Dissociable Neural Mechanisms of Cognition and Wellbeing in Youth versus Healthy Aging

Abstract

Mental health, cognition and their underlying neural processes in healthy aging are rarely studied simultaneously. Here, in a sample of healthy younger (n=62) and older (n=54) adults, we compared subjective mental health as well as objective global cognition across several core cognitive domains, with simultaneous electroencephalography (EEG). We found significantly greater symptoms of anxiety, depression, and loneliness in youth and in contrast, greater mental wellbeing in older adults. Yet, global performance across core cognitive domains was significantly worse in older adults. EEG-based source imaging of global cognitive task-evoked processing showed reduced suppression of activity in the anterior medial prefrontal default mode network (DMN) region in older adults relative to youth. Global cognitive performance efficiency was predicted by greater activity in the right dorsolateral prefrontal cortex in younger adults, and in contrast, by greater activity in right inferior frontal cortex in older adults. Furthermore, greater mental wellbeing in older adults related to lesser global task-evoked activity in the posterior DMN. Overall, these results suggest dissociated neural mechanisms underlying global cognition and mental wellbeing in youth versus healthy aging.

Introduction

Cognitive control is a dynamic ability of the human brain requiring multiple interacting mental operations. These operations fundamentally include attentive stimulus encoding (Badre, 2011; Hillyard & Anllo-Vento, 1998; Luna et al., 2015), online maintenance of goal-relevant information (Gazzaley & Nobre, 2012), suppression of competing goal-irrelevant distractions (Mishra et al., 2013), and continuous evaluation of the accuracy of selected actions based on feedback (Posner & Rothbart, 2009; van Noordt & Segalowitz, 2012). While it is well-evidenced that cognitive abilities decline even with healthy aging (Craik & Salthouse, 2000; A Gazzaley, 2013; Harada et al., 2013; Hedden & Gabrieli, 2004; Jackson & Owsley, 2003) and prominent changes occur in frontal cortex with aging (Zanto & Gazzaley, 2019), most studies do not simultaneously investigate mental wellbeing and neuro-cognitive processes.

In contrast to cognitive decline, mental wellbeing has been shown to improve with aging at least in wealthy English speaking countries, wherein a reduction of worry and stress has been observed with age (Steptoe et al., 2015). The World Health Organization defines wellbeing as stemming from positive mental health - a state ‘which allows individuals to realize their abilities, cope with the normal stresses of life, work productively and fruitfully, and make a contribution to their community’ (World Health Organization, 2004). This research is important given that greater wellbeing is associated with greater survival rates even when accounting for demographic differences. Notably, loneliness, anxiety, and depression, markers of mental ill-being, have all been evidenced to be greater amongst younger relative to older adults (Child & Lawton, 2017; Klap et al., 2003; Thomas et al., 2016). Further, cognition and wellbeing are related in that greater cognitive reserve in aging has been associated with greater wellbeing (Pettigrew & Soldan, 2019). At the same time, cognitive health alone does not convey successful aging, the latter being determined by several dimensions including functional, psychological, social, spiritual or environmental factors (Martineau & Plard, 2018). Overall, more research is needed at the intersection of cognition and wellbeing in aging, particularly with regards to whether these constructs have shared or dissociable neural correlates.

Evidence from neuroimaging studies suggests that age-related cognitive decline is associated with neurodegeneration of grey matter, white matter, and associated loss of connectivity between brain regions, which together contribute to deficits in information processing (Achard & Bullmore, 2007; Hong et al., 2015; Saykin et al., 2006). This neural loss has been observed in the frontal cortex as well as in cingulate cortex in older adults, underlying deficits in executive functioning in aging (Resnick et al., 2003; Van Petten et al., 2004). We also acknowledge the many cellular and molecular changes that occur with aging, but which are beyond the scope of this study (Khan et al., 2017; K. H. Wagner et al., 2016). Specifically, using functional neuroimaging methods, i.e., functional magnetic resonance imaging (fMRI) and electroencephalography (EEG), healthy aging has been associated with neuro-cognitive plasticity, i.e., learning abilities are intact in aging and there is some redistribution of neural network activations to different brain regions in older adults to support the same cognitive functions as observed in younger adults (Berry et al., 2010; Goh, 2011; Mishra et al., 2015; Mishra et al., 2014). In this regard, EEG particularly serves as an important tool to measure synaptic plasticity and directly assess brain network oscillations and corticocortical information transmission at millisecond time scales, distinct from the slower blood oxygen level dependent fMRI signals (Buzsáki & Draguhn, 2004; Cook & Leuchter, 1996; Massimini et al., 2009; Pascual-Leone et al., 2011).

At present, there is a lack of consensus regarding age-related neuroplasticity across different cognitive operations. Many cognitive neuroscience studies support no regional differences with aging in attention and working memory tasks, but simply decreased activations in the same frontoparietal cortical networks (Campbell et al., 2012; C. L. Grady et al., 1995; Kehoe et al., 2013; Pardo et al., 2007; Rypma & D’Esposito, 2000). Other cognitive studies report that older adults recruit greater neural activations to perform at par with younger adults, making their overall neural functioning less efficient (Achard & Bullmore, 2007; Berry et al., 2010). There is also accumulating evidence for compensatory neural processing in older adults (Cabeza et al., 2002; Hakun et al., 2015; Nashiro et al., 2012; Ohsugi et al., 2013; Riis et al., 2008). As per the compensation-related utilization of neural circuits hypothesis (CRUNCH) (Cappell et al., 2010; Reuter-Lorenz & Cappell, 2008; Schneider-Garces et al., 2010) declining neural efficiency in older adults is evidenced by greater neural circuit engagement than in young adults to meet task demands; hence, age-related over-activation is observed at least at low task loads, but under-activation and performance decline is evident at higher task loads. As such there is consensus that there are three interacting mechanisms at play in cognitive aging, including reserve, maintenance and compensation (Cabeza et al., 2018, 2019; Stern et al., 2019). Yet, most prior studies have focused on investigating neural activations in the context of single cognitive operations. More work is needed to understand global neuro-cognitive functioning across multiple cognitive domains and underlying brain networks in aging.

