Corrected: Author Correction Corrected: Publisher Correction PERSpECTIvES

when there is no direct association between OPINION activity levels and performance in older adults15,16. By contrast, other studies Maintenance, reserve and have used the term compensation more restrictively to describe situations in which age-related increases in brain activity are compensation: the cognitive directly correlated with better performance in older adults17. Moreover, it has also been neuroscience of healthy ageing unclear in the literature how the concepts of reserve, compensation and maintenance Roberto Cabeza, Marilyn Albert, Sylvie Belleville, Fergus I. M. Craik, relate to one another. Audrey Duarte, Cheryl L. Grady, Ulman Lindenberger, Lars Nyberg, To address this terminological confusion, Denise C. Park, Patricia A. Reuter-Lorenz, Michael D. Rugg, Jason Steffener the authors of this Opinion article met in 2017 and worked to sharpen the definitions and M. Natasha Rajah of these popular terms. Some differences Abstract | Cognitive ageing research examines the cognitive abilities that are in opinion about the definitions persist; preserved and/or those that decline with advanced age. There is great individual however, in this article, we emphasize the points of agreement. The terms variability in cognitive ageing trajectories. Some older adults show little decline in maintenance, reserve and compensation cognitive ability compared with young adults and are thus termed ‘optimally can of course be applied to aspects of ageing’. By contrast, others exhibit substantial cognitive decline and may develop ageing beyond the brain and cognition dementia. Human neuroimaging research has led to a number of important (such as bone changes). However, here, advances in our understanding of the neural mechanisms underlying these two we focus on their use in structural and functional neuroimaging studies in healthy outcomes. However, interpreting the age-related changes and differences in brain ageing humans (defined here as ageing in structure, activation and functional connectivity that this research reveals is an individuals who are apparently free of brain ongoing challenge. Ambiguous terminology is a major source of difficulty in this disease), although other related terms and venture. Three terms in particular — compensation, maintenance and reserve — methods are also discussed. Given this have been used in a number of different ways, and researchers continue to disagree focus, the use of the terms maintenance, about the kinds of evidence or patterns of results that are required to interpret reserve and compensation in this article should be considered to refer specifically to findings related to these concepts. As such inconsistencies can impede progress neurocognitive maintenance, neurocognitive in both theoretical and empirical research, here, we aim to clarify and propose reserve and neurocognitive compensation, consensual definitions of these terms. respectively. We do not discuss in detail how the mechanisms of reserve, maintenance In all parts of the world, the proportion of decline, as well as those of optimal ageing. and compensation interact with pathological older adults in our population is rapidly These mechanisms — that is, the putative processes (but see Box 1), but it is worth expanding1. Advances in medicine and causal explanations of age-related changes noting that this is an important question public health measures, rising standards — presumably exist at multiple levels of and that these three mechanisms may also of living, and improvements in education analysis (including the genetic, cellular attenuate pathological processes18. and nutrition have lengthened the human and systems levels). As the field has grown, lifespan. Cohort comparisons suggest several specific terms, including reserve4–6, Cognitive neuroscience of ageing that the debilitating effects of senescence maintenance7 and compensation8–13, have Ageing affects neurobiological functions are increasingly delayed to later ages2. been introduced. These terms have been at multiple levels19. It can alter genes and Nevertheless, advancing adult age continues used both to describe the qualitative and gene expression20–23, interfere with the to be associated with cognitive decline in quantitative differences in brain structure functions of cells and molecules24–33 and many individuals, and major challenges and function that occur with age and to lead to changes in the overall structure remain in our efforts to understand the advance current theories of the mechanisms and function of the brain34,35. In recent mechanisms of cognitive loss versus those of brain ageing and cognitive decline. decades, the advent and availability of MRI of optimal ageing (defined as the situation However, over the years, these terms have methods have significantly advanced our in which cognitive abilities are preserved been used inconsistently, creating confusion understanding of how the brain changes throughout ageing). and slowing progress14. For example, with age at the gross anatomical and Research in the cognitive neuroscience age-related increases in brain activity functional levels36–38. For example, healthy of ageing3 seeks to understand the neural have been interpreted as compensation ageing is known to be associated with grey mechanisms of age-related cognitive for declines elsewhere in the brain, even matter volume reductions and functional

Nature Reviews | Neuroscience Perspectives

Box 1 | Maintenance, reserve and compensation in Alzheimer disease and mild cognitive impairment

the trajectory of alzheimer disease (aD), which progresses from normal β-amyloid levels have shown both greater activity in the superior and cognitive performance to mild cognitive impairment (MCi) and then to lateral parietal cortex and the occipital cortex and better memory than full-blown dementia94 is, by definition, an example of poor brain older adults with low β-amyloid levels105. maintenance. However, the trajectory from healthy ageing to aD is in some cases, however, hyperactivation may reflect the underlying modulated by reserve and compensation6,95,96. studies have reported that neuropathology105 and excitotoxicity rather than compensation. it has been β-amyloid deposition — a putative biomarker of aD — is lower in older suggested that poor clearance of β-amyloid and tau proteins in the brain adults with higher scores on reserve proxies, such as education5, and in contributes to the accumulation of amyloid plaques and neurofibrillary those who participate in cognitively stimulating activities across the tangles, respectively, and that this increases the production of glutamate lifespan97. even in individuals in whom biomarkers of aD are present, and inhibits its recapture106,107, leading to hyperexcitation. Consistent with higher scores on reserve proxies are associated with a lower risk of this hypothesis, a study found that low doses of an antiepileptic drug progression from normal cognition to the onset of clinical symptoms98. reduced hippocampal hyperactivity and improved memory in individuals the neural bases of these protective effects remain to be identified. with MCi90. an intriguing possibility is that hyperactivation is, at first, Functional Mri studies have shown that individuals with MCi99–101 and compensatory but later reflects excitotoxicity100. Compensatory processes carriers of the apolipoprotein e (APOE) ε4 allele, a known risk factor for thus might characterize the early course of aD and contribute to its long late-onset aD102, show increased task-related activity in the brain regions prodrome. if this is the case, compensatory non-pharmacological first affected by aD: the hippocampus, cingulate and precuneus103,104. interventions could be used to reduce cognitive symptoms. in turn, some studies have associated greater activity in these regions with better hyperactivation has the potential to contribute to an early signature of cognition in individuals with MCi99,101, consistent with compensation. aD, and approaches might be developed to reduce hippocampal Furthermore, some cognitively normal older individuals with high hyperactivity or to promote reliance on unimpaired brain networks108. alterations in several regions that are crucial age-related differences in cognitive ability; or the operation of a brain network, as for higher cognitive function: that is, the this is important because it has been shown assessed by functional connectivity methods. prefrontal, medial temporal and parietal that a large proportion of the variance in In the case of compensation (Fig. 1c), the cortices3,39–41. Similarly, diffusion MRI cognitive performance observed among heightened cognitive demands that arise methods have shown age-related changes older adults already existed when they owing to the combined effects of task in white matter connectivity between were children51. Although they have their difficulty and age-related cognitive decline prefrontal and posterior cortical regions own limitations52, longitudinal study are counteracted by recruiting additional and within posterior sensory cortices42–44. designs avoid these problems. It is therefore neural resources, including those established These age-related declines in brain structure important to note that longitudinal studies by neural enhancement. It is important to and function are associated with cognitive have consistently demonstrated large emphasize that the mechanisms of reserve, decline in a variety of domains, including individual differences in rates of age-related maintenance and compensation are not episodic memory, working memory and cognitive decline53,54. static but dynamic and modifiable and that attention45,46. The individual differences in age-related it is likely that they are not only responsible One of the most fundamental and cognitive decline described above for modifying behaviour but are, in turn, urgent goals of research in the cognitive undoubtedly reflect a complex interaction modified by changes in behaviour. neuroscience of ageing is to understand between genetic and environmental Below, we propose one way to define and why some individuals experience faster factors. The outcomes of these factors have link the terms reserve, maintenance and cognitive decline than others during healthy been proposed to be partly mediated by compensation to provide greater coherence ageing47. Inter-individual variability in three interacting mechanisms: reserve, in an evolving field. Nevertheless, we cognitive ageing is striking. Indeed, in maintenance and compensation (Fig. 1a). recognize that these concepts are the subject cross-sectional studies, some 80-year-old In the case of reserve6 and maintenance7 of continuing debate. Our goal is therefore individuals can perform as well as, or (Fig. 1b), it has been proposed that the to initiate an open dialogue about what better than, some 40-year-old individuals processes of age-related neural decline these three concepts mean. For the sake of on cognitive tasks that assess functions (which include brain atrophy, synaptic loss simplicity, we use examples involving single often impaired by ageing (such as episodic and white matter degradation) are countered brain regions to illustrate our arguments; memory)48. However, when investigating by processes of neural enhancement: that however, each of the concepts is equally individual differences among older adults, is, the creation, replenishment and repair applicable to measures that account for it is important to consider the limitations of neural resources. Specifically, it has covariance between multiple brain regions, of such study designs when compared with been hypothesized that reserve supports including functional connectivity and longitudinal designs. For example, older the creation of neural resources that help multivariate activation patterns55. participants are typically recruited only from withstand the neural decline and that the subset of well-educated people who have maintenance supports the replenishment Reserve aged in relatively good health and are free and repair of these neural resources. Neural We believe that reserve should be defined as of brain disease, whereas the young adult resources here refers to the brain anatomy a cumulative improvement, due to genetic samples with which they are compared are and physiology that mediate cognitive and/or environmental factors, of neural more heterogeneous. Cross-sectional study processes. In its simplest form, a neural resources that mitigates the effects of neural designs can also be contaminated by birth resource could be the grey matter volume of decline caused by ageing or age-related cohort effects, including inter-generational a brain region or the white matter quality diseases. Reserve is hypothesized to result IQ increases (known as the ‘Flynn effect’)49,50. of a fibre tract. However, neural resources in the accumulation of neural resources before Moreover, cross-sectional designs cannot could also refer to the function of a region, the brain is affected by age-related processes distinguish between age-invariant and as measured by functional neuroimaging, and to take place over a period of years.

