Understanding the atrophy of the aging brain is essential to understanding age-related and individual differences in neurocognitive functions across a lifespan. There is a wealth of neuroimaging evidence, including structural magnetic resonance imaging (MRI) across several cross-sectional and longitudinal studies, which demonstrates that biological aging is characterized by reductions in gray matter volume (Raz et al., 2005; Resnick, Pham, Kraut, Zonderman, & Davatzikos, 2003; Rodrigue & Raz, 2004), density (Good et al., 2001; Smith, Chebrolu, Wekstein, Schmitt, & Markesbery, 2007), and cortical thickness (Dickerson et al., 2009; Fjell et al., 2006, 2009; Salat et al., 2004), with the prefrontal, parietal and hippocampal regions showing greater shrinkage compared to the temporal and occipital regions, which show more minimal decreases. For example, Raz et al. (2005) reported volumetric measurements of a number of structural brain MRI images in lifespan samples, tracking both cross-sectional and longitudinal change over 5 years. They showed gradual volumetric changes with age, but also found that these changes are not equivalent across brain regions, with the greatest age-related shrinkage occurring in the prefrontal, hippocampal, caudate, and cerebellar regions. Relative stability was observed in the entorhinal cortex and the visual regions of the brain (Raz et al., 2005).

Given that the performance of neurocognitive functions may stem from changes in the brain's structural cytoarchitecture, many studies have embarked on associating age-related differences in regional brain volume with individuals' cognitive functioning measured by a variety of neuropsychological tests. Although structure-cognition associations are not easily replicated and appear to be sensitive to subtypes of neurocognitive assessment and to sample composition, there is also clear evidence that these age-related reductions in brain structure may index specific cognitive decline in older adults. For example, age-related volumetric shrinkage in prefrontal gray matter has been related to lower fluid intelligence (Fjell et al., 2006; Head, Rodrigue, Kennedy, & Raz, 2008), poorer inhibitory/motor control and working memory performance (Head, Kennedy, Rodrigue, & Raz, 2009; Tapp et al., 2004), but better verbal fluency (Elderkin-Thompson, Ballmaier, Hellemann, Pham, & Kumar, 2008). Moreover, reduced hippocampal volume has been reported to be associated with poorer episodic memory (Head et al., 2008; Rodrigue & Raz, 2004).

2.2.2 Age-related reduction in micro-structural neural fibers and its cognitive consequences

In addition to the shrinkage in gray matter, biological aging is also associated with a reduction in white matter volume and density (Raz et al., 2005; Resnick et al., 2003; Salat et al., 2009; Walhovd et al., 2011). More recently, studies using diffusion tensor imaging (DTI), a non-invasive imaging technique, have suggested that white matter changes with aging involve a reduction in the integrity of white matter fibers that may be related to de-myelination along the axonal fibers and may result in less efficient conduction of neural signals and impaired transmission of information across the brain (Cox et al., 2016; Fan et al., 2019; Madden, Bennett, & Song, 2009; Sullivan & Pfefferbaum, 2006; Yang, Tsai, Liu, Huang, & Lin, 2016). The most consistent pattern of age-related difference in white matter fiber integrity is the greater reduction in the anterior regions of the corpus callosum, especially in the genu and body of the corpus callosum and pericallosal frontal white matter, relative to the posterior regions of the corpus callosum (Davis, Kragel, Madden, & Cabeza, 2011; Fan et al., 2019; Gordon et al., 2008; Gunning-Dixon & Raz, 2003; Sullivan & Pfefferbaum, 2006; Sullivan, Rohlfing, & Pfefferbaum, 2010), reflecting an anterior-to-posterior deterioration in white-matter diffusivity and anisotropy with age.