In this study, we used a rapid neuro-cognitive assessment suite (Balasubramani et al., 2021) for evaluating multiple cognitive task domains, of attention and inhibitory control, working memory, interference processing, and emotion bias. We examined differences in cognitive performance as well as EEG-based neural activations between younger adults at peak cognitive age (Fortenbaugh et al., 2015) versus healthy older adults. Additionally, we sampled subjective mental health in all study participants, specifically symptoms of the most prevalent neuropsychiatric conditions of anxiety and depression along with self-reports of loneliness and overall mental wellbeing. Hence, the primary objective of this study was to better understand global neuro-cognitive functioning in healthy aging in the context of subjective mental health differences with age.

Given the crucial role of brain networks in cognitive processing (Corbetta & Shulman, 2002; Dosenbach et al., 2006; Menon & Uddin, 2010; Power et al., 2011), here, we also leveraged EEG source imaging to investigate age-related processing in core frontal, parietal and cingulate network regions. EEG has high temporal resolution so that time windows with peak stimulus-evoked activity can be identified with millisecond accuracy. These peak activity windows can then be localized by EEG source imaging to known regions of interest (ROIs). EEG has low spatial resolution, yet ROI activations across subjects can be statistically tested for significance, hence, allowing direct comparison with regional activations identified in fMRI studies. Specifically, we focus on core regions important for executive control (i.e., dorsolateral prefrontal cortex: DLPFC), interference control and salient reorientation (i.e., inferior frontal cortex: IFC), spatial attention and working memory (i.e., superior parietal cortex: SPC), as well as regions of the default mode network (i.e., medial orbitofrontal cortex and rostral anterior cingulate cortex part of the anterior DMN, and posterior cingulate as the core region of the posterior DMN). We include the DMN regions in our study given that there is much evidence for age-related neural changes in this network at rest and in the context of attention and memory tasks (Andrews-Hanna et al., 2007; Chadick et al., 2014; Eudave et al., 2018; C. Grady et al., 2016; C. L. Grady et al., 2010; Lustig et al., 2003; Marstaller et al., 2015; Persson et al., 2014). By focusing on network specific regions, we can observe if we find broad evidence of frontal cortex aging or network region-specific effects (Campbell et al., 2012; Grady et al., 1995; Kehoe et al., 2013; Marstaller et al., 2015; Pardo et al., 2007; Resnick et al., 2003; Rypma & D’Esposito, 2000; Van Petten et al., 2004). We hypothesized that healthy aging will be associated with reduced performance abilities across cognitive domains tested, while wellbeing in older adults will not be impacted. We further hypothesized that neural correlates revealed by EEG will show age-related changes in early stimulus-evoked processing, and that source imaging of these task-based EEG signals may confirm fMRI study observations of compensatory mechanisms as well as reduced suppression of the DMN in older adults. An alternative hypothesis would be that cognitive performance and wellbeing are positively associated and that neural correlates that determine improved cognitive performance also underlie wellbeing.

Methods

Transparency and Openness

The de-identified data on which the study conclusions are based, the analytic code needed to reproduce analyses, and study materials are available, and the link to access this information is provided in the Author Note.

Participants

125 human subjects participated in the study, including 62 healthy young adults (mean age: 20.9 ± 0.22, range: 18–25 years, 39 females) and 54 healthy older adults (mean age: 70.7 ± 0.86, range: 60–88 years, 34 females); 9 participants (5 younger, 4 older) were excluded due to self-reported current diagnosis of mental illness and/or current medications for a neuropsychiatric disorder. Racial distribution of the sample is reported in Table 1. For all older adults, participants were confirmed to have a Mini-Mental State Examination (MMSE) score >26 to verify absence of apparent cognitive impairment (Arevalo-Rodriguez et al., 2015); no other IQ assessments were performed. All participants had normal/corrected-to-normal vision and hearing and no participant reported color blindness. Majority of participants (107 of 116) were right-handed. All participants had at least high-school education. All participants provided written informed consent for the study protocol approved by the University of California San Diego institutional review board (UCSD IRB #180140). All data were collected in the year prior to COVID-19 research restrictions placed in Spring 2020.

Table 1.

Demographic and mental health characteristics for younger (n = 62) and older (n=54) adult study participants

Demographics	Younger Adults	Older Adults	p-value
	Median ± mad	Median ± mad
Age	21 ± 1.36	69.50 ± 5.16
Gender n (%)			0.944
Male	23 (37.1)	20 (37.0)
Female	39 (62.9)	34 (63.0)
Race n (%)			<0.0001
Caucasian	21 (33.9)	53 (98.2)
Black/African American	0 (0.0)	1 (1.8)
Asian	25 (40.3)	0 (0.0)
Native American	2 (3.2)	0 (0.0)
Other	14 (22.6)	0 (0.0)
SES	5.0 ± 1.33	6.0 ± 1.21	0.061
Mental Health	Median ± mad	Median ± mad	Effect Size, p-value
Anxiety	3.0 ± 2.75	1.0 ± 2.41	0.77, <0.0001
Depression	3.0 ± 2.71	1.0 ± 1.96	0.84, 0.002
Loneliness	43.0 ± 8.90	34.0 ± 9.64	0.97, 0.005
Mental Wellbeing	25.0 ± 3.08	30.0 ± 3.32	−1.57, <0.0001

Note. Variables of gender and race were compared across age groups using χ² (Chi-Square) statistics. SES and all mental health variables were compared across age using Mann-Whitney U tests. Effect sizes for mental health comparisons are Cohen’s d equivalents with negative values denoting higher scores in older adults. mad: median absolute deviation.

Sample Size and Power

Our sample size was adequately powered to detect medium effect size group differences (Cohen’s d>0.5) at beta of 0.8 and alpha significance level of 0.05 as calculated using the G*Power software (Cumming, 2014; Faul et al., 2009). Of note, recent meta-analyses have shown that younger vs. older adult cognitive differences have medium-to-large effect sizes (Vallesi et al., 2021).

Demographics and Mental Health Assessments

All participants provided demographic information by self-report including, age, gender, race and ethnicity. Socio-economic status was measured on the Family Affluence Scale (Boudreau & Poulin, 2009); this scale measures individual wealth based on ownership of objects of value (e.g., car/computer) and produces a composite score ranging from 0 (low affluence) to 9 (high affluence).

All participants completed subjective mental health self-reports using standard instruments, including anxiety ratings on the 7-item Generalized Anxiety Disorder scale: GAD-7 (Spitzer et al., 2006), depression ratings on the 9-item Patient Health Questionnaire: PHQ-9 (Kroenke et al., 2001), loneliness ratings on the 20-item UCLA version-3 Loneliness scale (Russell, 1996), and overall mental wellbeing on the 7-item Shortened Warwick-Edinburgh mental wellbeing scale: SWEMWBS (Tennant et al., 2007). We calculated reliability for each of these standard self-reports instruments in our sample using the Cronbach’s alpha intra-class correlation measure.

