would arise from an individual's level of neural compensation (NC), the ability to develop new approaches to address a task in the face of neural substrate deterioration, by deploying neural resources or cognitive strategies alternative to those usually utilized by healthy individuals (Stern, 2009;Steffener et al., 2011).

Brain reserve in bilinguals
As mentioned above, the development of BR, fostered by lifelong experiential factors, begins well before senescence. Evidence supporting the role of bilingualism in fostering BR has indeed been reported both for young and for older adults. Several studies highlighted differences between bilinguals and monolinguals in gray and white matter density in areas involved in language/executive control and lexico-semantic processing (see Perani & Abutalebi, 2015;Bialystok et al., 2016), consistent with the aforementioned mechanisms underlying bilingual language processing, i.e., that language processing in bilinguals (unlike monolinguals) depends to a certain degree also on EC processes and engages more the lexico-semantic pathways.
Regarding gray matter investigations in young individuals, Mechelli et al.'s (2004) pioneering study revealed significantly increased GMV in bilinguals' left IPL. Moreover, the amount of increase was positively correlated with L2 vocabulary knowledge and inversely correlated with age of L2 acquisition. Follow-up studies replicated findings regarding the IPL (Della Olulade et al., 2015;Del Maschio et al., 2018;Heim et al, 2019) and also revealed increased GMV in the brains of young bilingual adults in a number of other regions including the ACC Del Maschio et al., 2018), cerebellum (Filippi et al., 2011;Pliatsikas et al., 2014), dorsolateral PFC (Stein et al., 2012;Olulade et al., 2015;Del Maschio et al., 2018), and Heschl's gyri (Ressel et al., 2012). Bilingualism has also been shown to affect the structure of the subcortical areas that are part of the language/executive control network. For instance, increased GMV was found in the left Pliatsikas et al., 2017) and right (Pliatsikas et al., 2017) putamen, the left caudate nucleus (Zou et al., 2012), the right thalamus Fig. 1 Schematic overview of the brain mechanisms underlying bilingualism-related protective effects against age-related neurocognitive decline. ACC anterior cingulate cortex, LIFG left inferior frontal gyrus, CAU caudate nucleus, LIPL left inferior parietal lobule (Pliatsikas et al., 2017), as well as bilaterally in the globus pallidus (Pliatsikas et al., 2017).
Results reporting increased GMV in adult bilinguals have been subsequently replicated in different samples of aging individuals, with senior bilinguals showing volumetric enhancements bilaterally in the IPL (Abutalebi et al., 2015a;Del Maschio et al., 2018), ACC (Abutalebi et al., 2015b;Del Maschio et al., 2018), PFC (Del Maschio et al., 2018), temporal pole (Abutalebi et al., 2014;Olsen et al., 2015), and orbitofrontal cortex (Abutalebi et al., 2014). An interesting example is Abutalebi et al.'s 2014 study, in which senior bilinguals showed increased gray matter compared to age-matched monolinguals bilaterally in the temporal poles and the orbitofrontal cortex. Moreover, bilinguals were reported to suffer from overall brain aging effects to a lesser extent than monolingual counterparts. Noteworthy is also the finding that GMV of the left temporal pole-an area related to lexical retrieval and semantics-was positively associated with increasing levels of L2 proficiency. In other words, bilinguals who mastered their second language best appeared to develop the greatest BR. Olsen et al. (2015) reported similar results in a cortical thickness study: confirming the finding that the process of learning an L2 per se induces neuroplastic changes in the bilingual brain, these authors found greater cortical thickness in the temporal pole of bilingual seniors, as compared to monolingual peers. Of note, the temporal poles, alongside the orbitofrontal cortex, are well-known to be among the first cortical areas to suffer from brain atrophy during nonpathological aging (Kalpouzos et al., 2009). The neuroplastic changes induced by bilingualism in this area suggest how multiple language usage may delay the onset of cognitive decline. Moreover, another recent study (Del Maschio et al. 2018), conducted GMV comparisons on several language control network's areas, between both young and senior bilingual and monolingual groups. This investigation revealed increased GMV in the ACC, PFC, and IPL, bilaterally, in bilinguals. Such findings suggest that bilingualism would help to develop BR in the EC network throughout the lifespan, beginning from early life stages. Yet, neuroprotective effects of bilingualisminduced BR do not seem to be confined to healthy aging but may also be extended to brain pathologies. Indeed, GMV loss in IPL is known to be linked to mild cognitive impairment (MCI; Saykin et al., 2006;Apostolova et al., 2007) and early stages of dementia (McDonald et al., 2009). In this light, the findings of a lifelong build-up of enhanced BR in the IPL that span from bilingual children (Della  to young adults (Mechelli et al., 2004;Del Maschio et al., 2018) to seniors (Abutalebi et al., 2015a;Del Maschio et al., 2018) acquire even greater relevance.