There is also converging evidence that age-related reductions in white matter microstructural integrity have direct consequences for cognitive performance (Bennett & Madden, 2014; Fan et al., 2019; Ritchie et al., 2015). With regard to processing speed, Kennedy and Raz (2009) found that slowed speed of cognitive processing was related to a reduced integrity in anterior white matter regions, suggesting that general slowing in older adults may stem from degraded neural transmission along the axonal fibers of the aging brain. Moreover, with regard to working memory, Charlton et al. (2006) first reported a significant correlation between working memory performance and white matter integrity, with Digit-Symbol sequencing performance negatively correlated with integrity in anterior, middle, and posterior portions of the aging brain. Zahr, Rohlfing, Pfefferbaum, and Sullivan (2009) further reported a link between working memory and the corpus callosum, with poor working memory in older adults relating to less integrity of the anterior portion of the corpus callosum. Additionally, Kennedy and Raz (2009) found that poorer working memory in older adults appeared to be associated with age-related reduction in white matter integrity in widespread networks of regions ranging from anterior (prefrontal, anterior corpus callosum) to posterior (temporal and occipital) regions of the brain. These DTI findings suggest that working memory performance depends on white matter integrity in a widespread network of the aging brain. Finally, with regard to the domain of cognitive control, increased task switching costs have been reported to be associated with reduced integrity of fronto-parietal white matter (Gold, Powell, Xuan, Jicha, & Smith, 2010), integrity of prefrontal (Gratton, Wee, Rykhlevskaia, Leaver, & Fabiani, 2009), anterior corpus callosum, superior parietal, and occipital white matter (Kennedy & Raz, 2009), as well as the genu of the corpus callosum (Madden et al., 2009). Also, higher Stroop interference costs were found to be associated with reduced anisotropy of the posterior white matter in older adults, particularly in parietal, splenium, and occipital regions of the aging brain (Kennedy & Raz, 2009). Moreover, Davis et al. (2009) used a fiber tracking technique on young and older adults, and reported that greater white matter integrity in anterior regions but not posterior regions of the aging brain was related to better executive functioning (Davis et al., 2009). In addition, our recent DTI findings demonstrated that increased micro-structural integrity in both anterior corpus callosum and right cingulum/angular fibers positively correlated with performance on a visuospatial task in older adults. The further mediation analysis revealed that the white matter integrity of the frontal inter-hemispheric fibers was a significant mediator of the age-and-visuospatial performance relation in older adults, but not in younger adults (Fan et al., 2019). These results suggest that, congruent with the findings in the working memory domain, performance of executive control in older adults appears to depend on white matter integrity in a widespread network of the aging brain.

Age-related patterns of change, including more activity and more distributed activity, especially in the fronto-parietal cortices, have been frequently reported across a variety of cognitive and motor domains, including language comprehension, working memory, cognitive control, and motor function (Cabeza et al., 2018; Grady, 2012; Park & Reuter-Lorenz, 2009). In functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) studies, healthy older participants exhibited either (1) greater bilateral activity than the more unilateral activity observed in their young counterparts, and/or (2) additional and more distributed activation in regions that are not activated in young adults (Cabeza, Anderson, Locantore, & McIntosh, 2002; Cabeza et al., 2004; Huang, Meyer, & Federmeier, 2012; Huang, Polk, Goh, & Park, 2012; Jimura & Braver, 2010). Therefore, the phenomena of greater and more distributed fronto-parietal activity in older adults has been identified as a general characteristic of age-related functional neural change (Cabeza et al., 2018, 2004; Park & Reuter-Lorenz, 2009).

Age-related greater activity in fronto-parietal regions is often interpreted as being compensatory and involved in the improvement or maintenance of behavioral performance in the face of age-related neurodegeneration (Cabeza et al., 2002; Davis, Dennis, Daselaar, Fleck, & Cabeza, 2008; Heuninckx, Wenderoth, & Swinnen, 2008; Huang, Meyer, & Federmeier, 2012; Huang, Polk, Goh, & Park, 2012; Vallesi, McIntosh, & Stuss, 2011). For example, Huang, Meyer, and Federmeier (2012) and Huang, Polk, Goh, and Park (2012) investigated the functional significance of posterior parietal cortex when young and older adults performed a physical-numerical interference paradigm (i.e., numerical Stroop paradigm) while undergoing fMRI. In this paradigm, stimuli were pairs of Arabic digits shown simultaneously in the middle of a computer screen. Two types of magnitude judgments were included. For the numerical judgment task, participants were asked to judge which digit was numerically larger while ignoring their physical sizes. For the physical size task, participants were required to decide which one was physically larger while ignoring the numerical magnitude of the digits. In congruent trials, the digit that was larger in numerical magnitude was also larger in physical size. In incongruent trials, the digit that was larger in numerical magnitude was smaller in physical size. We found that, in the Incongruent vs. Congruent conditions contrast, young adults engaged more right parietal cortex for the Magnitude task and activated more left parietal cortex for the Size task, whereas older adults showed bilateral parietal activation for both Size and Magnitude tasks. These findings suggested a task-specific neural recruitment in the parietal cortex, as the hemisphere that was engaged depended on the judgment required. Moreover, we demonstrated the age-related additional parietal activity (i.e., right parietal for the Size and left parietal for the Magnitude task in older adults) was associated with better performance. Finally, as in many previous studies, we showed that older adults also recruited their left prefrontal cortex during both tasks, and this common activation was also associated with better performance (Huang, Meyer, & Federmeier, 2012; Huang, Polk, Goh, & Park, 2012). These results provide evidence for task-specific compensatory recruitment in the parietal cortex as well as task-independent compensatory recruitment in the left prefrontal cortex during cognitive control in normal aging.