Neuro-cognitive Assessments

Each participant made a single study visit at the Neural Engineering and Translational Labs (NEAT Labs) located at the Altman Clinical and Translational Research Institute at the University of California San Diego. Cognitive assessments were deployed on the BrainE platform with simultaneous EEG (Balasubramani et al., 2021), delivered on a Windows-10 laptop at a comfortable viewing distance. The Lab Streaming Layer (LSL) protocol was used to time-stamp all stimuli and response events in all cognitive assessments (Kothe et al., 2019). Each cognitive assessment session lasted ~40 minutes, and consisted of four different cognitive tasks plus rest. Figure 1 shows the stimulus sequence in each of the four cognitive tasks. All four cognitive tasks had a standard trial structure of 500 ms central fixation “+” cue followed by task-specific stimulus presented for task-specific duration and with a task-specific response window. All stimuli were presented in a shuffled order across trials. Response in every task trial was followed by standard response feedback for accuracy as a smiley or sad face emoticon, presented 200 ms post-response for 200 ms duration, followed by a 500 ms inter-trial interval (ITI). At the end of each task block, participants received a percent block accuracy score with a series of happy face emoticons (up to 10) to promote engagement.

Figure 1. Cognitive assessments delivered on the BrainE platform.

Note. (A) Assessment dashboard with the wireless EEG recording setup. Individual task schematics are shown in (B) inhibitory control task, (C) Flanker interference processing task, (D) working memory task with perceptually thresholded stimuli, and (E) emotion bias task that presented neutral, happy, sad or angry faces. The individual whose face appears in (A) gave signed consent for their likeness to be published in this article. Photo in (E) is adapted from Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M., Hare, T. A., Marcus, D. J., Westerlund, A., Casey, B. J., & Nelson, C. (2009). The NimStim set of facial expressions: Judgments from untrained research participants. Psychiatry Research, 168(3), 242–249. https://doi.org/10.1016/j.psychres.2008.05.006

1. Inhibitory Control

Participants accessed a game-like task, “Go Wait” (Fakhraei et al., 2021; Shah et al., 2021). The basic task framework was modeled after the standard test of variables of attention (Greenberg & Waldman, 1993). In this two-block task, visual stimuli of colored rockets appeared in either the upper or lower central visual field. Post-fixation cue, a rocket stimulus of either blue target color or other iso-luminant nontarget color (brown, mauve, pink, purple, teal) was presented for 100 ms. For blue rocket targets, participants were instructed to press the spacebar on the laptop keyboard as quickly as possible (“go” trials). For non-target color rockets, participants withheld their response until the fixation “+” cue flashed briefly on the screen, at 2 sec post-stimulus for 100 ms duration (“wait” trials). Thus, participants were required to be flexible in their responses based on the stimulus cues. Both task blocks lasted 5 minutes and consisted of 90 trials per block with 30/60 target/nontarget ratio in block 1 and 60/30 ratio in block 2 (blue rockets). Four practice trials preceded the first task block. Total task time was 10 min.

2. Interference Processing

Participants accessed the game-like task, “Middle Fish”, which was an adaptation of the Flanker assessment (Eriksen & Eriksen, 1974; Lavie et al., 2004; Shipstead et al., 2012). Post-fixation cue on each trial, participants viewed an array of fish presented either in the upper or lower central visual field for 100 ms. On each trial, participants had a 1 sec response window to detect the direction of the middle fish (left or right) while ignoring the flanking distractor fish that were either congruent or incongruent to the middle fish, i.e., faced the same or opposite direction to the middle fish. 50% of task trials had congruent distractors and 50% were incongruent. A brief practice of 4-trials preceded the main task of 96 trials presented over two blocks for a total task time of 8 min.

3. Working Memory

Participants accessed a game-like task, “Lost Star”, which was based on the visuo-spatial Sternberg task (Sternberg, 1966). Post-fixation cue on each trial, participants viewed a spatially distributed test array of objects (i.e., a set of blue stars) for 1 sec. Participants were required to maintain the locations of these stars for a 3 sec delay period, utilizing their working memory. A probe object (a single green star of 1 sec duration) was then presented in either the same spot as one of the original test stars, or in a different spot than any of the original test stars. Participants were instructed to respond whether or not the probe star had the same or different location as one of the test stars. We implemented this task at the threshold perceptual span for each individual, which was defined by the number of test star objects that the individual could correctly encode without any working memory delay. For this, a brief perceptual thresholding period preceded the main working memory task, allowing for equivalent perceptual load to be investigated across participants (Lavie et al., 2004). During thresholding, the set size of test stars increased progressively from 1–8 stars based on accurate performance where 100% accuracy led to an increment in set size; <100% performance led to one 4-trial repeat of the same set size and any further inaccurate performance aborted the thresholding phase. The final set size at which 100% accuracy was obtained was designated as the individual’s perceptual threshold. Post-thresholding, the working memory task consisted of 48 trials presented over 2 blocks (Lenartowicz et al., 2014) with total task duration of 6 min.

4. Emotion Bias

Participants accessed the game-like assessment, “Face Off”, adapted from studies of attentional bias in emotional contexts (Grennan et al., 2021; S. López-Martín et al., 2015; Sara López-Martín et al., 2013). The task integrated a standardized set of culturally diverse faces from the NimStim database (Tottenham et al., 2009). We used an equivalent number of male and female faces, each face with four sets of emotions, either neutral, positive (happy), negative (sad) or threatening (angry), presented on equivalent number of trials in each task block. Post-fixation cue on each trial, participants viewed an emotional face with a superimposed arrow of 300 ms duration. The arrow occurred in either the upper or lower central visual field on equal number of trials, and participants responded to the direction of the arrow (left/right) within an ensuing 1 sec response window. Participants completed 144 trials presented over three equipartitioned blocks; a practice set of 4-trials preceded the main task. The total task duration was 10 min.

Resting State

Eyes closed resting state data were acquired for 3 min duration. For the purpose and scope of this study, analyses are focused on the cognitive task-related data, not the resting state data.

