Abstract
Alzheimer’s disease (AD) is an age-related neurocognitive disorder that is epidemic in the elderly population. Currently, there are limited pharmacological interventions, and this has heightened the urgency to identify potential preventable or modifiable risk factors that promote resilience to the neuropathological effects of AD. The regular use of two or more languages is one such factor that may increases cognitive reserve through the long-standing executive control involved in managing multiple languages in the brain. There is also evidence that bilingualism is associated with increased brain reserve or maintenance, particularly in frontal-executive structures and networks. This review examines the current, sometimes conflicting literature on bi/multilingualism and AD. These studies have confounding variations in the assessment of age of second language onset, language proficiency, language usage, and whether determining incidence of AD or age of symptom onset. Despite these limitations, most publications support the presence of increased frontal-executive reserve that compensates for the development of AD neuropathology and, thereby, delays the emergence of clinical symptoms of dementia by about 4-5 years. Although regularly speaking more than one language does not protect against AD neuropathology, the delay in its clinical expression has a potentially significant impact on the lifelong morbidity from this age-related disease. Learning other languages may be an important modifiable factor for delaying the clinical expression of AD in later life.
INTRODUCTION
Dementia is an age-related disorder that is growing in importance in proportion to the increasing age of the population. Alzheimer’s disease (AD), the most common cause of dementia, affects about 10% of people 65 years of age and older [1]. In the U.S., approximately 6.5 million people are living with AD with many others having all causes of dementia, and the prevalence of AD may increase to about 14 million by 2050 as the U.S. population ages [1]. Despite the slow development of disease-modifying drugs such as lecanumab, these drugs have limited efficacy or lack broad applicability, and there is a need to identify modifiable environmental risk or protective factors that can delay the onset and emergence of this disorder [2].
The frequent use of two or more languages in everyday life appears to be one such protective factor [3, 4]. Reports from around the world indicate that bilingualism (or multilingualism) may be a modifiable factor that can delay the clinical expression of dementia by four to five years [5–10]. Studies even indicate that learning a second language (“L2”) later in life, and practicing it, can have beneficial effects for maintaining cognition [11–13]. This has implications for decreasing the impact of dementia as just a few years delay in the clinical emergence of AD means many years of quality of life and a decreased lifetime prevalence of this age-related disorder. Postponing the onset of clinical symptoms of AD can result not only in improved health in later life, but also in major health economic savings [14].
BILINGUALISM AND RESILIENCE TO AD
Brain resilience, and its reserve components, is a critical concept for understanding the effects of bilingualism on the emergence of AD clinical syndrome [15, 16]. Although the terms resilience and reserve have been used inconsistently, recent collaboratory work has made substantial progress in defining and operationalized these terms [17, 18]. Resilience is the capacity to maintain cognition and function in the face of aging and disease and is composed of cognitive reserve (CR), brain reserve (BR), and brain maintenance (BM). CR involves optimizing the efficiency, capacity, and flexibility of neural networks and the recruitment of alternate networks [19]; BR and BM refer to the underlying neural substrate, with the former reflecting the status at any point in time and the latter the minimizing of loss over time. Bilingualism contributes to all three overlapping types of reserve, resulting in active functional (CR) and structural (BR, BM) protection against the neuropathological effects of AD [20, 21].
BILINGUALISM AND COGNITIVE RESERVE
Bilingualism or multilingualism enhances CR against AD [22, 23], similar to the effects of education, occupational attainment, and engagement in stimulating activities in decreasing the impact of cognitive aging and neuropathology [15, 25]. CR means a relatively better cognitive level per underlying brain level; hence better CR means less functional impairment per age-related brain changes or neuropathology [26]. Consequently, better CR could delay the emergence of clinical dementia despite more advanced underlying disease. In other words, CR that promotes cognitive skills before the onset of AD lessens the loss of function from the onset of neuropathology [15]. Factors that increase CR work through more efficient network utilization and may all impact on a core CR network [19], such as a domain-general frontoparietal control network [15, 27].
Bi/multilingualism promotes CR from the need to manage more than one language, all of whom are constantly and unconsciously active, even when just planning to speak in one language [28, 29]. Investigators have proposed the Adaptive Control Hypothesis where successful language production depends on attending to the target language, inhibiting and avoiding interference from non-target languages, and orderly code-switching between them [21, 30–39]. CR emerges from this need to optimize the executive control mechanisms engaged in regulating simultaneously active and competing languages [5, 40–54]. Active, and not just passive, bi/multilingualism, with the regular use of languages in contexts that require linguistic switching [55, 56], trains the brain through increasingly efficient and alternative functional connectivity [20, 57].