Several functional neuroimaging studies use similar designs to explore compensatory mechanisms of the aging brain, and also suggest that greater and more distributed neural activation in anterior brain areas (e.g., prefrontal cortex), in addition to being beneficial for behavioral performance, may serve a compensatory role for declining neural function in more posterior sites, including the medial temporal lobe (Cabeza et al., 2004; Gutchess et al., 2005; Park et al., 2003), and occipital visual areas (Cabeza et al., 2004; Davis et al., 2008; Goh, Suzuki, & Park, 2010). For example, given the observation of age-related cognitive declines in memory encoding and retrieval and volumetric shrinkage in the memory-related hippocampal area, Gutchess et al. (2005) conducted an fMRI study to investigate the functional relationship between prefrontal and medial temporal activation during an incidental memory encoding task when both young and older participants were scanned. The results showed that older adults had reduced activation compared to young adults in the left and right parahippocampus. More importantly, such reduced activation in the medial temporal lobe was significantly coupled with greater activation in the middle frontal cortex. Similarly, Goh et al. (2010) also demonstrated that older adults showed reduced activation in occipital visual areas during a face discrimination task. However, these decreased neural responses were associated with greater activation in the prefrontal cortex for older adults compared to their young counterparts. These findings are consistent with the Posterior Anterior Shift in Aging (PASA) hypothesis, which posits a neurocognitive compensatory role of prefrontal regions for age-related neural deterioration in posterior brain regions (Cabeza et al., 2004; Davis et al., 2008).

Overall, these findings suggest that the brain ages in a dynamic way, declining in some respects but maintaining the ability to engage adaptive neural functions even in advanced age.

2.3.2 Altered functional connectivity with age

There is converging evidence that biological aging is associated with significant changes in functional connectivity between different brain regions, resulting in less effective neural communication between different areas in the aging brain (Andrews-Hanna et al., 2007; Damoiseaux et al., 2008; Duzel, Schutze, Yonelinas, & Heinze, 2011; Persson, Lustig, Nelson, & Reuter-Lorenz, 2007). Functional connectivity analysis involves correlating low-frequency neuronal signal in resting-state functional MRI data (data are acquired as participants simply lay still in the MRI scanner) between predefined regions-of-interest (ROIs). Several studies have suggested that, compared to younger counterparts, there is a posterior-to-anterior gradient of decline in functional connectivity in older adults. The findings suggest that age-related reductions in occipital activity and increases in prefrontal activity (i.e., PASA) may be associated with less effective functional communication between neural networks of the aging brain. Moreover, some studies measuring functional connectivity during the resting state among predefined ROIs have demonstrated that intrinsic brain connectivity in the aging brain is related to the performance of older individuals on a variety of cognitive tasks, including executive function and processing speed (Damoiseaux et al., 2008), associative memory (Wang et al., 2010), and working memory (Sambataro et al., 2010).

Thus far, we have reviewed findings from human neuroimaging studies showing age-related changes in brain volume and cortical thickness, micro-structural neural fibers, functional activation, and functional connectivity. However, interpreting age-related cognitive declines in the context of structural and functional changes in the aging brain at multiple levels of analysis is an ongoing challenge. Aging continues to be associated with cognitive decline and preservation in many individuals; what are the putative causal explanations for the middle-aged and/or older adults who exhibit substantial cognitive loss (and may develop dementia over the next few years) and for those who show intact cognitive abilities over the age of 80? Can middle-aged and older adults rely upon neural resources (i.e., neurocognitive reserve) to compensate for age-related declines in cognitive function and structural brain volume? In the next section, we will discuss whether and how the concept of neurocognitive reserve interacts with age-related and individual differences in cognition, brain structure, and neural function.