www.nature.com/nrn Perspectives

A good example of a factor that promotes a reserve is education, which improves Genes Reserve neural resources during childhood and Individual differences young adulthood (possibly by enhancing Maintenance in cognitive ageing synaptic density56) and attenuates age-related (rate of decline) 57,58 Compensation cognitive decline in later adulthood . The Environment beneficial effect of education on cognitive performance might also be mediated partly by its effects on a range of other outcomes Ideal Typical b that have also served as proxy measures of reserve, including health, stress, profession

and lifestyle. In the ideal case, accumulated ces and reserve is enough to completely offset e demands age-related neural decline; however, in esour a more typical case, it only attenuates al r Reserve Reserve cognitiv this decline. Presumably, most reserve Neur accumulates during childhood and in young Childhood Old age Childhood Old age adulthood; however, it may also continue to build up in older age9, which arguably Seconds or hours underscores the importance of intellectual engagement throughout the lifespan. ces and e demands Some authors have used the more esour

al r Maintenance specific terms ‘brain reserve’ to refer to Maintenance aspects of reserve that are easily quantified cognitiv Neur in anatomical brain images and ‘cognitive Childhood Old age Childhood Old age reserve’ to refer to aspects that are difficult to detect in anatomical images and either c require functional imaging measures or cannot be delineated at the neural level given current technology6,59. Given that ces and

cognition depends on the brain, we believe sour e demands that this distinction is somewhat artificial al re Compensation Compensation and prefer to use only the term ‘reserve’. cognitiv However, we recognize that different aspects Neur Seconds or minutes Seconds or minutes of reserve require different technologies for their measurement and that some cannot Neural resources Cognitive demands be assessed with the current technology Increase in neural resources Increase in cognitive demands Decrease in neural resources Decrease in cognitive demands (but could be measurable in the future). Genetic60 and environmental6,61 factors, including longer education5, greater physical Fig. 1 | similarities and differences between reserve, maintenance and compensation. activity62, active participation in demanding a | Individual differences in cognitive ageing have been attributed to the effects of three interacting leisure activities63 and bilingualism64,65 affect mechanisms: reserve, maintenance and compensation. As illustrated in the schematic, these mech- individual differences in reserve. Because it anisms are assumed to mediate some (but not all) of the effects of interacting genetic and environ- is not possible to measure reserve directly, mental factors on cognitive ageing. b | Schematic charts illustrate the hypothesized changes in most studies of reserve have focused on a neural resources and cognitive demand that occur across the lifespan as a result of reserve and maintenance mechanisms. In the ideal scenario shown on the left, these mechanisms completely particular proxy of reserve and investigated counteract the effects of ageing, with resources meeting or exceeding demands throughout life. how individuals with high or low levels However, in the typical scenario, they only attenuate the effects of ageing. Reserve and maintenance of this proxy measure differ in their are both hypothesized to involve an increase in neural resources; however, they differ in terms of brain structure or function. For example, whether this increase occurs before or after the effects of ageing on neural function and the times- functional neuroimaging studies have cale of the changes. In the case of reserve, neural resources accumulate beyond what is required to compared the brain activity of individuals satisfy current cognitive demands, such that when these resources start to decline in old age, cog- who exhibit high or low scores in IQ, nitive decline is attenuated. It is important to note that although the graph shows resources accu- education level or occupational attainment, mulating during childhood and young adulthood, cognitive reserve can continue to accumulate in which are all considered proxy measures old age. In the case of maintenance, processes of neural decline are continuously offset by processes of reserve66. One such study found that of neural enhancement. Given that neural decline increases in old age, greater maintenance is also required to maintain the same level of performance. The figure shows neural decline and neural greater cognitive reserve, as measured using enhancement processes in alternation for illustration purposes only , as these processes can occur IQ and education–occupation as proxies, simultaneously. c | Schematic charts illustrate the hypothesized changes in neural resources and was associated with lower brain activity in cognitive demand that occur during short-term increases in cognitive demands as a result of com- a variety of brain regions (including the pensation mechanisms. In the ideal scenario, a task-related increase in cognitive demands is superior temporal and superior parietal ­completely counteracted by the recruitment of additional neural resources whereas, in the typical cortices) during cognitive processing, scenario, the additional resource recruitment reduces but does not eliminate the gap between task suggesting that reserve is linked to more demands and available resources.