Much data indicates that the need to manage more than one language relies on the same domain-general executive control mechanisms used for non-language tasks [33, 58–75]. Bilinguals, compared to monolinguals, may exhibit better general executive abilities [58, 76–80], attentional control and monitoring [36, 81], inhibitory control [82, 83], conflict management [62], switching ability [50, 85], and resistance to irrelevant, conflicting, or competing stimuli [58, 84]. This is evident on executive measures such as the flanker test [73, 87], the AX-continuous performance task [88], the Simon task [60–63], the Stroop test [59, 89], and also on digits backwards and the trail making test-B [45]. Bilingualism may also improve working memory [89], episodic memory and recall [90, 91], and cognitive screening and intelligence [92, 93]. Electroencephalographic studies further indicate more efficient attention and focused activation at rest among bilinguals compared to monolinguals [94], and evoked potentials suggest more monitoring, attentional, and other cognitive resources in bilinguals versus monolinguals [95, 96].
BILINGUALISM AND BRAIN RESERVE OR BRAIN MAINTENANCE
Without longitudinal data and baseline measures, it is difficult to distinguish BR, or the neurobiological physical status of the brain at any point in time, from BM, or the ongoing maintenance of brain structure; both reflect the underlying gray matter (GM) and white matter (WM) in the brain. Although all of a bi/multilingual’s languages recruit the same neuroanatomical language areas [97, 98], bilinguals show additional increases in volume of GM and integrity of WM [73]. These enhanced brain elements primarily occur in frontal (especially left inferior BA45 but also right) GM and their WM connections involved in language functions and in executive control [27, 99–141]. Specifically, bilinguals versus monolinguals have less age-related GM loss in dorsolateral prefrontal cortex and less extensive frontal activation or recruitment from more efficient neural compensation during executive control tasks, particularly if L2 is proficient or acquired early [11, 142–146]. Bi/multilingualism may also increase related regions such as the temporal pole (especially left and including hippocampus) [125, 148], the inferior parietal lobule (especially left) [77, 149], and subcortical regions [139, 150], including the caudate nuclei (especially left) involved in code-switching and controlling for verbal interference [101, 151–154], the putamen involved in L2 speech articulation [37, 156], and the cerebellum involved in producing fluent speech [157]. Finally, a recent review of 137 articles showed that increased L2 proficiency and its earlier age of acquisition was associated with increased GM and WM, particularly in the frontal lobes, anterior cingulate cortex, left inferior parietal lobule, and subcortical areas [158].
Bilingualism improves WM connectivity between the frontal and posterior areas [125, 135] and modulates the functional connectivity (FC) of the language network and the executive control network (ECN) [159]. Bilinguals versus monolinguals have increased WM integrity in the corpus callosum [12, 161], the superior longitudinal fasciculi [11, 162], the right inferior fronto-occipital fasciculus [135], and in parahippocampal cingulum and uncinate fasciculi [135, 163]. When bilinguals and monolinguals are matched for comparable levels of cognitive functioning, bilinguals tolerate lower WM integrity than monolinguals in the inferior longitudinal fasciculus, the inferior fronto-occipital fasciculus, the fornix, and parts of the corpus callosum [138]. Bilinguals appear to avoid the typical age-related tendency to shift task-related activation to more anterior regions with better preservation of anterior-to-posterior resting-state FC; stronger FC in language, ECN, and frontoparietal networks; and preserved FC in posterior brain regions and the default mode network, compared to age-matched monolinguals [123, 164–170]. In sum, these WM as well as GM changes can protect the brain against cognitive decline and neurodegenerative processes, helping delay the onset of the clinical symptoms of AD [171]. The cognitive exercise and training of bilingual language control continuously practiced over time appears to maintain these structures, particularly prefrontal ones involved not only in language but also in domain-general executive abilities such as switching [129, 173].