3.2.1 Indicators of NCR: Educational level and occupational complexity

A number of cross-sectional behavioral studies have been conducted to examine the relationship between the three common indicators of NCR and neurocognitive functions in healthy older adults. For educational level, years of schooling has been used as a proxy measure of NCR and related to cognitive decline, brain atrophy, and neural function in late adulthood, with the observation that highly educated individuals show better cognitive performance as evaluated by neuropsychological assessment and larger cortical thickness as well as regional brain volume (Liu et al., 2012; Mungas, Reed, Farias, & Decarli, 2009; Opdebeeck, Martyr, & Clare, 2015; Staff, Murray, Deary, & Whalley, 2004;). However, the relationship of this proxy measure with cognitive function appears to differ across different cognitive domains, with a strong correlation between educational level and several measures of memory (e.g., short-term memory, long-term memory and relational memory) (Angel, Fay, Bouazzaoui, Baudouin, & Isingrini, 2010; Lee, Lee, & Yang, 2012) but a weak correlation between educational level and several measures of executive functions (e.g., inhibitory control, working memory, and task switching) (Lee et al., 2012; Mueller, Raymond, & Yochim, 2013).

Similarly, behavioral studies examining the association between occupational status as a proxy measure of NCR and different types of cognitive performance with age have demonstrated that higher levels of complexity of occupation are related to better cognitive performance in healthy older people (Opdebeeck et al., 2015), with a moderate correlation with executive functions (Foubert-Samier et al., 2012) and a weak correlation with memory performance (Fritsch et al., 2007). Despite the fact that these findings are congruent with the notion of neurocognitive reserve, which posits that occupational status may provide a reserve capacity of the detrimental effects of biological aging and neural diseases on neurocognitive functions (Stern, 2002, 2009), the mechanism underlying how occupational complexity impacts the reserve system and modulates neurocognitive function is unclear thus far. One explanation is the cognitive aspect of occupational attainment. Given that executive planning and organizational skills appear to be important to maintaining functional independence, individuals who have higher levels of occupational demands may engage a variety of higher-order aspects of cognition, such as working memory, to actively maintain, manipulate, and monitor information flow during work. Alternatively, the social aspect of occupational complexity posits that a higher level of occupational demands usually couples with greater levels of social engagement in the work environment, resulting in frequent social activities, demanding processes related to emotional regulation, and controlled retrieval of long-term memory, all of which have been identified to be important determinants of later-life cognitive functioning.

3.2.2 Indicators of NCR: Cognitively stimulating leisure activities

Staying engaged in cognitively stimulating leisure activities has been reported to be associated with better cognitive functioning and a reduced risk of developing Alzheimer's disease in late life (Hultsch, Hertzog, Small, & Dixon, 1999; Shimamura, Berry, Mangels, Rusting, & Jurica, 1995), suggesting that sustained practice in mental activity (e.g., reading books, attending a play, doing volunteer work, and/or engaging in social interaction) can protect against age-related cognitive declines, attenuate neural dysfunction, and postpone the onset of dementia (Chan, Haber, Drew, & Park, 2014; Carlson et al., 2008, 2009). For example, in a multi-modal, cognitively-stimulating activity program, conducted in a community-based setting, Carlson et al. (2008) randomized 149 older adults into the program and control groups. The participants who were assigned to the program team were instructed to participate in a social service program, consisting of actively working with elementary school children on reading achievement, library support, and classroom behavior for at least 15 h per week during an academic year (4–8 months). The neuropsychological assessment included measurements of memory, executive function, and psychomotor speed at baseline and follow-up phases to examine the effects of the activity program. This demonstrated that older individuals with low income, low education and at high risk for accelerated development of cognitive impairment showed significantly improved executive function and memory performance, suggesting the beneficial effects of short-term cognitively stimulating activities on cognitive function in at least some populations of older adults. Moreover, Carlson et al. (2009) used a follow-up fMRI study to investigate whether such activities could yield functional changes in the aging brain. They demonstrated that the older adults who actively participated in this short-term cognitively stimulating social service showed increased brain activity in the left prefrontal cortex and anterior cingulate cortex over the 6-month program interval (Carlson et al., 2009).