Nature Reviews | Neuroscience Perspectives

effective use of cerebral networks4. It remains capacity (the total amount of neural resources beyond their current level, debatable what type of variable would resources available for cognition) and/or whereas maintenance is about returning serve as a good proxy measure of reserve. neural efficiency (the use of less neural them to their former higher level. We However, it is worth noting that it is critical resources — often operationalized as acknowledge that reserve has an impact on to specify the proxy factors and mechanisms neural activity — to perform a cognitive later maintenance because the accumulation that are assumed to build and constitute task)6,61. One example of how increased of reserve must be maintained. What may reserve a priori when developing one’s capacity is associated with a proxy measure most differentiate reserve and maintenance experiment. If reserve is defined merely of reserve is the aforementioned effect of are the mechanisms by which these as the factor that individuals with greater education on synaptic density56. An example factors influence healthy brain ageing and reserve have and then this factor is used to of an increase in neural efficiency is the cognition. In the case of reserve, these explain why some individuals have greater development of expertise in a particular factors cumulatively influence neural reserve, the argument is clearly circular14. domain through training, which in turn capacity and neural efficiency (and other When using functional neuroimaging is often associated with reduced regional mechanisms not yet identified owing to investigate reserve, it is important to brain activity70–73. The development of to limits of technology). In the case of distinguish between the across-individual expertise is associated with the presence maintenance, these factors influence activity differences that are related to reserve of richer and more differentiated conceptual neural mechanisms of repair and plasticity and those that are related to compensation representations, which can attenuate (and others not yet identified). Therefore, (see below). Differences related to reserve age-related decline in the domain of although the concepts of reserve and might be expected to manifest as trait-like expertise6,74–76. This idea may explain why maintenance are similar, we view them effects: that is, they would be evident older individuals can remain highly effective as complementary perspectives on how across a range of different cognitive in their specific professional domain77. environmental and biological factors domains and would correlate with multiple When adults with high levels of reserve, influence brain ageing and cognition. independent proxies of reserve, such as IQ as indicated by one or more reserve proxies, In principle, it is possible to distinguish and educational level. Individual differences do eventually display cognitive decline, different forms of maintenance and for reflecting compensation, by contrast, they do so at a rapid rate78. It is possible these to be operational to a different extent would be expected to differ according to that at some level, the burden of age-related in different individuals. For example, the nature of the cognitive challenge and neuronal decline becomes great enough maintenance can relate to different aspects to correlate with individual differences in to overcome the protective mechanisms of the brain, such as the grey matter or white task performance to a greater extent than of reserve, resulting in rapid cognitive matter, to different neurotransmitter systems with proxy measures of reserve. However, decline6,59. or to different brain regions, such as the complicating the distinction, reserve and hippocampus or the prefrontal cortex (PFC). the capacity for compensation may interact. Maintenance For example, given the link between exercise For example, highly educated individuals We propose that the term maintenance be and white matter integrity79, it is possible may show different activation patterns used to refer to the preservation of neural that individuals who exercise regularly than individuals with lower educational resources, which entails ongoing repair maintain white matter integrity better than attainment because their greater reserve and replenishment of the brain in response they maintain other aspects of the brain. allows them to deploy more effective to damage incurred at the cellular and However, because there is a close interaction compensatory processes. molecular levels owing to ‘wear and tear’7. between different brain structures and One analytical approach to the Maintenance occurs throughout the lifespan processes, it is also possible to talk about measurement of the neural correlates of but may become more critical in old age, as general brain maintenance. Maintenance reserve (and to other concepts that cannot neural deterioration becomes more severe. is often defined as a relative lack of decline be directly measured owing to current The timescale of maintenance processes is in one or more neural measures and technological limitations) is to regress likely to depend on the neural level at which not in absolute terms. Therefore, so-called out (control for) the effects of cognitive they take place (molecules, cells or systems). ‘brain maintainers’ include individuals performance on neural variables known to In the optimal case, repair processes who start with above-average levels of affect cognitive decline (including volume fully counteract decline. In the typical a neural measure in early adulthood and white matter hyperintensities) and then scenario, however, repair processes do not (such as a larger hippocampus) and to examine the correlation of the residual completely offset neural deterioration, maintain these high levels into older age, neural measures with a hypothesized leading to a gradual process of age-related as well as individuals who start below proxy of reserve (or other concept of neural deterioration. Some individuals average and maintain functioning at that interest)67–69. Using this approach, one study may be relatively spared from detrimental lower level. It is also theoretically possible decomposed variance in episodic memory brain changes in the first place, resulting that maintenance mechanisms differ for performance into a component predicted in a likelihood of displaying high levels of those who start with high levels of a neural by demographics, a component predicted by maintenance regardless of the capacity for measure and those who start with low levels pathology (as measured by structural repair. Thus, the efficacy of maintenance of the same measure. MRI) and a residual reserve component, depends both on the magnitude of decline The notion of maintenance is consistent which was then shown to moderate and the efficacy of repair. with evidence that adults who display stable cognitive decline69. The concepts of reserve and maintenance cognitive performance as they age tend to In addition, it is likely that different are clearly related to each other but, here, show minimal brain decline or pathology7. proxy measures of reserve may engage we highlight what distinguishes them: For example, a longitudinal structural different neural mechanisms and reflect although both involve enhancing current MRI study80 found that individuals aged different aspects of reserve, such as neural resources, reserve is about augmenting 65 years or older who exhibited little or

www.nature.com/nrn Perspectives

no episodic memory decline over a 4-year different brain regions using neuroimaging, maintenance and not reserve, one would period showed less hippocampal atrophy it is advisable to propose beforehand specific need to match performance groups by during the same period than individuals hypotheses about the relationship between proxy measures of reserve or statistically with substantial memory decline (Fig. 2a) specific brain measures (such as volume) control for proxy measures of reserve (see also ref.81). Similarly, in a longitudinal from particular brain regions (such as before testing for maintenance effects. fMRI study82, hippocampal activity during the hippocampus) and specific cognitive Nevertheless, when considering an episodic memory task was significantly measures (such as episodic free recall). At cross-sectional studies, it is important to higher in older adults whose memory the same time, the high degree of covariance remember that some differences between function was stable over the previous two between changes in different cognitive young and older adults could be due to decades (‘maintainers’) than in older adults abilities suggests that maintenance in one cohort effects, such as early life influences, whose memory abilities had declined over functional domain is likely to be related to and not ageing per se85. this period (‘decliners’) (Fig. 2b). Importantly, maintenance in another domain54. In summary, we consider maintenance higher hippocampal activity in maintainers As demonstrated by the examples to be a dynamic process that engages neural compared with decliners was observed after described above, maintenance mechanisms mechanisms of cellular repair and may matching both groups on initial memory are ideally investigated using longitudinal overlap to a large degree with mechanisms levels. However, because hippocampal data, as only within-person assessments of brain plasticity in adulthood86. Similar activity was not measured at initial testing can truly quantify change (or, in the case of to the concept of reserve, the mechanisms (20 years prior), it is unclear whether the maintenance, lack thereof)84. However, this contributing to maintenance are also likely maintainers exhibited higher hippocampal does not imply that cross-sectional brain to have both genetic87 and environmental activity than decliners at the start of data cannot be informative. For example, origins, with the latter including factors such the experiment. Whereas hippocampal successful maintenance could explain as diet, exercise and cognitive and social maintenance has been associated with cross-sectional findings that the brains of engagement7,62,88. Behavioural genetic studies episodic memory maintenance, the high-performing older adults look similar suggest that genetic and environmental maintenance of other brain regions is likely in anatomy and physiology to young brains, contributions to maintenance become to be associated with the preservation of whereas the brains of low-performing older increasingly correlated with advancing other cognitive abilities7. For example, adults look different from young brains7. age87. The specific mechanisms remain to PFC maintenance could be associated with However, as reserve may also contribute to be determined but are likely to include both the maintenance of cognitive control83. To differences in neural resources available neural components (such as neurogenesis) reduce the number of multiple comparisons to distinct performance groups, to argue and non-neural components (such as when investigating the maintenance of that this difference is indeed related to vascular changes). abOld maintainers Compensation 0.10 Old decliners We propose that the term compensation Young adults should be used to refer to the cognition-enhancing recruitment of 0.08 neural resources in response to relatively high cognitive demand. Compensation olume change

olume at T1) is temporally linked to variations in 0.00 cognitive demands and can occur rapidly, in a matter of seconds. As explained

ity in left hippocampus 0.04 below, we reserve the term compensation (% BOLD signal change )

olume at T2/v for neural recruitment that enhances (v ear hippocampal v cognitive performance. In the ideal case, fMRI activ

4-Y the cognition-enhancing recruitment is –0.10 sufficient to meet the task demands, whereas 0.00 in the typical scenario, it is insufficient Maintainers Decliners to match the demands. Our definition of Fig. 2 | stable cognitive performance is associated with brain maintenance. a | Graph showing that compensation is not limited to healthy individuals aged 65–80 years old who showed minimal episodic memory decline on verbal immediate and pathological ageing; it also applies to free recall and delayed cued recall tasks over a 4-year period (referred to as maintainers) also showed the cognition-enhancing recruitment of less hippocampal volume decline over the same period than decliners (graph created using data from resources in response to task demands REF.80). b | Also consistent with the concept of maintenance, a group of old maintainers (individuals with in other age groups and other forms of a mean age of 68.8 years who showed no significant episodic memory decline on verbal immediate free pathology. It is possible, however, that recall and delayed cued recall tasks over a period of two decades compared with young adults with a compensation mechanisms differ across mean age of 35.3 years) displayed levels of hippocampal activity comparable to those of young adults these different populations. during a fMRI study of face–name associative encoding82. By contrast, old decliners showed longitudinal In functional neuroimaging studies, episodic memory decline in the aforementioned verbal tasks and exhibited significantly lower hip- pocampal activity during associative encoding than young adults and old- maintainers. Consistent with the term compensation is often used to the idea that there are individuals who exhibit high levels of maintenance and those who exhibit low describe a situation in which brain activity levels of maintenance, maintainers and decliners were defined independently of their absolute levels or functional connectivity is greater or of memory and hippocampal activity. However, it is impossible to know whether the maintenance more widespread in older adults than it is observed involved repair or just an absence of a decline. BOLD, blood-oxygen-level-dependent. Part b in younger adults89. Greater brain activity is adapted with permission from REF.82, Society for Neuroscience. or connectivity is sometimes interpreted as