The beneficial contribution of bilingualism to BR in senescence extends also to white matter tracts. In a study employing diffusion tensor imaging (DTI), Luk et al. (2011) analyzed fractional anisotropy (FA) values in senior bilingual and monolingual subjects to investigate white matter differences. The results showed higher levels of white matter preservation for the bilingual group in parts of the corpus callosum extending bilaterally into the superior longitudinal fasciculus and to the right inferior fronto-occipital and uncinate fasciculi. As before, bilingualism appears to provide protection in brain areas that typically suffer from age-related deterioration (Pfefferbaum et al., 2005;Gunning-Dixon et al., 2009). Moreover, Olsen et al. (2015) reported higher frontal lobe white matter volumes in senior bilinguals as compared to agematched monolinguals. Finally, Anderson et al. (2018) showed that senior bilinguals maintain higher structural integrity in the left superior longitudinal fasciculus.
Put together, these findings suggest the route via which bilingualism may contribute to preventing healthy and pathological age-related decline, fostering the development of BR. As pointed out above, beside the passive protection provided by BR, the notion of CR comprises the beneficial effects that may be achieved through two different, yet interlinked, active neural mechanisms, NR and NC. Below, we review evidence suggesting that bilingualism may foster the development of CR in both ways.

Efficiency and flexibility: neural reserve in bilinguals
Similar to BR, NR also develops along the entire lifespan with evidence suggesting a role of bilingualism in fostering NR coming from different age groups, spanning from youth to senescence. Luk et al. (2010) administered a modified version of the Flanker task to young bilinguals and monolinguals. A NoGo trial condition, to which participants were instructed not to respond, was added to congruent and incongruent trials. These trials were introduced in order to distinguish between the suppression of interfering responses (as in standard Flanker's incongruent trials) and the general inhibition of any behavioral response. Both for the incongruent (interference suppression) and for the NoGo trials (response inhibition), bilinguals activated a diffuse network including bilateral regions of the inferior frontal and temporal cortices and subcortical regions. Monolinguals, instead, recruited this network for NoGo trials only, activating a more circumscribed network when facing incongruent trials. The authors made two important conclusions based on these findings. First, they concluded that bilingualism leads to differential patterns of neural activation when encountering conflicting information. Second, they interpreted the recruitment of a common network for response inhibition and interference suppression as a sign of enhanced efficiency of bilinguals' EC function. At first glance, the activation of a more widespread network to support a cognitive function may appear as indicating lower, rather than higher, neural efficiency. Nonetheless, one has to consider the bigger picture: while monolinguals showed the necessity to activate a second, differential network for the more demanding interference suppression, bilinguals, due to extensive training in such ability, seemed to recruit regions in the general attention control network, being able to process both kind of trials through a common network.
In another study using the Flanker task with young adults,  reported that bilinguals activated the ACC to a lesser extent than monolinguals while outperforming them on the task, thus showing a more efficient use of this structure to optimize behavioral performance. This result suggests that bilinguals' experience in resolving conflict can better ''tune'' their ACC for conflict monitoring. The study, which employed two runs of the Flanker task, provided further evidence confirming an enhanced flexibility of bilinguals' ACC: while behavioral conflict effect was significantly smaller in the second run for bilinguals, this was not the case for monolinguals. This pattern was also observable in fMRI data: besides activating the ACC to a lesser extent already in the first run, bilinguals showed a dramatic decrease in the signal in the ACC in the second run, while monolinguals did not. Similarly, Rodriguez-Pujadas et al. (2014) reported a more efficient use of the ACC by bilinguals, as compared to monolinguals, during a stop-signal task.