These behavioral and neuroimaging findings support the hypothesis that intellectual and social engagement that requires cognitive effort has long-term benefits for older adults, and, more importantly, provides promising evidence of neurocognitive plasticity in later adulthood. Moreover, recent behavioral studies examining the association between cognitively stimulating leisure activities as a proxy measure of NCR and cognitive performance with age have demonstrated a moderate correlation with executive function (Eskes et al., 2010; Lin, Friedman, Quinn, Chen, & Mapstone, 2012) but non-significant correlations with memory (Lin et al., 2012; Murphy, & O ’ Leary, 2010). These results imply that cognitively stimulating leisure activities as one of the Indicators of NCR may play a more domain-specific role in cognitive function.

3.2.3 Is bilingualism an indicator of NCR?

In addition to the typical protective factors discussed above, a number of researchers have suggested that lifelong language experiences (e.g., bilingualism, literacy, and verbal fluency, etc.) could also be protective factors and serve as an indicator of NCR, mitigating against age-related neurocognitive decline in normal aging (Grant, Dennis, & Li, 2014). There is evidence that bilinguals develop a higher-level control mechanism that allows them to function in one language at a time, while still maintaining the ability to switch between languages (Lee & Tzeng, 2016). Different language systems present sharp contrasts along several dimensions, including orthography, phonology, semantics, and grammar. For example, in many Indo-European languages, nouns are marked for gender, number, and definiteness, while verbs are marked for aspect, tense, and number. In Chinese, most of these grammatical markers for nouns and verbs are absent. Another example is that Chinese has a more flexible word order compared with English, which is more rigidly fixed in its SVO order. As a result, word order variation in Mandarin Chinese poses a great challenge to English learners who learn Chinese as a second language.

Bilingualism may play an important role at older ages, potentially protecting against age-associated cognitive decline. Research with older adults has shown that, in older adulthood, lifelong bilinguals' neurocognitive abilities are advantaged relative to their monolingual counterparts (Gold, 2015; Gold, Johnson, & Powell, 2013; Schweizer, Ware, Fischer, Craik, & Bialystok, 2012). The simultaneous use of two different languages has also been demonstrated to be associated with functional brain changes and different connectivity patterns. It has been shown, for example, that bilingual older adults have higher white matter integrity (Luk, Bialystok, Craik, & Grady, 2011) and gray matter intensity (Mechelli et al., 2004) than monolinguals. Furthermore, bilinguals also showed a 4- to 5-year delay in the onset of symptoms of Mild Cognitive Impairment (MCI) and Alzheimer's disease (AD) compared to monolinguals (Bialystok et al., 2007; Craik, Bialystok, & Freedman, 2010). This could be due to the constant engagement of the executive control system to monitor, select, and switch between multiple languages across the lifespan. However, precisely how bilingual experiences contribute to neurocognitive reserve at the neural level is still poorly understood thus far.

Beyond bilingualism in particular, it has been suggested that literacy and/or verbal fluency may be better indicators of cognitive reserve than educational attainment (Manly, Schupf, Tang, & Stern, 2005; Manly, Touradji, Tang, & Stern, 2003). Research has uncovered important individual differences among older adults in their patterns of neural recruitment and use of language comprehension processes associated with literacy acquisition and verbal fluency. For example, higher verbal fluency (rapid cued generation) predicts increased likelihood of young-like response patterns during comprehension (Federmeier et al., 2010). Taken together, language abilities continue to reflect cerebral changes throughout a lifespan, and are important indicators of cognitive health for the elderly.

3.3.1 Neural reserve: Structural scaffolding of cognition with age

Neural reserve refers to existing brain networks and brain structures that are more efficient, with greater capacity for coping with brain damage or increased task demands. Individuals with higher levels of NCR (e.g., higher educational level) show more efficient brain networks or greater brain capacity and tend to be less susceptible to disruption. Therefore, the concept of neural reserve provides a theoretical way to explore how NCR influences age-related and individual differences in brain structure and brain network through advanced structural neuroimaging techniques, such as anatomical MRI, diffusion MRI, Computerized Tomography (CT), etc. Individuals with higher levels of NCR should be able to better cope with age-related changes, resulting in less cognitive decline and reduced incidence of dementia with age. Here, we propose that the mechanism of neural reserve could be a structural scaffolding of neurocognitive reserve to protect against the impact of brain anatomical declines (e.g., brain volume, gray-matter density, neural myelination) and shrinkage.