Electroencephalography (EEG)

EEG data were collected in conjunction with all cognitive tasks using a 24-channel semi-dry and wireless electrode cap and SMARTING™ amplifier. Signals were acquired at 500 Hz sampling frequency at 24-bit resolution. The LSL protocol was used to time-stamp EEG markers and integrate cognitive markers (Kothe et al., 2019), and files were stored in xdf format.

Behavioral Analyses

Behavioral data for all cognitive tasks were analyzed for signal detection sensitivity, d’, computed as z(Hits)-z(False Alarms) (Heeger & Landy, 2009); all d’ values were divided by max theoretical d’ of 4.65 to obtain scaled-d’ in the 0–1 range. Task speeds were calculated as log(1/RT), where RT is response time in seconds. Task efficiency was calculated as a product of d’ and speed (Barlow, 1980; Vandierendonck, 2017). d’, speed, and efficiency metrics were checked for normal distributions prior to statistical analyses. Global task efficiency (average of all four cognitive task efficiencies) was used in combined behavioral/neural analyses. All analyses were done after behavioral outlier removal, where outliers >3 median absolute deviations (mad) from the median were removed.

Neural Analyses

A uniform processing pipeline was applied to all EEG data based on the cognitive event markers. The pipeline included 1) data pre-processing, 2) computing event related spectral perturbations (ERSP) for all channels, and 3) cortical source localization of the EEG data filtered within relevant theta, alpha, and beta frequency bands.

1. Data pre-processing utilized the EEG processing software EEGLAB toolbox in MATLAB (Delorme & Makeig, 2004). EEG data were first resampled at 250 Hz and filtered in the 1–45Hz range to exclude ultraslow DC drifts at <1Hz and high-frequency noise produced by muscle movements and external electrical sources at >45Hz. EEG data were average electrode referenced and epoched to cognitive task-relevant stimuli based on the LSL time-stamps, within the −1.0 to +1.0 sec event time window. The epoched data were then cleaned using the autorej function of EEGLAB, which automatically removed noisy trials (>5sd activity outliers rejected over max 8 iterations; 5.89 ± 4.09% of trials rejected per participant across tasks). EEG data were further cleaned by excluding signals estimated to be originating from non-brain sources, such as electrooculographic, electromyographic or unknown sources, using the Sparse Bayesian learning (SBL) algorithm (https://github.com/aojeda/PEB) explained below (Ojeda et al., 2021). Participants were excluded if their peak channel activity for any channel exceeded 5 standard deviations from global channel activity across all subjects (n = 116), thereby, excluding EEG data for three older adults; thus, 62/62 younger adult participant data and 51/54 older adult participant data were used for neural analyses.

2. For event-related potentials (ERPs) at the scalp, we analyzed the global cognitive ERPs that were averaged to stimuli presentations across all cognitive tasks. These ERPs had peak activations at 100–300 ms post-stimulus onset across all younger and older subjects, with no latency differences across groups.

3. Cortical source localization was performed to map the underlying neural source activations using the block-sparse Bayesian learning (BSBL-2S) algorithm (Ojeda et al., 2021). This is a two-step algorithm in which the first-step is equivalent to low-resolution electromagnetic tomography (LORETA) (Pascual-Marqui et al., 1994). LORETA estimates sources subject to smoothness constraints, i.e., nearby sources tend to be co-activated, which may produce source estimates with a high number of false positives that are not biologically plausible. To guard against this, BSBL-2S applies sparsity constraints in the second step wherein blocks of irrelevant sources are pruned. This data-driven sparsity constraint of the SBL method reduces the effective number of sources considered at any given time as a solution, thereby reducing the ill-posed nature of the inverse mapping. In other words, one can either increase the number of channels used to solve the ill-posed inverse problem or impose more aggressive constraints on the solution to converge on the source model when channel density is low/moderate; 24 channels in this case. The two-stage SBL has been benchmarked to produce evidence-optimized inverse source models at 0.95AUC relative to the ground truth while without the second stage <0.9AUC is obtained (Ojeda et al., 2018, 2021). Prior research also provides support that sparse source imaging constraints can be soundly applied to low channel density data (Ding & He, 2008; Stopczynski et al., 2014), and we have also shown that cortical source mapping with this method has high test-retest reliability (Balasubramani et al., 2021).

Using BSBL-2S, source space activations were estimated and the root mean square signals were partitioned into regions of interest (ROIs) and artifact sources. ROIs were based on the standard 68 brain region Desikan-Killiany (DK) atlas (Desikan et al., 2006) using the Colin-27 head model (Holmes et al., 1998). Activations from artifact sources contributing to EEG noise from non-brain sources such as electrooculographic, electromyographic or unknown sources, were removed to clean the EEG data. Cleaned subject-wise trial-averaged EEG data were then source localized in each task to estimate the corresponding cortical signals. The envelope of source signals was computed in MATLAB (envelop function) by a spline interpolation over the local maxima separated by at least one-time sample. Within the source signals, our neural analyses focused on post-stimulus encoding in the 100–300 msec range. This epoch window was chosen because it encompassed the peak global activity across all-task-averaged signals across all younger and older subjects. To calculate peak global activity latencies, all task signals were averaged for each subject across all electrodes and the timing of the max ERP amplitude was calculated in the 1 sec post-stimulus epoch. The peak global activity window was the range of the peak ERP amplitude latencies across all subjects, and was fixed between younger and older adults. There were no significant differences between peak global activity latencies between younger and older adult subjects. Post-stimulus neural response data were baseline corrected using the −750 msec to −550 msec time window prior to stimulus presentation.

Statistical Analyses

Demographic characteristics were compared between younger and older adults using χ² (Chi-Square) statistics derived from non-parametric group comparisons. All mental health reports were non-parametrically distributed, hence, were compared between older and younger adults using Mann-Whitney U tests, with family-wise error rate (fwer) Bonferroni correction applied for multiple comparisons.

Cognitive performance metrics of d’, speed and efficiency were compared across tasks using repeated measures analyses of variance (rm-ANOVA) with within-subject factor of task-type and between-subjects factor of age group (younger vs. older). These analyses subsequently also used a mental health covariate, derived from principal component analysis of the mental health reports, in repeated measures analyses of covariance (rm-ANCOVAs). The Greenhouse-Geisser correction was applied to adjust for lack of sphericity.