Before discussing the bilingualism-induced effects on NR in seniors, it is useful to briefly report the findings of a study carried out in children. In an eventrelated design study deploying a Stroop and a Simon task, Mohades et al. (2014) compared activation between simultaneous bilingual, successive early bilingual, and monolingual children. Increased activation in the ACC was found for both bilingual groups in the Stroop task in incongruent (i.e. those tapping on conflict resolution) versus congruent trials. In a similar fashion, in the incongruent versus congruent trials of the Simon task, bilingual children showed increased activation, compared to the monolinguals, in the left superior temporal gyrus, bilateral posterior cingulate gyrus, right middle frontal gyrus, and right caudate nucleus. Nonetheless, a more recent study utilizing the same Simon paradigm with senior individuals (Ansaldo et al., 2015) reported that, during incongruent trials, bilinguals recruited the left inferior parietal lobule (IPL) while monolinguals preferentially recruited the right middle frontal gyrus. Whereas monolinguals relied on a standard inhibitory control region of the frontal lobe, known to be particularly vulnerable to age-related impairment, bilingual seniors activated brain regions not typically related to inhibitory control, possibly indicating enhanced resilience. Such pattern was indeed interpreted by the authors as a sign of enhanced NR, reflecting the highly efficient EC developed by lifelong bilinguals, whose experience allowed for optimal conflict resolution without the activation of the widespread network described for the relatively inexpert bilingual children in Mohades et al. (2014). This interpretation is corroborated by the finding that senior monolinguals, who lack such lifelong experience in resolving conflict, activated instead parts of the same network that Mohades et al. (2014) found in children. Again, in a study deploying a nonverbal switching task, Gold et al., (2013b) reported reduced switch costs (brain activation for switch compared to non-switch trials) for senior bilinguals, as compared to monolingual peers, in the left DLPFC, VLPFC and ACC. Moreover, this activation pattern did not differ, for senior bilinguals, from that of young bilinguals and monolinguals. Finally, bilingualism has been shown to affect the functional connectivity of seniors' neural networks: for instance, a resting-state functional connectivity study by Grady et al. (2015) revealed that bilingual seniors, as compared to age-matched monolinguals, maintained stronger functional connectivity in the Default Mode-and Prefrontal Executive Control networks. A more recent study by De Frutos-Lucas et al. (2020) deploying magnetoencephalography (MEG) reported for bilingual seniors higher restingstate functional connectivity than age-matched monolinguals in five occipito-parietal clusters, whose functional connectivity is known to be particularly disrupted in AD (e.g. Nakamura et al., 2017;Yu, et al., 2017).
Taken together, the findings reviewed above indicate that extensive training may foster NR in bilinguals, through enhancements of EC network's efficiency and flexibility. Such enhancements would, in turn, result in neuroprotection from age-related cognitive decline for bilingual users by allowing them to compensate for eventual decreases in their cognitive functioning during healthy or pathological aging.