Nature Reviews | Neuroscience Perspectives

being beneficial to older adults without any mutually exclusive such that one or more easier, whereas a younger adult may choose additional supporting evidence. However, may co-occur within or across individuals. freestyle, which is faster but harder. It is we believe that two basic criteria must be important to note that selecting a neural fulfilled to attribute any greater activity or Compensation by upregulation. One form of implementation of a behavioural strategy connectivity observed in older adults to compensation relates to the enhancement need not be as deliberate as choosing a compensation. First, it should be clear what of cognitive performance by boosting a swimming stroke — the important point is is being compensated for: that is, evidence neural process in response to task demands. that the process selected by older adults should be available that the increased In such cases, the processes recruited by is also available to young adults but is less activation in older adults is directly older adults would be the same as those likely to be the one that supports their or indirectly related to some insufficiency or engaged by younger adults, and the primary performance. An experimental example of gap between available neural resources and difference between the ages would be compensation by selection is shown task demands (the supply–demand gap)86,89. quantitative: older adults would engage the in Fig. 3c. This gap may be due to an age-related process to a greater extent than younger Unlike compensation by upregulation, reduction in neural resources (for example, adults. Compensatory upregulation could compensation by selection involves a resulting from brain atrophy, reduced blood explain the frequent finding that at least qualitative difference in the cognitive flow, neurotransmitter deficits or reduced some age-related activity increases are processes engaged by older and younger neural specificity), to an increase in task evident within the brain regions that individuals and hence is likely to be demands, or to both. This can be explained younger adults recruit during the same associated with the recruitment of using a metaphor: using eyeglasses for task91. Although reports of greater activity in different brain regions rather than with the reading compensates for an insufficiency older adults could reflect inefficiency13, this recruitment of the same region with in visual acuity, and the magnitude of the interpretation is less likely when the greater differential levels of activity. However, to supply–demand gap depends both on activity is shown to correlate positively with avoid mistakenly attributing overactivation the visual deficit of the individual and the cognitive performance. in a particular brain region in older adults size of the letters. In the context of ageing, It is worth noting that young adults to selection rather than upregulation, the supply–demand gap is primarily may also upregulate activity in response it is important to examine the effects of due to the age-related decline in neural to increased task demands but that the manipulating task demands. Furthermore, resources. That is, we assume that, owing to demand threshold for such upregulation in some cases, older adults may show age-related neural decline, some older adults may be higher in young adults than it is for activation in a different region than have difficulty implementing cognitive older adults25,37. As task difficulty increases, young adults, suggesting selection, but operations that would not have taxed their neural activity (particularly in frontal further investigation may show that this younger selves. regions) tends to increase up to a certain region is a different component of the Second, evidence should be available level34, beyond which activation asymptotes functional network recruited by younger that the enhanced activation in older adults and ultimately declines37,38 (Fig. 3a). It is adults, suggesting compensation by is related to a beneficial effect on cognitive assumed that the asymptote reflects the upregulation. One possible method to performance. To be compensatory, the limit of available neural resources and that distinguish compensation by upregulation use of eyeglasses should be associated the final decline reflects the breakdown in from compensation by selection is to use with better reading performance than cognitive performance when these resources repetitive transcranial magnetic stimulation when eyeglasses are not used. This is a are exceeded. Several studies37,39–42 have to determine the effects of altering the point on which our view departs from found that, consistent with the reduced function of a particular region in younger some uses of the term compensation in availability of neural resources in these and older individuals (Box 2). the literature, in which the term is often individuals, older adults show a greater It is also important to note that applied to any age-related increase in increase in activity, a lower asymptote compensation by selection is related to brain activity or to the recruitment of and an earlier decline than younger adults individual differences in reserve: some additional brain regions in older but not (Fig. 3a,b). Because age-related differences in older adults may have a larger repertoire of young adults, regardless of the relationship brain activity can depend on task difficulty, alternative neural strategies to implement a with performance15,16. In our view, without researchers should ideally investigate given behaviour than others, and this may a link to performance, these findings multiple levels of task demands. reflect differences in accumulated reserve. should be simply described as age-related differences (increases or decreases) in Compensation by selection. Another Compensation by reorganization. activity and not as compensatory activity. mechanism of compensation is the Compensation may also occur when older Linking compensation to successful recruitment, by older adults, of neural adults use a neural mechanism to respond to performance (Box 2) helps to distinguish circuitry associated with cognitive processes ageing-induced losses that is not available compensation from activation differences that are available to but not engaged by to younger individuals86. The closest due to inefficiency, dedifferentiation or young adults under the same objective analogy to this type of compensation pathology90 (Box 1). task conditions. For example, older adults would be the development of new neural We believe that it is necessary to may engage a less effective but also less mechanisms following brain damage. For distinguish between three different demanding process, whereas younger example, there is evidence that recovery mechanisms or forms of compensation adults may prefer a more effective but from aphasia following a left-hemisphere (all incorporating the two criteria above): more demanding one. To explain this stroke is associated with the recruitment upregulation, selection and recruitment idea with a metaphor, during a swimming of right hemispheric regions that do not of additional processes (Fig. 3). We posit competition, an older adult may prefer support language processes in the normal that these forms of compensation are not to swim breaststroke, which is slower but brain92. Although these alternative circuits

www.nature.com/nrn Perspectives

may be less effective than the original Box 2 | Linking compensatory activity to successful cognitive performance ones, their recruitment may still benefit performance. In the case of ageing, there are three basic methods by which brain activations measured by functional imaging can be such reorganization could underlie the linked to successful cognitive performance. Correlation across participants is the most common approach and the only one available when well-established finding that older adults using blocked functional Mri (fMri) designs. if the activity of a brain region that is recruited to a often show more bilateral patterns of brain greater extent by older adults than by younger adults is positively correlated with performance in 44–46 activity than younger adults (Fig. 3d). If older adults, the finding is consistent with compensation. a potential problem associated with the new regions engaged by older adults correlation across participants is known as simpson’s paradox: the direction of association at the are sometimes recruited by younger population level may be different in the subgroups or the individuals composing the population109. adults in other similar (or more difficult) For example, if one hypothesizes that increased activity compensates for brain atrophy, then a conditions, this would support the notion positive activity–cognition correlation should be expected in individuals with high brain atrophy; of compensation by selection. It is, however, however, if one calculates the correlation using all individuals (including those with minimal brain worth noting that reorganization due to atrophy), a negative correlation could be found (see the figure, part a, which shows a hypothetical ageing and due to brain damage differs in data set that illustrates this point). thus, correlations should be conducted within the group in which compensation is assumed to take place. several ways, including the fact that the time Correlation within participants requires an event-related design (using fMri or electroencepha- course of ageing is slow, whereas the course lography (eeG)) and bypasses simpson’s paradox, but it requires the reasonable assumption that of brain injury is fast. It is not clear which compensatory processes vary from trial to trial. the activity–performance association should be time course is better for reorganization: stronger in individuals with greater brain decline (that is, those in whom there is a larger supply– a slow change gives the brain more time demand gap). Consistent with this idea, a study17 found that older adults with worse white matter to adapt, but a fast change provides a clear quality and worse cognitive performance showed greater success-related fMri activity (defined trigger for reorganization. Clarification of as the difference between activity for hits and activity for misses), an effect that they described as this issue awaits future research. “less wiring, more firing”. this negative association between structure and function was found in the Distinguishing between a narrow frontal lobes and in the medial temporal lobes, depending on whether individual differences were 17 and a broad notion of compensation has based on executive function or memory function scores (see figure b, based on data from ref. ). Non-invasive brain stimulation is another approach. if a brain region is engaged during a task been suggested to be useful. According by older but not younger adults, then disrupting or enhancing the function of this region using to this distinction, compensation by brain stimulation should have a greater impact on task performance in older than younger reorganization would meet the stringent adults. For example, a study found that repetitive transcranial magnetic stimulation (rtMs) of the criterion of compensation in the right prefrontal cortex (PFC) disrupted episodic memory retrieval in all individuals, regardless of narrow sense because a new process age, whereas rtMs of the left PFC disrupted retrieval in older adults but not in younger adults110. is generated in response to a loss. By this suggested that the left PFC region contributed to retrieval only in older adults (see the contrast, compensation by upregulation figure, part c). Brain stimulation goes beyond correlations by establishing a causal link between or by selection would only qualify as localized brain activity and performance. compensation in the broader sense ab because these forms of compensation rely Executive function Correlation within group on an already existing process (which is Correlation across all individuals upregulated) or an already existing strategy High brain atrophy (which is selected) and do not require the Low brain atrophy evolution of a new process or structure. 0.44 0.3 it y opy) 0.43 0.2 Conclusions ity 0.1 0.42 elated In this Opinion article, we have tried to 0.0 ain activ 0.41