We are aware that neural reserve may be shaped by an individual's innate capacity and life experience (Barulli & Stern, 2013; Stern, 2009). Genetic and environmental factors may afford some individuals with more efficient networks (such that they require less activity to achieve the same behavioral performance, due to intact brain structure), higher capacity networks (such that they have more resources available to recruit as task demands increase), or greater flexibility in brain network selection in response to cognitive challenges.

3.3.2 Neural compensation: Functional scaffolding of cognition with age

On the other hand, the mechanism of neural compensation refers to alterations in cognitive processing and compensatory brain function in response to cognitive challenges and/or brain pathology. Here, we propose that neural compensation is a functional scaffolding of neurocognitive reserve that contrasts with the structural scaffolding of neurocognitive reserve (i.e., neural reserve). We should note that both mechanisms are complimentary and testable if we employ different neuroimaging techniques. Neural compensation could be observed by measuring functional recruitment of neural resources (e.g., fMRI) in response to age-related declines in brain structure, including gray matter shrinkage, white matter reduction, or connectivity changes, which, in turn, could be quantified by measuring brain volume, gray-matter density, and neural myelination.

There is a wealth of literature showing that the functional brain ages in a dynamic way, declining in some respects but maintaining the ability to engage adaptive neural functions even in advanced age (Dennis & Cabeza, 2008; Huang, Meyer, & Federmeier, 2012; Huang, Polk, Goh, & Park, 2012; Park & Reuter-Lorenz, 2009). Thus, neural compensation is a system that is actively engaged to maintain or improve cognitive performance to accomplish cognitive tasks through the recruitment of additional, distributed, and compensatory neural networks. Previous studies have demonstrated that healthy older adults with higher levels of NCR are more able to recruit neural resources and additional, alternative neural networks in the face of brain changes due to normal aging or disease-related neurodegeneration in pathological aging (Alexander et al., 1997; Perneczky et al., 2006). These findings suggest that those elderly populations with higher levels of reserve are more capable of recruiting alternative and/or additional brain networks in response to cognitive changes after pathological aging, especially within the prefrontal regions (Korsnes & Ulstein, 2014; Scarmeas & Stern, 2004; Stern, Gazes, et al., 2018), suggesting a reserve-associated neural compensatory mechanism for age-related neurological disorders.

We are aware that a few hypotheses regarding neural compensatory processing/mechanism have been proposed in the cognitive and clinical neuroscience of aging, as well as in neuro-rehabilitation (e.g., stroke). We will next review some of the hypotheses that take this into account and that may thus relate to neurocognitive reserve.

3.3.3 Neural compensation, the CRUNCH hypothesis, and the GOLDEN view of the aging brain

The Compensation-Related Utilization of Neural Circuits Hypothesis (CRUNCH) was proposed to explain how compensatory mechanisms are modulated by different levels of task demands (Reuter-Lorenz & Cappell, 2008). The model postulates that, due to age-related decline in neural processing and neural efficiency, older adults will recruit greater levels of neural activity, while maintaining task performance, at low levels of task demands, but will show age-related decreases in neural activity at higher levels of task demand, where older adults' performance will also suffer because a resource ceiling has been reached (Reuter-Lorenz & Cappell, 2008; Reuter-Lorenz & Lustig, 2005). Moreover, CRUNCH suggests that, compared to young adults, older adults have limited functional processing resources and reach their limitation of processing earlier, resulting in a less flexible modulation of neural activity in response to increasing task demands. This notion is consistent with the concept of neurocognitive reserve, which predicts that older adults with lower level of reserve (i.e., limited functional processing resources in CRUNCH) would have a reduced capacity/efficiency of their neural reserve, resulting in less neural compensatory activation in response to increasing task demands. In addition to explaining age-related patterns, the pattern of CRUNCH could also be observed at the level of individual differences. Therefore, processing inefficiency would cause low performing individuals (whether they be young, middle-aged, or older adults) to recruit additional neural resources at low levels of task demand. As the level of task demand became higher, the high performing individuals would also need to engage additional neural resources to achieve the task goal, possibly resulting in additional, more distributed, compensatory activity even in young individuals.