Facing pathological aging: neural compensation in bilinguals
As follows from the definition of neural compensation, an individual with high levels of NC would still be able to maintain nearly normal cognitive functioning even in the face of brain atrophy and neurodegeneration. Existing literature offers examples of studies that deployed neuroimaging to investigate differences in NC between bilinguals and monolinguals, by comparing neural substrate integrity in groups matched for cognitive functioning, dementia severity or duration. The first such investigation was Schweizer et al.'s (2012) study, which utilized computer tomography to assess differences in brain atrophy between bilingual and monolingual AD patients, matched for education and disease severity. At comparable levels of cognitive impairment, bilinguals appeared to have significantly greater brain atrophy in areas typically affected by AD-related disruption, namely the temporal horn ratio (Zhang et al., 2008), third ventricle ratio and the radial width of the temporal horn (Frisoni et al., 2002a, b). Given that the level of cognitive impairment was matched between the two linguistic groups, such results suggest that bilingual individuals had higher NC, allowing to compensate for the cerebral damage to a greater extent than monolinguals. Similarly, Gold et al. (2013a), compared white matter integrity between senior bilinguals and monolinguals matched for performance levels on a range of cognitive tests, as well as for potentially confounding CR factors including education, socio-economic status (SES), and intelligence. Bilinguals showed lower white matter integrity in the corpus callosum, inferior longitudinal fasciculus, inferior fronto-occipital fasciculus, and the fornix. These white matter tracts are part of the brain's memory circuitry and are typically disrupted in AD and MCI (Stebbins & Murphy, 2009). Given that the cognitive status was comparable between the two groups, the authors interpreted their findings as supporting the view that bilingualism promotes the development of CR. Indeed, senior bilinguals were able to perform at a level comparable to that of monolinguals in spite of lower resources available, suggesting higher levels of NC. This result may, at first glance, seem to contradict the abovementioned studies (i.e. Luk et al., 2011;Olsen et al., 2015;Anderson et al., 2018) showing better maintenance of white matter integrity in senior bilinguals as compared to monolinguals. However, we must consider that the samples of the abovementioned studies included healthy aging individuals, while Gold and colleagues' participants may have had a higher incidence of preclinical AD, as suggested by the specific white matter tracts showing reduced integrity in the bilingual sample. This may explain why, in this particular study, bilinguals showed reduced white matter integrity when compared to healthy monolinguals. Similar results were reported in a study by Perani et al. (2017) utilizing fluorodeoxyglucose positron emission tomography (FDG PET) to investigate brain metabolism and neural connectivity in bilinguals and monolinguals with AD matched for disease duration and general cognitive functioning. The study revealed that cerebral hypometabolism was more severe in bilingual subjects suggesting higher levels of NC than monolinguals. Moreover, increased connectivity was found in the EC and Default Mode networks in bilinguals compared to monolinguals; these functional changes were positively correlated with the degree of lifelong bilingualism. Another study utilizing FDG PET (Kowoll et al., 2016) produced similar results: bilingual seniors diagnosed with MCI and probable AD showed significantly higher levels of glucose hypometabolism in frontotemporal and parietal regions and in the left cerebellum, compared to monolingual peers matched for age, gender and disease severity. Similar differences, denoting enhanced NC in bilingual individuals, were also found in the temporal and parietal cortices, structures typically involved in AD pathology. Finally, a recent study deploying a combination of cross-sectional and longitudinal designs (Costumero et al., 2020) compared the total amount of brain parenchyma (i.e. white matter ? gray matter) in samples of MCI bilinguals and monolinguals matched for cognitive status. Bilinguals showed significantly lower parenchymal volumes, suggesting that they were able to successfully compensate for brain atrophy in order to reach levels of cognitive performance comparable with monolinguals. Additionally, the longitudinal analysis revealed that monolinguals lost more parenchyma and suffered more cognitive decline than bilinguals in a mean follow-up period of 7 months.
The studies reviewed above support the role of bilingualism in fostering CR development, in the form of enhanced NC, as bilinguals appear to be able to maintain levels of cognitive functioning comparable to monolinguals', even in the face of far more severe brain deterioration.