In channel-wise ERP data as well as source-localized neural data outliers >5 standard deviations from the mean activity were removed prior to statistical analyses. In this study, we were not powered to compare neural data for each of the four tasks by age while correcting for multiple comparisons across tasks. Hence, all main analyses focused on global cognitive processing for which stimulus-locked ERPs were averaged across all four cognitive tasks. Channel-wise ERPs corresponding to global cognition (i.e., mean stimulus-evoked responses across all four cognitive tasks) were compared between younger and older adults using t-tests within the peak 100–300 ms time window since this window encompassed the peak global activity across all-task-averaged signals across all younger and older subjects. Permutation testing was applied to account for multiple comparisons across time. For this, 100 iterations of random permutations were performed across the time vector; only continuous time segments that survived significance at p<.05 using permutation testing (>97.5%le of the right tail of the random vector permutation distribution) were reported (Nichols & Holmes, 2002). ERP topographies were visualized using the EEGLAB topoplot function. Corresponding task-specific ERPs are shown in Figure S1.

For source-reconstructed data within the peak global activity time window of 100–300 ms post-stimulus averaged across all four tasks, group comparisons were confined to specific regions of interest that are critical for cognitive control; these regions included left/right DLPFC (caudal middle frontal ROI in the DK atlas), superior parietal cortex (SPC), inferior frontal cortex (IFC: inferior frontal pars opercularis ROI in DK atlas), anterior DMN (represented by mean activity of medial orbitofrontal and rostral anterior cingulate ROIs in the DK atlas), and posterior DMN (represented by isthmus cingulate ROI in DK atlas). For lateral regions of the DLPFC, IFC and SPC, we tested the left and right sided activations separately, while for the medial anterior and posterior DMN regions we combined (i.e., averaged) left/right activations given uncertainty in source reconstruction for medial localizations. False discovery rate corrections were applied for multiple comparisons in cortical source space (fdr, p ≤ 0.05).

We further investigated how the source-reconstructed global stimulus-evoked peak neural activity data predicted global cognitive performance efficiency as well as mental health. For this, robust linear regression models were generated that included predictors of source neural activity (in ROIs as specified above), age and the interaction of age by source neural activity. The dependent variables in these models (cognitive efficiency and the mental health principal component) were log transformed to obtain normal distributions. Robust regression was used to minimize outlier influence (Lane, 2002). Age by neural activity interactions observed in the regression models were followed up by Spearman’s nonparametric correlations separately within each age group. All model results were fdr-corrected for comparisons across multiple ROIs (fdr, p ≤ 0.05).

Effect sizes were calculated and reported for all parametric age-group comparisons as the Cohen’s d estimate, 0.2: small; 0.5: medium; 0.8 large (Cohen, 1988); for non-parametric variables, the equivalent effect size estimate derived from median and median absolute deviation measures was used (Ricca & Blaine, 2020). For regression models, standardized beta coefficients were reported to reflect effect sizes. The Bayes Factor was also computed for regression models using the MATLAB Bayes Factor toolbox (Krekelberg, 2021), and interpreted as Bayes Factor log₁₀(BF₁₀)<0.5: evidence against H₀ (null hypothesis) is weak; log₁₀(BF₁₀): 0.5–1: evidence against H₀ is substantial; log₁₀(BF₁₀): 1–2: evidence against H₀ is strong; log₁₀(BF₁₀)>2: evidence against H₀ is decisive (Kass & Raftery, 1995; Liang et al., 2008).

Results

Demographics and Mental Health

Comparisons of these data between healthy younger versus older adults are shown in Table 1. These age groups did not differ in gender or SES composition but differed by race (p < 0.001), with older adults showing greater composition of Caucasian participants compared to younger adults.

All standard self-report measures used to assess mental health showed high, i.e., >0.8 reliability scores (Cronbach’s alpha: anxiety: 0.85, depression: 0.82, loneliness: 0.95, wellbeing: 0.86). Comparing these by age, younger adults showed strikingly greater anxiety, depression and loneliness severity, while older adults showed greater overall mental wellbeing (Table 1).

Cognitive Performance

Performance metrics on the four cognitive tasks are reported in Table 2. For each of the four assessments of inhibitory control, interference processing, working memory and emotion bias, we report signal detection sensitivity scaled-d’, speed, response time (RT in sec), and efficiency (product of d’ and speed) measures. For the inhibitory control task, metrics in the table are reported for both go and wait stimuli. For working memory, item-span was also calculated. Repeated measures (rm-) ANOVAs were conducted on the scaled-d’, speed, RT and efficiency measures across the four tasks with age as a between-subjects factor and task as a within-subjects factor. For the inhibitory control task, go stimuli performance was taken as representative for the task in the across-task rm-ANOVAs, as this condition was similar to the fast and accurate performance demands common across tasks.

Table 2.

Behavioral performance across tasks for younger (n = 62) and older adult participants (n = 54)

Cognitive Task	Younger Adults mean ± sem	Older Adults mean ± sem	Effect Size, p-value
Inhibitory Control
scaled-d’: go	0.72 ± 0.02	0.48 ± 0.03	1.27, <0.0001
Speed: go	0.37 ± 0.01	0.29 ± 0.01	1.05, <0.0001
response time: go	0.44 ± 0.01	0.52 ± 0.01	−1.05, <0.0001
Efficiency: go	0.27 ± 0.01	0.12 ± 0.01	1.96, <0.0001
scaled-d’: wait	0.77 ± 0.01	0.67 ± 0.04	0.48, <0.0001
response time: wait	2.22 ± 0.01	2.12 ± 0.02	0.87, <0.0001
Interference Processing
scaled-d’	0.79 ± 0.02	0.73 ± 0.02	0.39, 0.03
speed	0.32 ± 0.01	0.22 ± 0.01	1.31, <0.0001
response time	0.48 ± 0.01	0.60 ± 0.01	−1.57, <0.0001
efficiency	0.25 ± 0.01	0.14 ± 0.01	1.44, <0.0001
Working Memory
scaled-d’	0.39 ± 0.02	0.47 ± 0.03	−0.42, 0.03
speed	0.37 ± 0.01	0.33 ± 0.01	0.52, 0.03
response time	0.87 ± 0.03	0.93 ± 0.02	−0.30, 0.10
efficiency	0.14 ± 0.01	0.16 ± 0.01	−0.26, 0.22
span	5.66 ± 0.30	3.06 ± 0.33	1.09, <0.0001
Emotion Bias
scaled-d’	0.76 ± 0.02	0.71 ± 0.02	0.33, 0.10
speed	0.32 ± 0.01	0.21 ± 0.01	1.44, <0.0001
response time	0.49 ± 0.01	0.62 ± 0.01	−1.70, <0.0001
efficiency	0.24 ± 0.01	0.15 ± 0.01	1.18, <0.0001

Note. Data are shown as mean ± standard error of mean (sem). Speed is log(1/RT) and RT is reported in sec. Efficiency is scaled-d’ × speed. Working memory span was tested for up to 8 items. Cohen’s d effect sizes are shown with negative values denoting higher scores in older adults. p-values represent t-tests for age comparisons; statistically significant variables are shown in bold font.