Learning a second language, no matter the age, is associated with long-term structural changes in connectivity [33, 174–176]. Learning L2 later in life is associated with wider representation in cerebral cortex and involvement of a larger number of cortical areas [177, 178]. A 4-month course of L2 training in older adults enhanced FC in the right inferior frontal gyrus (IFG), superior frontal gyrus, and left superior parietal lobule [12]. In another study, individuals intensively learning L2 over 3 months showed a significant increase in hippocampus volume and cortical thickness in the left middle frontal gyrus, IFG, and superior temporal gyrus [140]. In a third study, those who underwent 5 months of language learning showed structural changes indicating a close association between the left IFG and proficiency in L2 [179]. Altered resting state FC in language and control networks subsequent to L2 training programs is also associated with improved global cognition in older adults [180].
BILINGUALISM AND MILD COGNITIVE IMPAIRMENT
Studies show that bi/multilingualism delays the onset of mild cognitive impairment (MCI) [56, 181–183]. An early report that included MCI patients as well as AD patients, found a later age of onset, and subsequent reports have corroborated this finding [181, 183]. In a report from India of 115 persons with MCI, bilinguals had a clinical onset of cognitive complaints 7.4 years later than monolinguals (65.2 versus 58.1 years; p = 0.004), a finding not explained by years of education [182], and in a more recent report from Canada of 158 consecutive MCI patients, bilinguals received a diagnosis of MCI at a later age than monolinguals (77.8 and 75.5 years, respectively) [183]. In a prospective study of 964 seniors followed over about a 6-year period, the incidence of MCI in later life was significantly lower in those who had foreign language education during adolescence [184]. A recent systematic review reported a 3.2-year delay between bilinguals and monolinguals in age of diagnosis of MCI, although this difference did not reach statistical significance [4]. The decreased risk of MCI and early dementia depends on being an active, not a passive, bilingual [56, 163], may increase with number of languages (>2) spoken [185], applies to even closely related languages [163], and may be specific to single-domain amnestic MCI [186].
Several investigations indicate that bi/multiling-uals with MCI differ in underlying brain structure when compared to monolinguals with MCI. One investigation of structural changes at the time of MCI diagnosis found significantly more GM loss primarily in the frontal gyri, the right cingulate gyrus, and the precuneus, with greater volume loss in the right hemisphere [187]. Others have also found greater neuropathology in initial stages of MCI among bilinguals compared to monolinguals [181, 186], and one prospective MCI study showed that bilinguals, compared to monolinguals, had less parenchymal volume (more atrophy), especially in areas related to atrophy in dementia [55]. In contrast, one investigation of multilinguals with MCI showed a higher temporal tissue density in the medial temporal lobes and in areas related to language and cognitive control, before declining in temporal lobe density with the development of AD, suggesting that very early on there is increased GM compensating for episodic memory impairment before the development of greater neuropathology in bi/multilinguals [188]. Active bilinguals with MCI exhibit lower WM integrity in the fornix but higher integrity in the parahippocampal cingulum and uncinate fasciculus than is the case for passive bilinguals [163]. Finally, there was reduced glucose uptake in frontotemporal, parietal and cerebellar regions of bilinguals diagnosed with MCI and probable AD, relative to a sample of monolinguals matched for age, gender ratio, and disease severity [189].
BILINGUALISM AND ALZHEIMER’S DISEASE
Retrospective studies and systematic reviews report that bilinguals, compared to monolinguals, have a delay in the onset of the clinical symptoms of dementia despite the neuropathology of AD. In 2007, Bialystok and colleagues initially reported that bilingual patients with probable AD were diagnosed about 4 years later than monolinguals with probable AD [5]. Since then, predominantly retrospective studies from around the world suggest that bilingualism delays clinical symptoms of dementia by 4-5 years [5, 190–197] (See Table 1). A recent systematic review with metanalysis of 16 cross-sectional and 5 longitudinal studies concluded that bilingualism delayed the onset of AD symptoms by 4.7 years (CI 3.3, 6.1; t 8.06, p < 0.001), but did not find a lower risk of developing AD [4]. Another meta-analyses of 21 studies confirmed an overall effect of bilingualism in delaying AD symptom onset and weaker evidence that bilingualism prevents the occurrence of the disease itself [198]. A third meta-analysis of 8 studies found that bilinguals with AD exhibit symptoms 4.05 years later (95% CI: 1.87–6.22) and were diagnosed 2.0 years later (95% CI: 0.08–3.92) than monolingual participants [199]. A fourth systematic review and meta-analysis of 34 studies found “mixed results” with some studies showing a lifelong bilingual advantage in symptom delay but not others [200]. In a different approach, an investigation of the relationship between the degree of multilingualism and dementia incidence in 93 countries revealed that speaking more than one language induced a protective effect against the development of dementia [194]. This delay has occurred regardless of the dominant or first language (“L1”) [197]. Pervasive multilingualism could have been a factor in the low prevalence of cognitive complaints and dementia among older adults in Luxembourg [52], and, in the Lothian Birth Cohort (Scotland), those who learned a second language achieved better cognitive results in late life [93]. Some studies, however, indicate a delay in AD only among low-educated (<11 years) bilinguals [195]; whereas, others found a bilingualism-related delay of AD only in immigrant, but not in non-immigrant bilinguals [196, 201].