Br –0.1 elucidate three important concepts that ain activ are widely used in studies of brain and 0.40 br –0.2 Success-r actional anistr

(fr 0.39 –0.3 cognitive ageing: reserve, maintenance and White-matter qualit y compensation. We propose that reserve is Cognitive performance High Low High Low used to refer to the accumulation of brain Performance level resources during the lifespan, maintenance Memory function to the preservation of these resources Baseline c Left PFC TMS via constant recovery and repair, and 50 compensation to the deployment of these Right PFC TMS resources in response to task demands. 45 0.47 0.4 In other words, reserve is about how much 40 opy) 0.46 0.3 ity you have, maintenance is about how well you ecognition 35 keep it, and compensation is about when 0.45 elated 0.2 30 0.44 0.1 ected r

and how you use it. r ain activ

(hits–false alarms) 25 Although we have discussed reserve, 0.43 br 0.0 Cor Success-r actional anistr

(fr 0.42 –0.1 maintenance and compensation in separate 20 White-matter qualit y Young adults Older adults sections, they can operate concurrently and High Low High Low affect each other. For example, if education Performance level augments reserve by increasing synaptic Part c is adapted with permission from REF.89, from Ch.37 ‘Frontal Lobes and aging: Deterioration and density, this can attenuate age-related Compensation’ by roberto Cabeza and Nancy a. Dennis from “Principles of Frontal Lobe Function”, cognitive decline if the new synapses are 2e edited by stuss, D. t. & Knight, r. t. (2013), by permission of Oxford university Press.

Nature Reviews | Neuroscience Perspectives

preserved via maintenance. Thus, when maintain it; it is also necessary to deploy Such studies will result in stronger models considering the potential effects of, for these resources during task performance in of successful cognitive ageing, which are example, cognitive training on reserve, response to task demands, that is, to engage essential for the interpretation of findings one has to consider whether age-related in compensation. Future neuroimaging from studies of pathological ageing (Box 1). deficits in maintenance will render any research should aim to more directly link In addition, we note that, despite our positive effects less effective. Likewise, it the predictions derived from studies of focus on healthy ageing, the concepts we is not enough to accumulate reserve and maintenance, reserve and compensation. have discussed can also be applied in other domains, including child development, a b acute brain injury, neurodegeneration 0.08 Young adults Older adults and psychiatric illness. We therefore hope 0.07 that the present paper will help promote ity 0.06 0.05 consensus in these domains as well. That ity 0.04 is, individuals with a neurological disease 0.03 or disorder may compensate for their 0.02 0.01 disorder-related deficits in ways similar to fMRI activ Young adults 0.00 those described here for healthy older adults. ight DLPFC activ R Older adults (change in MR signal) –0.01 We also note with caution that most studies –0.02 –0.03 of cognitive ageing to date have been limited Baseline 123456 4 letters5 letters7 letters to testing samples of high-functioning, Task demands Working memory load highly educated and mostly Caucasian cdhealthy older adults. To develop more Recollection-related Familiarity-related representative models of cognitive and brain Rhinal cortex ageing, we strongly recommend expanding inclusion criteria to encompass individuals Hippocampus from diverse backgrounds93. This will, of course, require funding sufficient to support 1.8 0.7 Young adults the large, ideally longitudinal, studies that 1.6 0.6 ession ession

ate 1.4 such research requires, with an emphasis on 1.2 0.5 egr egr combining longitudinal observations with

1.0 e slope 0.4 0.8 0.3 intervention studies to gauge the long-term

dized r 0.6 dized r 0.2 effects of physical exercise, cognitive coefficient of coefficient of 0.4 negativ exponential r training and other variables. Such studies 0.2 0.1 Low-performing High-performing

Standar 0.0 Standar 0.0 older adults older adults will allow researchers to better understand Young Old Young Old age-related changes in brain and cognition in terms of biological ageing (senescence), Fig. 3 | compensation mechanisms: upregulation, selection and reorganization. a | Compensation variations in environmental and genetic by upregulation occurs when neural activity increases in response to greater task demands. The hypo- thetical relationship between cognitive demand and brain activity illustrated in the graph is that as task factors for a given birth cohort, and the demands increase, activity first rises, then asymptotes and finally declines13,89. Because of reduced secular trends in health, education and neural resources, this demand–activity function is hypothesized to be shifted to the left in older adults, technology that will determine the ageing and hence, they would tend to show greater activity in the same regions as younger adults at lower trajectories of future generations. levels of task difficulty but lower activity at higher levels of task difficulty. b | An example of compensa- Roberto Cabeza1*, Marilyn Albert2, Sylvie Belleville3, tion by upregulation that is consistent with the hypothetical function in panel a: in a functional MRI Fergus I. M. Craik4, Audrey Duarte5, Cheryl L. Grady4, (fMRI) study , older adults showed greater working memory-related activity in the right dorsolateral Ulman Lindenberger6, Lars Nyberg7, Denise C. Park8, prefrontal cortex (DLPFC) than younger adults at lower levels of task demands but less activity at higher Patricia A. Reuter-Lorenz9, Michael D. Rugg8, 10 11 levels of task demands when there was a higher working memory load111. c | Compensation by selection Jason Steffener and M. Natasha Rajah occurs when older adults engage in a process not currently recruited by young adults but available to 1Center for Cognitive Neuroscience, Department of young adults, who may use it in other tasks or conditions. An example of compensation by selection is Psychology and Neuroscience, Duke University, shown. This fMRI study compared the effects of ageing on the rich form of memory known as recollec- Durham, NC, USA. tion and the less precise form of memory known as familiarity , measured in the same recognition mem- 2Departments of Psychiatry and Neurology, John ory task. Compared with younger adults, older adults showed reduced recollection-related activity in Hopkins University, Baltimore, MD, USA. the hippocampus but increased familiarity-related activity in the rhinal cortex. Thus, older adults com- 3Research Center of the Institut Universitaire de pensated for deficits in an optimal but demanding process (recollection) by recruiting a suboptimal but Gériatrie de Montréal, Montreal, Quebec, Canada. less demanding process (familiarity)112. d | The idea of compensation by reorganization is that older 4Rotman Research Institute, Baycrest Health Sciences, adults may use a neural mechanism to respond to ageing-induced losses that is not available to younger Toronto, Ontario, Canada. individuals. An example of compensation by reorganization is shown. During an episodic memory 5School of Psychology, Georgia Tech, Atlanta, retrieval task , young adults and low-performing older adults showed unilateral frontal activity , whereas GA, USA. high-performing older adults showed bilateral frontal activity , suggesting a reorganization of the epi- 6Max Planck Institute for Human Development and sodic retrieval network12. Part a is adapted with permission from REF.89, from Ch.37 ‘Frontal Lobes and Max Planck UCL Centre for Computational Psychiatry Aging: Deterioration and Compensation’ by Roberto Cabeza and Nancy A. Dennis from “Principles of and Ageing Research, Berlin, Germany. Frontal Lobe Function”, 2E edited by Stuss, D. T. & Knight, R . T. (2013), by permission of Oxford University 7Departments of Radiation Sciences and Integrated Press. Part b is adapted with permission from REF.111, Elsevier. Part c is adapted with permission from Medical Biology, UFBI, Umeå University, Umeå, REF.112, Daselaar, S. M. et al. Effects of healthy aging on hippocampal and rhinal memory functions: an Sweden. event-related fMRI study. Cereb. Cortex (2006) 16(12), 1771–1782, by permission of Oxford University 8Center for Vital Longevity, University of Texas, Dallas, Press. Part d is adapted with permission from REF.12, Elsevier. TX, USA.