In the speed rm-ANOVA, we observed a significant main effect of age (F_1,108 = 96.83, p < 0.0001), of task (F_3,324 = 64.55, p < 0.0001), and a significant age × task interaction (F_3,324 = 7.98, p < 0.0001). Concomitantly, for response time, we also observed a significant effect of age (F_1,103 = 59.50, p < 0.0001), of task (F_3,309 = 391.80, p < 0.0001), and a significant age × task interaction (F_3,309 = 2.73, p < 0.0001). Similar results were obtained for d’, that is, a significant effect of age (F_1,97 = 5.55, p = 0.02), of task (F_3,291 = 102.28, p < 0.0001) and a significant age × task interaction (F_3,291 = 15.16, < 0.0001). The efficiency metric that is a product of d’ and speed, thereby, showed a significant effect of age (F_1,108 = 78.08, p < 0.0001), of task (F_3,324 = 14.27, p < 0.0001) and a significant age × task interaction (F_3,324 = 35.46, p < 0.0001).

Post-hoc t-tests and effect sizes for age comparisons on each task metric are shown in Table 2. Older adults had lower d’ on the inhibitory control and interference processing tasks but not on the thresholded working memory task or the emotion bias task. Speed was significantly slower on all tasks for older adults, with greater speed differences by age on the inhibitory control, interference processing and emotion bias tasks and less so on the working memory task. Notably, response times were significantly longer in older adults across all tasks when speeded response was required. For the inhibitory control task wait trials, older adults were significantly more impulsive i.e., older adults waited for less duration than younger adults, and there were no RT differences found on the working memory task. Efficiency was significantly lower for older adults on all tasks except for the perceptually thresholded working memory task. Notably, the perceptual threshold for item-span for the working memory task was significantly lower for older adults (Mann-Whitney U test, p < 0.0001), hence, if efficiency on the working memory task was weighted by item span, then younger adults had greater span-weighted efficiency (p < 0.0001).

Effects of Mental Health on Cognitive Performance

We conducted a principal component analysis (PCA) on the mental health variables of anxiety, depression, loneliness, and mental wellbeing for which the top PC accounted for 84.11% variance of factors, corresponding to greater self-reports of loneliness, anxiety, and depression, and inversely related to mental wellbeing. We used this mental health PC as a covariate in the repeated measures analyses (rm-ANCOVA) for all cognitive task efficiencies as well as for global task efficiency (average of the four task efficiencies); we used efficiency in this analysis as reflective of cognitive performance because it encompasses both d’ and speed measures. The mental health PC had no significant effects on global cognitive performance efficiency (p = 0.97) while the effect of age remained significant in these models (F_1,88 = 102.70, p < 0.0001).

Neural Activations

ERPs were compared between younger and older adults within the peak global cognitive activity time window of 100–300 msec post-stimulus with permutation corrections applied for multiple comparisons in time (see Figure 2A). Global cognitive activity was calculated as the average of stimulus-locked ERPs across all four cognitive tasks with baseline corrections applied within each task. We focused on these global cognitive signals because we were not powered to compare evoked responses separately on each of the four cognitive tasks while correcting for multiple comparisons across tasks. Figure S1 provides the corresponding age-related ERP comparisons in each task. Global ERPs in the peak stimulus-evoked activity time window showed greater frontal negativity and greater parietal positivity in younger adults relative to older adults.

Figure 2. Global cognitive processing differences in scalp and source space neural activity between younger and older adults.

Note. Differences are shown for global stimulus-locked ERPs averaged across all cognitive tasks depicted in red: younger, blue: older, and black: younger minus older difference; error bars are 95%le confidence intervals. Scalp topographies are shown on the right in (A). Significant global ERP differences in (A) were compared in the peak 100–300 ms window and permutation corrected (100 iterations) for multiple comparisons in time (p<0.05, solid black horizontal lines on top of ERP plots). (B) Global peak evoked activity in source ROIs ± standard error of mean, normalized (for display purposes) to the absolute max neural activity observed in that ROI across all subjects. Only the anterior DMN showed significant differences between young and older adults (*** t-stat = −3.415, df = 95, p = 0.0009, effect size = −0.69).

We then performed corresponding age comparisons of source-reconstructed activity (i.e., source localization of global peak evoked activity averaged in the 100–300 msec window across all tasks) within specific cognitive control regions of interest, including DLPFC, inferior frontal cortex (IFC), superior parietal cortex (SPC), anterior DMN and posterior DMN (see Figure 2B). The anterior DMN was the only region that showed significant age-related differences, with reduced activity suppression (relative to baseline) observed in this region in older adults in comparison to youth (t-stat = −3.415 ± 0.0005, df = 95, p = 0.0009, effect size = −0.69). All results were fdr-corrected for multiple comparisons across ROIs. Corresponding source activity results in each task are shown in Figure S1.

Relationships between Task-Global Neural Source Activations, Cognitive Performance and Mental Health

To understand which brain region activations contribute to (a) cognitive performance efficiency and (b) mental health, we constructed robust linear regression models separately for these two dependent variables. Independent variables in these models included age group, activity in brain regions of interest and the age group × neural activity interaction. Models were fdr-corrected for comparisons across multiple ROIs, i.e., left/right DLPFC, left/right IFC, left/right SPC, medial anterior DMN and medial posterior DMN.

We found that global cognitive efficiency was predicted by neural activity by age interactions in specific brain regions (see Figure 3A–C): (1) right DLPFC (age interaction β = 0.33 ± 0.11, p = 0.0043; main effect β = −0.25 ± 0.10, p = 0.0096; model R² = 0.54, log₁₀(BF₁₀) = 12.35); (2) right IFC (age interaction β = −0.93 ± 0.26, p = 0.0006, main effect β = 0.95 ± 0.25, p = 0.0003, model R² = 0.54, log₁₀(BF₁₀) = 12.87); and (3) left SPC that tended towards but did not reach significance (age interaction β = 0.22 ± 0.13, p = 0.088, main effect β = −0.22 ± 0.11, p = 0.0435, model R² = 0.5, log₁₀(BF₁₀) = 11.09).