Possible mechanisms underlying bilingualisminduced reserve
One of the least studied aspects in this area is how the bilingualism-induced BR and CR interplay throughout the lifespan to provide neuroprotection in later life stages. Del Maschio et al.'s (2018) study is particularly useful to shed light on this point. The authors deployed a Flanker task to assess EC in young and senior bilinguals, compared to young and senior monolingual controls. They also performed voxelbased morphometry (VBM) analysis in EC network's areas and conducted a joint behavioral-neuroimaging analysis to investigate how BR and CR interact to affect behavioral EC performance during aging. Their initial finding, consistent with previous literature (e.g., Bialystok et al., 2004;Gold et al., 2013b;Bialystok et al., 2014) revealed a behavioral EC advantage for bilingual seniors, but not for young adults, over their monolingual counterparts (see Valian, 2015, for a review on the issue). Furthermore, as already discussed, bilinguals in both age groups showed greater GMV in multiple EC brain areas, namely the ACC, PFC, and IPL bilaterally, thus showing that bilingualism fosters the development of BR from early life stages. Nonetheless, the most interesting results came from the conjunct behavioral-VBM analysis. In a pattern consistent across all the aforementioned regions, in which bilinguals showed enhanced BR compared to monolinguals, bilingual seniors appeared not to rely on their neural EC substrate to optimize behavioral EC performance and outperform monolingual counterparts. In other words, senior bilinguals' performance was independent of the GMV decline and remained at optimal level even in the face of GMV reductions in the EC network areas, while senior monolinguals' performance dropped drastically with brain atrophy in such regions. The authors suggested that this result may reflect premorbid differences in cognitive strategies and brain networks recruited to perform EC tasks. Even if the scope of the VBM analysis did not allow an ''online'' investigation of the mechanisms underlying such phenomenon, the ensuing hypothesis was that lifelong bilingualism would help to develop BR by extensive training of EC network, which would in turn lead to enhancements in CR through increased network's efficiency and flexibility (i.e. NR) and thus to the development of different strategies to cope with agerelated brain deterioration (i.e. NC). Similarly, a recent longitudinal investigation with senior bilinguals and monolinguals reported that bilingualism mitigated the semantic memory loss linked with enthorinal cortex thinning (Arce Rentería et al., 2019).
Some attempts have also been made at modeling the mechanisms through which the neuroprotective effect could develop and act. For instance, Grant et al. (2014) suggested that the mechanisms behind the beneficial effect may be understood in terms of the posterior-to-anterior shift in aging (PASA) model (Davis et al., 2007;Dennis & Cabeza, 2008). PASA argues that, in the face of age-related decrease in neural activity in posterior brain areas, older adults would increasingly rely on frontal regions in order to maintain optimal levels of behavioral performance. In the case of bilingualism, such goal would instead be achieved through the preservation of posterior areas and their connections with the frontal cortex. Bilinguals, due to their less expressed posterior deterioration, would experience less of the typical PASA shifting. This account is based on the results of studies above (see ''Brain reserve in bilinguals'' section) reporting increased GMV in posterior regions such as the temporal pole and IPL (Abutalebi et al., 2014(Abutalebi et al., , 2015a and enhanced frontal to posterior connectivity in bilingual seniors (Luk et al., 2011).
In similar fashion, Grundy et al. (2017) developed a model named bilingual anterior to posterior and subcortical shift (BAPSS), arguing that expert bilinguals would rely less on the recruitment of frontal and executive regions and more on the recruitment of posterior/subcortical regions to perform EC tasks, compared to monolinguals. The model stems from five different types of findings emerging from bilingual research: (1) bilinguals show increased GMV in posterior and subcortical regions of the brain, compared to monolinguals (e.g. Abutalebi et al., 2013;Pliatsikas et al., 2014;Abutalebi et al., 2015a;Wei et al., 2015;Burgaleta et al., 2016;Pliatsikas et al., 2017); (2) bilinguals show greater white matter integrity than monolinguals (e.g. Coggins et al., 2004;Luk et al., 2011;Gold et al., 2013a;Felton et al., 2017); (3) bilinguals show less frontal activation than monolinguals with better or equivalent performance on EC tasks (e.g. Waldie et al., 2009;Rodríguez-Pujadas et al., 2014), and (4) bilinguals show stronger functional connectivity between brain regions during EC task performance. Thus, bilinguals may be able to better distribute the effort, without tapping excessively into their frontal resources (Luk et al., 2010;Grady et al., 2015;Li et al., 2015;Costumero et al., 2015). Finally, (5) EEG studies show that bilinguals rely on earlier processing stages than monolinguals to achieve comparable levels of EC performance (Fernandez et al., 2013(Fernandez et al., , 2014Moreno et al., 2014;Sullivan et al., 2014;Barac et al., 2016). Following these findings, Grundy and colleagues argue that senior bilinguals are better able to shift from more demanding, late, topdown processing, to more automatic early processing during EC performance due to their lifelong experience in controlling two languages simultaneously. This framework may provide an explanation of the mechanisms underlying bilingualism's neuroprotective effect in aging: a lesser reliance on frontal regions and a stronger one on subcortical/posterior regions would allow bilinguals to maintain optimal cognitive performance by experiencing less of the typical PASA shifting and maintaining more resources that can be deployed with increasing task requirements.