www.nature.com/nrn Perspectives

9Department of Psychology, University of Michigan, 23. Papenberg, G., Salami, A., Persson, J., Lindenberger, U. 50. Trahan, L. H., Stuebing, K. K., Fletcher, J. M. & Ann Arbor, MI, USA. & Backman, L. Genetics and functional imaging: Hiscock, M. The Flynn effect: a meta-​analysis. Psychol. effects of APOE, BDNF, COMT, and KIBRA in aging. Bull. 140, 1332–1360 (2014). 10 Interdisciplinary School of Health Sciences, University Neuropsychol Rev. 25, 47–62 (2015). 51. Deary, I. J., Whiteman, M. C., Starr, J. M., Whalley, L. J. of Ottawa, Ottowa, Ontario, Canada. 24. Campisi, J. Cellular senescence and apoptosis: how & Fox, H. C. The impact of childhood intelligence on cellular responses might influence aging phenotypes. later life: following up the Scottish mental surveys of 11Departments of Psychiatry & Psychology, Exp. Gerontol. 38, 5–11 (2003). 1932 and 1947. J. Pers Soc. Psychol. 86, 130–147 McGill University and Douglas Hospital Research 25. Musiek, E. S. & Holtzman, D. M. Mechanisms linking (2004). Centre, Montreal, Quebec, Canada. circadian clocks, sleep, and neurodegeneration. 52. Nyberg, L., Pudas, S. & Lundquist, A. in Cogntiive Science 354, 1004–1008 (2016). Neuroscience of Aging 2nd edn (eds Cabeza, R., *e-mail: [email protected] 26. Hullinger, R. & Puglielli, L. Molecular and cellular Nyberg, L. & Park, D. C.) (Oxford Univ. Press, 2016). https://doi.org/10.1038/s41583-018-0068-2 aspects of age-related​ cognitive decline and Alzheimer’s 53. Raz, N. et al. Regional brain changes in aging healthy Published online 10 October 2018 disease. Behav. Brain Res. 322, 191–205 (2017). adults: general trends, individual differences and 27. Mattson, M. P. & Arumugam, T. V. Hallmarks of brain modifiers. Cereb. Cortex 15, 1676–1689 (2005). 1. Beard, J. R. et al. The world report on ageing and aging: adaptive and pathological modification by 54. Ghisletta, P., Rabbitt, P., Lunn, M. & Lindenberger, U. health: a policy framework for healthy ageing. Lancet metabolic states. Cell Metab. 27, 1176–1199 (2018). Two thirds of the age-​based changes in fluid and 387, 2145–2154 (2016). 28. Santoro, A. et al. Innate immunity and cellular crystallized intelligence, perceptual speed, and 2. Gerstorf, D. et al. Secular changes in late-​life cognition senescence: the good and the bad in the developmental memory in adulthood are shared. Intelligence 40, and well-being:​ towards a long bright future with a short and aged brain. J. Leukoc. Biol. 103, 509–524 (2018). 260–268 (2012). brisk ending? Psychol. Aging 30, 301–310 (2015). 29. Backman, L., Lindenberger, U., Li, S. C. & Nyberg, L. 55. Stern, Y., Gazes, Y., Razlighi, Q., Steffener, J. & 3. Cabeza, R., Nyberg, L. & Park, D. C. Cognitive Linking cognitive aging to alterations in dopamine Habeck, C. A task-​invariant cognitive reserve network. Neuroscience of Aging: Linking Cognitive and Cerebral neurotransmitter functioning: recent data and future Neuroimage 178, 36–45 (2018). Aging 2nd edn (Oxford Univ. Press, New York, 2017). avenues. Neurosci. Biobehav. Rev. 34, 670–677 (2010). 56. Piras, F., Cherubini, A., Caltagirone, C. & Spalletta, G. 4. Sole-Padulles,​ C. et al. Brain structure and function 30. Sampedro-Piquero,​ P., Alvarez-​Suarez, P. & Begega, A. Education mediates microstructural changes in bilateral related to cognitive reserve variables in normal aging, Coping with stress during aging: the importance of a hippocampus. Hum. Brain Mapp. 32, 282–289 (2011). mild cognitive impairment and Alzheimer’s disease. resilient brain. Curr. Neuropharmacol. 16, 284–296 57. Stern, Y. Cognitive reserve and Alzheimer disease. Neurobiol. Aging 30, 1114–1124 (2009). (2018). Alzheimer Dis. Assoc. Disord. 20, 112–117 (2006). 5. Arenaza-Urquijo,​ E. M. et al. Association between 31. Rosario, E. R., Chang, L., Head, E. H., Stanczyk, F. Z. & 58. Scarmeas, N. et al. Cognitive reserve-​mediated educational attainment and amyloid deposition across Pike, C. J. Brain levels of sex steroid hormones in men modulation of positron emission tomographic the spectrum from normal cognition to dementia: and women during normal aging and in Alzheimer’s activations during memory tasks in Alzheimer disease. neuroimaging evidence for protection and disease. Neurobiol. Aging 32, 604–613 (2011). Arch. Neurol. 61, 73–78 (2004). compensation. Neurobiol. Aging 59, 72–79 (2017). 32. Morgan, D. G. The dopamine and serotonin systems 59. Stern, Y. Cognitive reserve. Neuropsychologia 47, 6. Barulli, D. & Stern, Y. Efficiency, capacity, compensation, during aging in human and rodent brain. A brief 2015–2028 (2009). maintenance, plasticity: emerging concepts in cognitive review. Prog. Neuropsychopharmacol. Biol. Psychiatry 60. Soldan, A. et al. Relationship of medial temporal lobe reserve. Trends Cogn. Sci. 17, 502–509 (2013). 11, 153–157 (1987). atrophy, APOE genotype, and cognitive reserve in 7. Nyberg, L., Lovden, M., Riklund, K., Lindenberger, U. 33. Freeman, G. B. & Gibson, G. E. Dopamine, acetylcholine, preclinical Alzheimer’s disease. Hum. Brain Mapp. 36, & Backman, L. Memory aging and brain maintenance. and glutamate interactions in aging. Behavioral and 2826–2841 (2015). Trends Cogn. Sci. 16, 292–305 (2012). neurochemical correlates. Ann. N. Y. Acad. Sci. 515, 61. Bialystok, E., Craik, F. I. M. & Luk, G. Bilingualism: 8. Grady, C. L. Age-​related changes in cortical blood 191–202 (1988). consequences for mind and brain. Trends Cogn. Sci. flow activation during perception and memory. 34. Valenzuela, M. J., Breakspear, M. & Sachdev, P. 16, 240–250 (2012). Ann. N. Y. Acad. Sci. 777, 14–21 (1996). Complex mental activity and the aging brain: 62. Prakash, R. S., Voss, M. W., Erickson, K. I. & 9. Rajah, M. N. & D’Esposito, M. Region-​specific changes molecular, cellular and cortical network mechanisms. Kramer, A. F. Physical activity and cognitive vitality. in prefrontal function with age: a review of PET and Brain Res. Rev. 56, 198–213 (2007). Annu. Rev. Psychol. 66, 769–797 (2015). fMRI studies on working and episodic memory. 35. Hibar, D. P. et al. Common genetic variants influence 63. Scarmeas, N. & Stern, Y. Cognitive reserve and lifestyle. Brain 128, 1964–1983 (2005). human subcortical brain structures. Nature 520, J. Clin. Exp. Neuropsychol. 25, 625–633 (2003). 10. Persson, J. & Nyberg, L. Altered brain activity in 224–229 (2015). 