Studies supporting delay in clinical symptoms of Alzheimer’s Disease (AD) or dementia
In contrast, several prospective studies have failed to find a bilingual advantage in the onset of dementia symptoms [201–208] (See Table 2). These studies can be much more difficult to accomplish than retrospective studies and may be hampered by inconsistencies in reported L2 proficiency and language usage, and possible factors such as dense code switching (mixed language, e.g., “Spanglish”). In particular, the discrepancy between mostly positive retrospective studies and many negative prospective studies may relate, in part, to the conflating of incidence rates of dementia and the age of symptom onset [198]. Another variable is the number of languages used. For example, in a Longitudinal Analysis of the Nun Study (325 religious sisters who were older than 75 years of age) [208], self-reported multilingualism did not delay the onset of dementia; however, participants speaking four or more languages (but not two or three) were significantly less likely to develop dementia than monolinguals [209].
Studies less supportive of a delay in clinical symptoms of Alzheimer’s Disease (AD) or dementia (with reservations noted)
Bilinguals versus monolinguals have more advanced AD neuropathology at comparable stage or severity of memory impairment and dementia, consistent with a delay in dementia symptom onset, but not a delay in the neuropathology of AD. The CR effect of bilingualism acts through neural compensation and efficiency, increased frontal and other areas of the brain, and the modulation of brain network connectivity [22, 144]. First, the greater CR is apportioned to general cognition and memory at the expense of language control [11, 188], thus optimizing compensatory use of brain resources in the presence of more pathology in AD [49]. Second, bilinguals with dementia may have more advanced temporal neuropathology in mesial areas typically associated with AD, than monolinguals with similar severity of dementia [10, 211]. In comparison, multilinguals with early-stage AD may have a thicker cortex than that of monolinguals in the frontal and related areas [188]. In a reverse study, where patients were matched on WM measures, the bilinguals were less impaired on cognitive measures (using propensity scores) than the monolinguals [212]. A compensation effect on brain neurodegeneration is reflected in increased connectivity in the ECN and related language network and in the default mode network in bilinguals with probable AD [126, 213]. In two fluorodeoxy-glucose positron emission tomography (FDG-PET) studies, [26, 144], bilingual patients with AD had more severe glucose hypometabolism compared to monolingual patients with AD, and this hypometabolism correlated with degree or extent of bilingualism. Finally, despite greater neuropathology with comparable cognitive impairment, one study found lower levels of t-tau in their cerebrospinal fluid, a biomarker of AD neuropathology, among bilinguals versus monolinguals [45].
At some point, the neuropathology overwhelms the CR. Investigators report a faster MCI-AD conversion rate in bilinguals with average conversion times from MCI to dementia of 1.8 years for bilinguals compared to 2.8 years for monolinguals, followed by a more rapid decline among bilinguals as they develop AD [183]. In those who are cognitively impaired or have a diagnosis of AD, higher CR is associated with an accelerated decline and often disproportionate language problems [214–216]. In patients with AD, naming with the dominant language was associated with cortical thickness of the entorhinal cortex and middle temporal gyrus, but naming with the nondominant language was correlated with thickness of the left caudal anterior cingulate cortex, a central cognitive control region involved in error monitoring and task switching [217]. Older bilinguals may struggle inhibiting cross-language intrusions particularly from L1 when speaking L2 and maintaining fluency in L2 [100, 218]. As dementia progresses to affect frontal language control, bilinguals may regress to predominant or exclusive use of the dominant language (with L1 intrusions during L2) [144, 219–222]. This “regression to L1” may indicate imminent cognitive impairment or incipient dementia in bilinguals [223] and suggests that dementia symptoms emerge among bilinguals once frontal lobes and their tracts are affected and their “executive compensation” for dementia symptoms begins to fail [197].