To conclude this section, it is worth to briefly discuss an explanation provided by Guzmán-Velez & Tranel (2015) that has, up to now, received relatively little attention. The noradrenergic theory of CR (Robertson, 2013) argues that factors known as enhancers of BR and CR would upregulate the noradrenergic system, in turn resulting in compensatory mechanisms (e.g., enhanced neurogenesis and synaptogenesis, increases in the production of brainderived neurotrophic factor (BDNF), increased GMV and connectivity) and disease-modification mechanisms (e.g., reduced amyloid burden, plaque size and aggregation, anti-inflammatory processes, and rescue of cholinergic and dopaminergic cells). Bilingualism, among other factors promoting BR and CR, would constitute no exception, contributing to develop a larger, more efficient and better-wired brain, capable of reorganizing itself to cope with age-related deterioration.

Future directions: resistance and resilience
A recently proposed account (Arenaza-Urquijo & Vemuri, 2018) attempts to reorganize the terminology in the field of cognitive aging by grouping the various sub-concepts of reserve and the concept of brain maintenance (see Nyberg et al., 2012) under a common umbrella. This account proposes a single general distinction between the concepts of resistance and resilience to dementia. The concept of resilience (approximately overlapping with that of reserve) refers to an individual's ability to sustain a betterthan-expected cognitive performance in relation to their degree of brain pathology. The concept of resistance (approximately overlapping with that of brain maintenance) refers to avoiding the appearance of brain pathology. Resistance is usually assessed by investigating the level of dementia biomarkers in an individual's brain such as beta amyloid or tau protein, while resilience is assessed by investigating levels of cognitive performance associated with brain pathology or levels of functional and structural neurodegeneration. As aforementioned, many studies seem to support the role of bilingualism in promoting resilience via better maintenance of gray and white matter structures, higher neural efficiency or flexibility, or better maintenance of cognitive ability in the face of neural deterioration. However, there is little research investigating the effects of bilingualism on resistance to age-related impairment. Nonetheless, one may argue that an indication of bilingualism's role in enhancing resistance may be the delay of AD onset observed in many studies as reported above. Moreover, a study directly investigating the effects of bilingualism on AD biomarkers (Estanga et al., 2017) reported a moderation effect of bilingualism between age and a cererebrospinal fluid (CSF) biomarker of AD, t-tau protein, in a cohort of 278 individuals matched for genetic, sociodemographic and cognitive profiles. Moreover, bilingualism in this study was associated with better scores on different cognitive tasks and lower prevalence of preclinical AD and it appeared to mitigate the relationship between age and worsening of EC performance. These findings suggest that bilingualism may promote both resilience and resistance to dementia and age-related cognitive decline, in general. We believe that such key neurobiological aspects of the relationship between bilingualism and cognitive aging deserve more attention in future research.

Conclusions-the socio-economic impact of bilingualism
Preservation of a healthy neurocognitive function is becoming more and more important in the steadily aging world that we live in. Taken together, the findings reviewed in this paper suggest that bilingualism may foster the development of brain and cognitive reserve during the lifespan, thus pointing to bilingual experience as an important personal, educational, and societal factor that may mitigate cognitive decline during the aging process. As a result, bilingualism may play an important role in improving the quality of life of older populations and reducing the burden on public health. Nevertheless, it is important to secure the practices that can ensure such effects in real life. A study using Cantonese-English bilinguals in Hong Kong showed that the neuroprotective effect against gray matter loss may be only present in individuals who keep using and practicing their L2 after retirement (Abutalebi et al., 2015a) suggesting that governments should develop policies aiming to help maintaining the second language use in senior citizens. With the difficulty to develop effective pharmacological therapies to mitigate the effects of cognitive decline, ''ecological'' solutions such as bilingualism and other factors inducing cognitive and brain reserve should be key in promoting healthy aging.