Figure 3. Regional brain activity, specifically, source-localized global peak evoked activity, showing significant relationships with global cognitive performance efficiency (A-C) and the mental health principal component (D).

Note. Regions include (A) right DLPFC where younger but not older adults show a positive neural relationship with efficiency (age interaction β = 0.33 ± 0.11, p = 0.0043), (B) right IFC where older but not younger adults show a positive neural relationship with efficiency (age interaction β = −0.93 ± 0.26, p = 0.0006), and (C) left SPC where older adults tend to have a negative neural relationship with efficiency (age interaction β = 0.22 ± 0.13, p = 0.088). (D) The mental health PC is predicted by a neural activity by age interaction in the posterior DMN where older adults show a positive neural relationship with mental health symptoms (age interaction β = −0.43 ± 0.21, p = 0.043). Brain regions are demarcated in red or blue depending on whether younger or older adults drive the regression results.

The right DLPFC × age interaction was driven by a significant positive relationship with global efficiency in younger adults but not older adults (right DLPFC activity vs. global efficiency correlation - younger: Spearman’s rho(62) = 0.30, p = 0.03, older: rho(51) = −0.04, p = 0.80). In contrast, the right IFC × age interaction was driven by a significant positive relation with global efficiency in older adults (right IFC activity vs. global efficiency correlation - younger: rho(62) = 0.09, p = 0.53, older: rho(51)= 0.33, p = 0.03).

Finally, the robust regression models for the principal component of mental health only found an effect of activity localized to the posterior DMN with a significant age interaction driven by older adults (see Figure 3D, age interaction β = −0.43 ± 0.21, p = 0.043; main effect β = 0.21 ± 0.11, p = 0.063; model R² = 0.16, log₁₀(BF₁₀) = 0.229; younger rho(62) = −0.25, p = 0.14; older rho(51) = 0.34, p = 0.03). Thus, older adults with greater posterior DMN activity during global cognitive processing, showed worse mental health symptoms.

Discussion

In this study, we investigated cognition, mental health and EEG-derived neural activations in 62 younger (age 18–25 years) and 54 older (age 60–88 years) healthy adults. Participants showed significant age-related differences in their subjective mental health, specifically greater anxiety, depression, and loneliness in younger adults contrasted by greater mental wellbeing in older adults. We measured cognitive processing in four fundamental task domains – inhibitory control, interference processing, working memory, and emotion bias. Younger adults showed significantly greater cognitive performance across tasks relative to older adults, with greater accuracy on the inhibitory control, interference processing, and working memory (WM) tasks, and faster response speed across all task stimuli where speeded responses were required. For the visuo-spatial WM task, item span was also greater in younger adults. In neural activity data analyzed in core cognitive control regions, we found that healthy older adults significantly differed from younger adults only in anterior DMN activity. When neural data were regressed with cognitive performance and mental health, we found distinct regional contributions of right DLPFC, right IFC and left SPC activity to cognition, and posterior DMN to mental health symptom severity.

All of the cognitive tasks we implemented had time-limited trials and hence, required rapid decision-making. Overall, our cognitive results support substantial previous literature showing that older adults have particularly slower response speed during goal-relevant stimulus processing relative to younger adults (Harada et al., 2013; Hedden & Gabrieli, 2004; Mishra et al., 2013; Zysset et al., 2007). Our WM task results are supported by prior studies, which show that older adults have difficulty processing high working memory loads (Kirova et al., 2015; Salthouse, 1994; Wang et al., 2011). Our results on subjective mental health align with recent research showing reports of greater loneliness, anxiety, and depression amongst younger relative to older adults (Child & Lawton, 2017; Klap et al., 2003; Thomas et al., 2016). Notably, recent analyses of data during the COVID-19 pandemic as a contextual stressor, also find greater wellbeing in older relative to younger adults (Carstensen et al., 2020; Klaiber et al., 2021; Vahia et al., 2020), although all of our data were collected pre-pandemic. Thus, we find that healthy aging in our sample is significantly associated with enhanced wellbeing even though cognitive performance is impacted with aging. Interestingly, we also note that subjective mental health was not a significant covariate in age-related cognitive performance analyses, so subjective mental health and cognition can be dissociable in aging.

While prior studies have usually focused on neural recordings in one specific cognitive domain, here we analyzed global stimulus-evoked signals that were acquired across all four cognitive tasks in order to focus on more generalizable neural mechanisms. Data were acquired using the BrainE neuro-cognitive platform that we have shown to be reliable and convenient for such data collection (Balasubramani et al., 2021). To allow for comparison with the literature on brain networks in aging, we analyzed source-reconstructed EEG data confined to specific cortical regions that are important for cognition. These regions included the dorsolateral prefrontal cortex that has a crucial role in executive control (Aizenstein et al., 2009; Dosenbach et al., 2006; Menon & Uddin, 2010; A. D. Wagner et al., 2001), the inferior frontal cortex that is involved in interference processing and salient reorientation (Brass et al., 2005; Corbetta & Shulman, 2002; Mishra et al., 2014; Zanto et al., 2011), the superior parietal cortex that is important for spatial attention and goal-directed cognition (Corbetta & Shulman, 2002; Vossel et al., 2014) as well as regions of the anterior and posterior default mode network that have been shown to code for task-irrelevant mind-wandering and thus may interfere with cognitive processing (Andrews-Hanna et al., 2007; Poerio et al., 2017; Raichle et al., 2001).