64. Bialystok, E., Craik, F. I. M. & Freedman, M. healthy seniors: what does it mean? Prog. Brain Res. 36. Sperling, R. Potential of functional MRI as a biomarker Bilingualism as a protection against the onset of 157, 45–56 (2006). in early Alzheimer’s disease. Neurobiol. Aging 32 symptoms of dementia. Neuropsychologia 45, 11. Park, D. C. & Reuter-​Lorenz, P. The adaptive brain: (Suppl. 1), S37–S43 (2011). 459–464 (2007). aging and neurocognitive scaffolding. Annu. Rev. 37. Grady, C. L. The cognitive neuroscience of ageing. 65. Alladi, S. et al. Bilingualism delays age at onset of Psychol. 60, 173–196 (2009). Nat. Rev. Neurosci. 13, 491–505 (2012). dementia, independent of education and immigration 12. Cabeza, R., Anderson, N. D., Locantore, J. K. & 38. Tromp, D., Dufour, A., Lithfous, S., Pebayle, T. & status. Neurology 81, 1938–1944 (2013). McIntosh, A. R. Aging gracefully: compensatory brain Despres, O. Episodic memory in normal aging and 66. Anthony, M. & Lin, F. A. Systematic review for activity in high-​performing older adults. Neuroimage Alzheimer disease: insights from imaging and functional neuroimaging studies of cognitive reserve 17, 1394–1402 (2002). behavioral studies. Ageing Res. Rev. 24, 232–262 across the cognitive aging spectrum. Arch. Clin. 13. Reuter-Lorenz,​ P. A. & Cappell, K. A. Neurocognitive (2015). Neuropsychol. https://doi.org/10.1093/arclin/acx125 aging and the compensation hypothesis. Curr. Direct. 39. Walhovd, K. B. et al. Consistent neuroanatomical (2017). Psychol. Sci. 17, 177–182 (2008). age-related​ volume differences across multiple 67. Reed, B. R. et al. Cognitive activities during adulthood 14. Nilsson, J. & Lövdén, M. Naming is not explaining: samples. Neurobiol. Aging 32, 916–932 (2011). are more important than education in building reserve. future directions for the “cognitive reserve” and “brain 40. Rajah, M. N., Maillet, D. & Grady, C. L. in The Wiley J. Int. Neuropsychol. Soc. 17, 615–624 (2011). maintenance” theories. Alzheimer’s Res. Ther. 10, 34 Handbook of Cognitive Neuroscience of Memory 68. Zahodne, L. B. et al. Quantifying cognitive reserve in (2018). (eds Addis, D., Barense, M. & Duarte, A.) 347–361 older adults by decomposing episodic memory variance: 15. Cabeza, R., Anderson, N. D., Houle, S., Mangels, J. A. (Wiley Publishers, New York, 2015). replication and extension. J. Int. Neuropsychol. Soc. 19, & Nyberg, L. Age-​related differences in neural activity 41. Nyberg, L. et al. Longitudinal evidence for diminished 854–862 (2013). during item and temporal-​order memory retrieval: frontal cortex function in aging. Proc. Natl Acad. Sci. 69. Reed, B. R. et al. Measuring cognitive reserve based a positron emission tomography study. J. Cognitive USA 107, 22682–22686 (2010). on the decomposition of episodic memory variance. Neurosci. 12, 1–10 (2000). 42. Salat, D. H. et al. Age-​related changes in prefrontal Brain 133, 2196–2209 (2010). 16. Cabeza, R. et al. Age-​related differences in neural white matter measured by diffusion tensor imaging. 70. Bernardi, G. et al. How skill expertise shapes the brain activity during memory encoding and retrieval: Ann. N. Y. Acad. Sci. 1064, 37–49 (2005). functional architecture: an fMRI study of visuo-​spatial a positron emission tomography study. J. Neurosci. 43. Giorgio, A. et al. Age-​related changes in grey and and motor processing in professional racing-​car and 17, 391–400 (1997). white matter structure throughout adulthood. naive drivers. PLOS ONE 8, e77764 (2013). 17. Daselaar, S. M. et al. Less wiring, more firing: Neuroimage 51, 943–951 (2010). 71. Adamson, M. M. et al. Higher landing accuracy in low-performing​ older adults compensate for impaired 44. Madden, D. J. et al. Adult age differences in functional expert pilots is associated with lower activity in the white matter with greater neural activity. Cereb. Cortex connectivity during executive control. Neuroimage 52, caudate nucleus. PLOS ONE 9, e112607 (2014). 25, 983–990 (2015). 643–657 (2010). 72. Kim, W. et al. An fMRI study of differences in brain 18. Arenaza-​Urquijo, E. M. & Vemuri, P. Resistance versus 45. Craik, F. I. M. & Salthouse, T. A. The Handbook of activity among elite, expert, and novice archers at the resilience to Alzheimer disease: clarifying terminology Aging and Cognition (Lawrence Erlbaum Associates, moment of optimal aiming. Cogn. Behav. Neurol. 27, for preclinical studies. Neurology 90, 695–703 (2018). Mahwah, NJ, 2000). 173–182 (2014). 19. Raz, N. & Daugherty, A. M. Pathways to brain 46. Lindenberger, U. Human cognitive aging: corriger la 73. Kozasa, E. H. et al. Effects of a 7-day meditation retreat aging and their modifiers: free-​radical-induced fortune? Science 346, 572–578 (2014). on the brain function of meditators and non-​meditators energetic and neural decline in senescence (FRIENDS) 47. World Health Organisation. World Report on Ageing during an attention task. Front. Hum. Neurosci. 12, model-a mini-review. Gerontology 64, 49–57 (2018). and Health (eds Beard, J., Officer, A. & Cassels, A.) 222 (2018). 20. Miller, R. A. Age-​related changes in T cell surface (WHO, Luxembourg, 2015). 74. Li, Y. et al. Sound credit scores and financial decisions markers: a longitudinal analysis in genetically 48. Habib, R., Nyberg, L. & Nilsson, L. G. Cognitive and despite cognitive aging. Proc. Natl Acad. Sci. USA heterogeneous mice. Mech. Ageing Dev. 96, 181–196 non-​cognitive factors contributing to the longitudinal 112, 65–69 (2015). (1997). identification of successful older adults in the Betula 75. Lindenberger, U., Kliegl, R. & Baltes, P. B. Professional 21. Roy, A. K. et al. Impacts of transcriptional regulation study. Aging Neuropsychol. Cogn. 14, 257–273 expertise does not eliminate age-​differences in on aging and senescence. Ageing Res. Rev. 1, (2007). imagery-based​ memory performance during 367–380 (2002). 49. Ronnlund, M. & Nilsson, L. G. Flynn effects on adulthood. Psychol. Aging 7, 585–593 (1992). 22. Foster, T. C. Role of estrogen receptor alpha and beta sub-​factors of episodic and semantic memory: parallel 76. Morrow, D., Leirer, V., Altieri, P. & Fitzsimmons, C. expression and signaling on cognitive function during gains over time and the same set of determining When expertise reduces age-​differences in performance. aging. Hippocampus 22, 656–669 (2012). factors. Neuropsychologia 47, 2174–2180 (2009). Psychol. Aging 9, 134–148 (1994).