Previous fMRI studies demonstrated that individual differences in working memory capacity are an important predictor of individual differences in fronto-parietal activation. For example, Schneider-Garces et al. (2010) utilized a modified version of the Sternberg memory task with five levels of working memory set size (2–6 letters in the memory set). They demonstrated that, averaged across memory set size, older adults showed poorer performance and greater neural activation in fronto-parietal regions than young adults. In addition, when performing trend analyses separately at low (set size 2–4) and high (set size 4–6) levels of memory demand, young adults were able to increase prefrontal activity as the set size increased whereas older adults were engaging more prefrontal activity at lower set sizes, which did not further increase, and even decreased, at higher set sizes. However, further analysis demonstrated that once individual differences in performance (i.e., subjective levels of task demands) were accounted for, older and young adults showed similar patterns of brain activation (Schneider-Garces et al., 2010). These results provide support for the idea that age-related functional activation in neural efficiency and capacity could stem from individual differences in working memory capacity.

It should be noted here that the findings described above (i.e., Schneider-Garces et al., 2010) also support the idea of the GOLDEN aging framework advanced by Fabiani (2012). The Growing of Life Differences Explains Normal Aging (GOLDEN) model was proposed to interpret the behavioral and neuroimaging results associated with normal aging processes. GOLDEN emphasizes that normal aging represents maturational processes resulting in progressive shifts in the distribution of mental abilities and brain activation with age (Fabiani, 2012), including reductions in working memory capacity, impairment of inhibitory control, and an increase in prefrontal activation. Moreover, GOLDEN suggests that age-related neural efficiency and capacity might stem from individual differences in working memory capacity. Given the neurocognitive reserve hypothesis' suggestion that age-related and individual differences in the efficiency and capacity of the neurocognitive network are associated with individual variations in life exposure (e.g., educational and occupational attainment, engagement in cognitively stimulating leisure activities, and bilingualism), the GOLDEN model seems to be congruent with the view of neurocognitive reserve that normal healthy aging involves the continuation of processes that are already present earlier in life and that continually shape the function and structure of the brain over time (Barulli & Stern, 2013; Martins, Joanette, & Monchi, 2015). The GOLDEN model, however, is more conservative in predicting cognitive as well as neural changes when neurodegenerative disease impacts cognitive impairments beyond normal aging processes.

3.3.4 Limitations of the reserve hypothesis

Despite the fact that the neurocognitive reserve hypothesis provides a theoretical account for the disjunction between the degree of brain damage/pathology and clinical manifestations of age-related and individual differences in brain structure and function, there are still some limitations and unanswered questions in this model. First, it is hard to quantify and weigh different NCR indicators. Several factors are attributed to the reserve, including educational level, occupational complexity, and leisure activity. However, these factors might involve different life experiences that are hard to quantify. For example, when examining the impact of education on lifetime cognitive abilities, two participants with identical educational levels may not have equivalent levels of behavioral performance, brain reserve, and neural compensation because each individual shows diverse learning curves and test performance in different disciplinary areas (mathematics, science, and literature). Moreover, it could be even more difficult to quantify the effects of education when different educational settings and countries have to be compared. Second, there is no distinct and clear definitions for these specific measures. For instance, leisure activity is overly simplified and includes almost all social and human activities. According to Stern (2009), leisure activity includes “knitting or music or other hobby, walking for pleasure or excursion, visiting friends or relatives, being visited by relatives or friends, physical conditioning, going to movies or restaurants or sporting events, reading magazines or newspapers or books, watching television or listening to the radio, doing unpaid community volunteer work, playing cards or games or bingo, going to a club or center, going to classes, and going to church or synagogue or temple.” Therefore, it becomes even harder to dissociate the influences of different life experiences and their contributions to the reserve. Finally, the process of the reserve is still unknown. How does a person's life experiences and lifestyle “store up” his or her brain or cognitive reserve? The two components of reserve (including neural reserve and neural compensation) seem like post hoc observation of the outcomes of epidemiological and brain imaging studies. These components show how the aging brain copes with brain damage, but provide no clues about the progress of the reserve. For example, how do our life experiences shape our brain network and cognitive functioning and determine the degree of the reserve? What are the moderating factors for developing neural reserve and neural compensation? What is the interaction between the brain reserve and cognitive reserve?

Moreover, there is thus far no clear evidence from longitudinal investigations of reserve and brain atrophy and neural compensation about premorbid intelligence, education, and the rate of structural changes.