RELATED NEUROLOGICAL CONDITIONS
There is support for the CR of bi/multilingualism from other neurological conditions [148]. For example, bi/multilingualism may protect from cognitive decline post-stroke and delay the onset of post-stroke dementia [224, 225]. Bi/multilingualism may also delay the onset of other dementias, such as behavioral variant frontotemporal dementia [226]. In Huntington’s disease, a higher frequency in using bilingualism was associated with better performance in inhibitory control and set-shifting and higher GM volume in the inferior frontal gyrus [227]. Finally, mixed results from studies of primary progressive aphasia (PPA) appear to reflect differences in underlying neuropathology. A study involving the logopenic variant of PPA, which is most associated with AD pathology, showed a 5-year delay in age of symptom onset (not observed in amnestic AD in this study) [228]. In contrast, others have not found a delay in nonfluent or semantic variants of PPA, which are associated with frontotemporal lobar degeneration (FTLD) [226]. Among a mixed group of PPAs, which may have mostly FTLD pathology, there were initial word-finding difficulties in L2 [229], and, in a report of 16 semantic variant patients, there was a tendency to lose L2 or the less proficient language [230].
LIMITATIONS
The research on bilingualism and dementia has been controversial primarily due to methodological differences and inconsistencies in the characterization of bilingualism and related variables across studies [47, 231–240]. The effects of bi/multilingualism vary greatly with age of acquisition, L2 proficiency (i.e., the level of mastery attained in L2), and language usage or exposure (the amount of daily use of both languages) [159, 241]. The main sources of discrepancy between studies may be interindividual variability in L2 acquisition [242], followed by variation in the objective determination of L2 proficiency without relying on self-reports [232, 233]. For example, a major prospective study that failed to find a bilingual difference did report that speaking L2 “very well” had a 14% lower risk of developing dementia than those who reported “not at all” well [202]. In addition, studies may fail to assess whether delays in clinical dementia onset are correlated with the use of bilingualism in dual-language contexts where languages are kept separate, rather than in dense code-switching where languages are mixed [243]. Another important factor that affects language usage is the context or the social situation, such as home or familial versus formal or work environment [30, 238–240]. Immigration status, comparability of education, degree of acculturation, and language of test administration are factors that can impact the assessment of bi/multilingualism. There are clear individual differences in language background and experiences that argue for a more inclusive assessment of the “bilingual experience” [20]. A further source of inconsistency among prospective studies is the focus on proportion of cohort affected with AD rather than on the delays in symptom onset or manifestations; it is not the lifelong occurrence of AD that is primarily affected by bilingualism, but, rather, the age of emergence of clinical symptoms. Bilingualism in dementia research often does not analyze the relatedness of languages and the number of shared cognates between them. Inhibition of the non-target language may be more effortful for close languages with the same or similar words, or cognates. Studies of bilingualism in AD must control for, or at least consider, these multiple variables and other complexities of the language experience. Finally, when invoking resilience and reserve, which cannot be measured directly, studies of the effects of bilingualism need to apply their precise operational definitions in research that captures individual variability, not only of the bilingual experience, but also of other lifelong variables that affect cognition across the lifespan [17, 20].
CONCLUSIONS
The aggregate of the literature on bi/multilingualism is that it increases frontal-executive CR, thereby delaying the emergence of clinical symptoms of dementia in the presence of the progressive neuropathology of AD. Early on, as bilinguals develop the initial brain changes of AD, they can successfully divert frontal-executive resources towards compensation for early memory and other cognitive impairments [47, 222]. The benefit of this delay is that it has a potentially significant impact on lifelong morbidity with this age-related disease, as more individuals die without experiencing as many years with clinical dementia or even its expression. Learning a second language, even in older age, may be an important mechanism for promoting CR and for delaying a functional decline from early dementia in later life through these compensatory executive mechanisms [244, 245]. Future research can evaluate the many individual factors involved in the bilingual experience and how they may contribute to resilience and the specific types of reserve [20]. Bi/multilingualism in AD deserves more attention and dedicated research as a modifiable preventative factor for the clinical symptom onset of AD.
Footnotes
ACKNOWLEDGMENTS
The author has no acknowledgments to report.
FUNDING
The author has no funding to report.
CONFLICT OF INTEREST
The author has no conflict of interest to report. The author is an Editorial Board Member of this journal but was not involved in the peer-review process nor had access to any information regarding its peer-review.