Nature Reviews | Neuroscience Perspectives

77. Vaci, N., Gula, B. & Bilalic, M. Is age really cruel to 93. Falk, E. B. et al. What is a representative brain? science: a practical guide. Frontiers Psychol. 4, 14 experts? Compensatory effects of activity. Psychol. Neuroscience meets population science. Proc. Natl (2013). Aging 30, 740–754 (2015). Acad. Sci. USA 110, 17615–17622 (2013). 110. Rossi, S. et al. Age-​related functional changes of 78. Stern, Y., Albert, S., Tang, M. & Tsai, W. 94. Albert, M. S. et al. The diagnosis of mild cognitive prefrontal cortex in long-​term memory: a repetitive Rate of memory decline in AD is related to education impairment due to Alzheimer’s disease: transcranial magnetic stimulation study. J. Neurosci. and occupation. Neurology 1, 1942–1947 recommendations from the National Institute on 24, 7939–7944 (2004). (1999). Aging-Alzheimer’s​ Association workgroups on 111. Cappell, K. A., Gmeindl, L. & Reuter-​Lorenz, P. A. Age 79. Ten Brinke, L. F. et al. Aerobic exercise increases diagnostic guidelines for Alzheimer’s disease. differences in prefontal recruitment during verbal cortical white matter volume in older adults with Alzheimers Dement. 7, 270–279 (2011). working memory maintenance depend on memory vascular cognitive impairment: a 6-month 95. Driscoll, I. & Troncoso, J. Asymptomatic Alzheimer’s load. Cortex 46, 462–473 (2010). randomized controlled trial. Alzheimers Dement. disease: a prodrome or a state of resilience? 112. Daselaar, S. M., Fleck, M., Dobbins, I. G., Madden, D. J. 11, 606 (2015). Curr. Alzheimer Res. 8, 330–335 (2011). & Cabeza, R. Effects of healthy aging on hippocampal 80. Gorbach, T. et al. Longitudinal association between 96. Brickman, A. M. et al. White matter hyperintensities and rhinal memory functions: an event-related hippocampus atrophy and episodic-​memory decline. and cognition: testing the reserve hypothesis. fMRI study. Cereb. Cortex 16, 1771–1782 (2006). Neurobiol. Aging 51, 167–176 (2017). Neurobiol. Aging 32, 1588–1598 (2011). 81. Persson, J. et al. Longitudinal structure–function 97. Landau, S. M. et al. Association of lifetime cognitive Acknowledgements correlates in elderly reveal MTL dysfunction with engagement and low β-​amyloid deposition. This manuscript presents a summary of discussions from a cognitive decline. Cereb. Cortex 22, 2297–2304 Arch. Neurol. 69, 623–629 (2012). 2-day symposium held at McGill University, Montreal, (2012). 98. Soldan, A. et al. Relationship of cognitive reserve and Canada, from 31 May–2 June 2017, which was funded by a 82. Pudas, S. et al. Brain characteristics of individuals cerebrospinal fluid biomarkers to the emergence of Canadian Institutes of Health Research (CIHR) Planning and resisting age-​related cognitive decline over two clinical symptoms in preclinical Alzheimer’s disease. Dissemination Grant 373172 awarded to M.N.R. and R.C. decades. J. Neurosci. 33, 8668–8677 (2013). Neurobiol. Aging 34, 2827–2834 (2013). and by institutional funds from Duke University (North 83. Miller, E. K. & Cohen, J. D. An integrative theory of 99. Dickerson, B. C. et al. Medial temporal lobe Carolina, USA) and the Douglas Hospital Research Centre prefrontal cortex function. Annu. Rev. Neurosci. 24, function and structure in mild cognitive impairment. (Montreal, Canada). R.C. is supported by a grant from the US 167–202 (2001). Ann. Neurol. 56, 27–35 (2004). National Institutes of Health (NIH; RO1-AG19731). M.A. is 84. Raz, N. & Lindenberger, U. Only time will tell: 100. Huijbers, W. et al. Amyloid-​β deposition in mild supported by a grant from the NIH National Institute on cross-sectional​ studies offer no solution to the cognitive impairment is associated with increased Aging (NIA; P50-AG005146). S.B. is supported by a grant age-​brain-cognition triangle: comment on Salthouse hippocampal activity, atrophy and clinical progression. from the National Sciences and Engineering Research Council (2011). Psychol. Bull. 137, 790–795 (2011). Brain 138, 1023–1035 (2015). of Canada (NSERC; RGPIN-2016-06132). F.I.M.C. is sup- 85. Kovari, E., Herrmann, F. R., Bouras, C. & Gold, G. 101. Clement, F. & Belleville, S. Compensation and disease ported by a grant from the Natural Sciences and Engineering Amyloid deposition is decreasing in aging brains: an severity on the memory-​related activations in mild Research Council (NSERC; A8261). A.D. is supported by a autopsy study of 1,599 older people. Neurology 82, cognitive impairment. Biol. Psychiatry 68, 894–902 grant from NIH (R56-AG049793). C.L.G. is supported by 326–331 (2014). (2010). a grant from CIHR (MOP-143311). U.L. is supported by the 86. Lovden, M., Backman, L., Lindenberger, U., Schaefer, S. 102. Bookheimer, S. Y. et al. Patterns of brain activation in Max Planck Society. L.N. is support by a scholar grant from & Schmiedek, F. A. Theoretical framework for the people at risk for Alzheimer’s disease. N. Engl. J. Med. the Knut and Alice Wallenberg Foundation. D.C.P. is sup- study of adult cognitive plasticity. Psychol. Bull. 136, 343, (450–456 (2000). ported by a grant from the NIH (R01-AG006265). P.A.R.-L. 659–676 (2010). 103. Rajah, M. N. et al. Family history and APOE4 risk for is supported by a grant from the NIH (R21-AG045460). 87. Beam, C. R. & Turkheimer, E. Phenotype-​environment Alzheimer’s disease impact the neural correlates of M.D.R. is supported by a grant from the NIH correlations in longitudinal twin models. Dev. episodic memory by early midlife. Neuroimage Clin. (RF1-AG039103). M.N.R. is supported by grants from CIHR Psychopathol. 25, 7–16 (2013). 14, 760–774 (2017). (MOP 126105) and NSERC (RGPIN-2018-05761). 88. Lovden, M., Ghisletta, P. & Lindenberger, U. Social 104. Celone, K. A. et al. Alterations in memory networks participation attenuates decline in perceptual speed in mild cognitive impairment and Alzheimer’s disease: Author contributions in old and very old age. Psychol. Aging 20, 423–434 an independent component analysis. J. Neurosci. 26, R.C., M.A., S.B., F.I.M.C., A.D., C.L.G., U.L., L.N., D.C.P., (2005). 10222–10231 (2006). P.A.R.-L., M.D.R., J.S. and M.N.R. researched data for the 89. Cabeza, R. & Dennis, N. A. in Principles of Frontal 105. Elman, J. A. et al. Neural compensation in older people article, made a substantial contribution to the discussion of Lobe Function (eds Stuss, D. T. & Knight, R. T.) with brain amyloid-​beta deposition. Nat. Neurosci. 17, content, wrote the article and reviewed and/or edited the (Oxford Univ. Press, New York, 2013). 1316–1318 (2014). manuscript before submission. 90. Bakker, A. et al. Reduction of hippocampal 106. Esposito, Z. et al. Amyloid β, glutamate, excitotoxicity hyperactivity improves cognition in amnestic mild in Alzheimer’s disease: are we on the right track? Competing interests cognitive impairment. Neuron 74, 467–474 Cns Neurosci. Ther. 19, 549–555 (2013). The authors declare no competing interests. (2012). 107. Rudy, C. C., Hunsberger, H. C., Weitzner, D. S. & 91. Spreng, R. N., Wojtowicz, M. & Grady, C. L. Reliable Reed, M. N. The role of the tripartite glutamatergic Publisher’s note differences in brain activity between young and old synapse in the pathophysiology of Alzheimer’s disease. Springer Nature remains neutral with regard to jurisdictional adults: a quantitative meta-​analysis across multiple Aging Dis. 6, 131–148 (2015). claims in published maps and institutional affiliations. cognitive domains. Neurosci. Biobehav. Rev. 34, 108. Belleville, S. et al. Training-​related brain plasticity in 1178–1194 (2010). subjects at risk of developing Alzheimer’s disease. Reviewer information 92. Cabeza, R. Hemispheric asymmetry reduction in older Brain 134, 1623–1634 (2011). Nature Reviews Neuroscience thanks C. Brayne, R. Dixon, W. adults. The HAROLD model. Psychol. Aging 17, 109. Kievit, R. A., Frankenhuis, W. E., Waldorp, L. J. & Jagust and the other anonymous reviewer(s) for their 85–100 (2002). Borsboom, D. Simpson’s paradox in psychological contribution to the peer review of this work.

www.nature.com/nrn