In our sample, we did not find strong age-related neural activity differences underlying cognition across the different brain regions studied; only the anterior DMN region encompassing the rostral anterior cingulate and medial orbitofrontal ROIs showed significant differences in global stimulus-evoked activity. Specifically, neural activity in younger adults was suppressed relative to baseline in the anterior DMN, while this effect was not observed in older adult. Overall, older adults had greater anterior DMN activity during global stimulus-evoked processing. This finding is aligned with prior evidence of age-related modulation of DMN activity showing reduced de-activation during cognitive tasks in aging with increased age (Andrews-Hanna et al., 2007; Chadick et al., 2014; Eudave et al., 2018; C. Grady et al., 2016; C. L. Grady et al., 2010; Lustig et al., 2003; Marstaller et al., 2015; Persson et al., 2014). Eudave and colleagues (2018) showed impaired speed of processing during simulated driving in older adults related to greater DMN activity. Similarly, Chadick et al. (2014) showed greater interference during working memory in aging related to anterior DMN (medial prefrontal cortex) activity. In healthy adults, DMN activity is usually suppressed during attentionally-demanding tasks (Buckner et al., 2008; Li et al., 2012; Raichle et al., 2001) and is atypically overactive in psychiatric disorders (Mohan et al., 2016; Sheline et al., 2009; Zhou et al., 2020). Hence, enhanced DMN activity during cognitive control tasks, or reduced deactivation in aging has been related to decreased ability to suppress distractions (Grady et al., 2010; Lustig et al., 2003). Longitudinal functional imaging in the context of an episodic memory task revealed that cognitively intact older adults show reduced deactivation of the DMN with increasing age (Persson et al., 2014). Notably, disruption of DMN connectivity in normal aging has been associated with disruptions in white matter integrity and poor cognitive performance across a range of domains (Andrews-Hanna et al., 2007), also verified in more recent studies that show changes in not just intra- but inter-network DMN connectivity with aging (Amer et al., 2016; Brown et al., 2019; Campbell et al., 2013; Contreras et al., 2017; C. Grady et al., 2016). These functional changes in the DMN have been further shown to progress to dysfunctional impairment in Alzheimer’s disease (Dennis & Thompson, 2014; Greicius et al., 2004; Lustig et al., 2003). Our EEG source imaging results thus support previous fMRI findings, which show that aging corresponds with an inability to suspend DMN activity during active cognitive control, weakening the ability to distinguish distractions from relevant information, hence, impacting task performance (A Gazzaley, 2013).

Finally, we constructed regression models to interrogate the brain region activations that contribute to global cognitive efficiency as well as mental health symptoms. Here, we found several age by neural activity interactions that were not apparent in overall activity magnitude comparisons, and instead driven by interindividual variation in brain activity. Younger adults exhibited greater activity in the right dorsolateral prefrontal cortex with greater performance efficiency, while in older adults, greater activity in the right inferior frontal cortex was associated with greater efficiency. A decreased role for the DLPFC region with increased age has been demonstrated in prior work (Campbell et al., 2012; C. L. Grady et al., 1995; Kehoe et al., 2013; Pardo et al., 2007; Rypma & D’Esposito, 2000), which has also been linked to age-related frontal grey matter loss (Marstaller et al., 2015; Resnick et al., 2003; Van Petten et al., 2004). That older adults with superior cognitive performance had greater inferior frontal activations may suggest a greater need for interference control (Brass et al., 2005; Zanto et al., 2011). We have previously shown that the inferior frontal cortex is capable of significant age-related brain plasticity in the context of distractor processing (Mishra et al., 2014), which aligns with the important role of this region in cognitive processing in aging. Overall, the differential neural activity associations with cognitive performance in younger versus older adults potentially support compensatory neural mechanisms in aging (Cabeza et al., 2002, 2018; Hakun et al., 2015; Nashiro et al., 2012; Ohsugi et al., 2013; Reuter-Lorenz & Cappell, 2008; Riis et al., 2008). Additionally, we found that mental health symptom severity (i.e., greater anxiety, depression, loneliness and reduced wellbeing) is predicted by an age by neural activity interaction in the posterior DMN region. It is logical to investigate cognitive information processing in the context of mental health, given much research that shows a relationship between cognitive dysfunction and symptoms of mental illness (Goodkind et al., 2015; Koban et al., 2021; Millan et al., 2012; Yager & Feinstein, 2017). Specifically, we found that greater posterior DMN activity in older adults during cognitive processing is related to more aggravated mental health symptoms. While our study focuses on symptom reports in healthy older adults, this result is supported by functional as well as structural MRI studies that show an association between posterior DMN dysfunction and anxiety/depression/loneliness symptoms identified in aging (Andreescu et al., 2011; Düzel et al., 2019; McLaren et al., 2016).

As per limitations of the study, we present results from a cross-sectional aging sample with moderate sample size. Our screen for healthy aging is limited to the Mini-Mental State Exam and self-reported illness diagnoses and medications indicative of neuropathology; future studies will benefit from integration of more comprehensive neuropsychological screening. Furthermore, even within non-pathological healthy aging, distinct cognitive and neural correlates may emerge for successful or ‘optimal’ agers (those that are thriving and flourishing) vs. those that are healthy yet not thriving (Cabeza et al., 2018; Martineau & Plard, 2018; Riis et al., 2008; Thomas et al., 2016). These distinct subsets need to be carefully studied to inform programs aimed at promoting successful aging. Our cognitive evaluations are limited to response speed-driven assessments, and we did not interrogate tasks that allow for extended cognitive reflection or non-binary decision-making. Our EEG-derived brain measures have high feasibility and affordability that facilitate scale-up, yet the source reconstructions are based on inverse model estimates that may not be exact. Hence, future research must address the generalizability of our findings in larger samples.

Overall, our research contributes towards a more integrated understanding of age-related cognitive and mental health differences while also simultaneously studying underlying neural processing. Our results confirm prior cognitive neuroimaging research in aging with affordable and scalable EEG-based source reconstruction methods. The results highlight interindividual variability in cognitive mechanisms in aging rather than significant activity differences between age groups. Our approach has important clinical translational significance as EEG acquired during cognitive tasks can be an accessible tool alongside resting state measures to predict risk for cognitive decline and monitor Alzheimer’s progression (Babiloni et al., 2021; Borhani et al., 2021). Ultimately, such research that integrates neural mechanisms, cognitive functioning and mental health may be used to identify future targets for age-related interventions that both enhance wellbeing and mitigate age-related cognitive decline.

Public Significance.

This study found that healthy older adults show greater mental wellbeing relative to younger adults, yet have relatively poorer cognitive performance. Mechanisms investigated using brain wave recordings showed that older adults did not suppress brain functions that are irrelevant to cognitive tasks as much as younger adults, particularly in the brain’s default mode network (DMN). Notably, older adults who had lesser activity in the DMN during cognitive tasks reported greater wellbeing. Our findings have potential utility to provide brain markers that can help monitor and mitigate age-related cognitive decline and simultaneously enhance wellbeing.