Age-Related Learning And Working Memory Impairment in The Common Marmoset Part 3

To determine whether there is a relationship between age and the amount of experience necessary to acquire the rules of the DRST, we calculated the number of trials each marmoset completed before their performance significantly deviated from chance (i.e., the length of the Novice Phase). 

For example, if we only master some basic knowledge points when studying a certain course, then our memory of this knowledge point is still relatively shallow. But if we delve into these knowledge points through practice, exploration, and thinking during the learning process, and continuously apply them in practice, then our experience will continue to accumulate, and this information will be more deeply integrated. in our memory.

In addition to the field of study, the improvement of memory through experience can also be fully reflected in life. For example, when we deal with a certain problem, if we have experienced similar problems or similar scenarios, then we can think of a solution faster. This is because our experience helps us describe the problem more clearly in our minds. , come up with solutions faster.

The increase in experience not only improves our memory but also makes us more confident. As the saying "Experience is the best teacher" says, only by accumulating experience through continuous practice and summarizing lessons from it can we be better able to handle similar things and become more confident.

In short, the relationship between experience and memory is inseparable. Only with positive efforts and continuous accumulation of experience through practice and thinking can our memory be fully trained and improved, making us smarter, more confident, and more mature. It can be seen that we need to improve memory, and Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material that has many unique effects, one of which is to improve memory. The efficacy of minced meat comes from the various active ingredients it contains, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know supplements to improve memory

Using the marmoset's age at the end of the Novice Phase, a Spearman's rank-order correlation revealed a strong, positive association that was statistically significant, indicating that aging marmosets required more experience to initially acquire the DRST rules (Fig. 3D; rs(13) = 0.65, p = 0.009). 

One nine-year-old male marmoset ("FL" in Fig. 1) failed to perform above chance levels, despite completing 7000 DRST trials, and so this animal was excluded from the statistical analyses conducted on the Learner and Expert Phases, described below.

In the Learner Phase, marmosets demonstrated rapid performance improvement. To determine whether there were aging-associated differences in the rate of improvement in the Learner Phase, we assessed the maximum learning rate for each monkey. 

The maximum learning rate was calculated by finding the largest derivative of the mean FSL curve, computed over a rolling 100-trial window. A Spearman's rank-order correlation revealed a significant negative correlation between average marmoset age over the 100-trial window, and maximum learning rate, demonstrating that with increasing age, learning rates slowed (Fig. 3E; rs(13) = 0.61, p = 0.015).

In the Expert Phase, marmosets reached maximal levels of performance and were typically plateaued. We determined each marmoset's maximum level of performance by calculating the mean FSL achieved during the Expert Phase. 

A Spearman's rank-order correlation was used to assess the relationship between the average age of each marmoset during the Expert Phase and maximum FSL. It revealed a significant negative correlation between these variables, indicating that marmoset aging is associated with decreased working memory capacity (Fig. 3F; rs(13) = 0.54, p = 0.037).

Additional information about cognitive performance and how marmosets approached the task can be discerned by assessing patterns of errors made while performing the DRST. 

Instances when marmosets end a trial with an FSL of two (i.e., make an error on TDL3 when there are three objects on the screen), are especially informative since errors can only be made by selecting the first object of the trial (primacy error) or the most recently correct object of the trial (perseverative error). Prior studies of learning have found that animals have a natural tendency to continue to perform behaviors for which they have recently been rewarded, a behavioral pattern that can yield perseverative errors. 

Thus, we hypothesized that, before acquiring the rules of the DRST (i.e., in the Novice Phase), marmosets would make more perseverative errors relative to primacy errors, and that once the rules of the DRST had been successfully acquired (i.e., Learner and Expert Phases), marmosets would make more primacy errors relative to perseverative errors. 

To examine this we calculated the proportion of primacy and perseverative errors marmosets made on TDL3 trials during each of the three Phases (Fig. 4A) and found that there was a significant interaction between error type (perseverative, primacy) and Phase (Novice, Learner, Expert; Scheirer Ray Hare test: H (2) = 58.24, p = 2.0
1013). 

Across Phases, there was a significant decrease in the fraction of perseverative errors and a corresponding increase in the fraction of errors that were primacy (Friedman's test; x2 (2) = 22.93, p = 1.05
105 ). Specifically, this reached statistical significance between the Novice and Learner Phases and between the Novice and Expert Phases but not between the Learner and Expert Phases (pairwise post hoc Nemenyi tests; Novice vs Learner: p = 0.01; Novice vs Expert: p = 6.55
106; Learner vs Expert: p = 0.16). 

In line with our hypothesis, the predominant error type changed with learning of the task such that during the Novice Phase, errors were predominantly perseverative, whereas primacy errors were most prevalent in the Learner and Expert Phases (Wilcoxon signed-rank tests for paired samples; Novice: Z = 3.38, p = 6.10
105; Learner: Z = 2.641, p = 0.0054; Expert: Z = 2.527, p = 0.0084).

To determine whether the slowed learning rates of aging marmosets could be explained by a protracted time course of the predominant error type change ("crossover"), we quantified the number of TDL3 trials before the crossover for each animal (Fig. 4B, C). 

A Spearman's rank-order correlation revealed a significant positive correlation between age at the time of the crossover event and the number of trials to the crossover, demonstrating that, on TDL3 trials, aging marmosets switch to predominantly primacy errors after more task experience than younger marmosets (Fig. 4C; rs(13) = 0.72, p = 0.002).

Since the TDL3 error analyses revealed an age-related difference in the rate of error type switching from perseverative to primacy, we went on to examine whether there was an age-related change in the magnitude of perseveration as well. Previous studies have found that aged macaque monkeys make more perseverative errors on the DRST than do young macaques (Moss et al., 1997). 

Therefore, we hypothesized that higher levels of perseveration could underlie the slowed learning rate of aging marmosets (Fig. 3E) and may also explain the overall age-related working memory impairment (Fig. 3F). 

To test this hypothesis, we calculated the proportion of perseverative errors across TDLs during each of the three Phases. We found that aging was associated with more perseverative errors, but only in the Learner Phase (Fig. 4D–F; Spearman's rank-order correlations using average age during a phase; Novice: rs(14) = 0.035, p = 0.90; Learner: rs(13) = 0.65, p = 0.009; Expert: rs(13) = 0.411, p = 0.13). 

These results show that higher levels of perseveration underlie the slowed learning rate of aging marmosets and that perseverative errors do not account for age-related working memory impairment.

Trials of increased difficulty were reliably performed during the Expert Phase. The pattern of errors marmosets made on these trials provides an opportunity to examine whether marmosets succumb to working memory interference. To do this, we quantified the distribution of errors according to the distance in the past the incorrectly chosen stimulus was presented (i.e., "back"). 

For example, if a marmoset made an error when there were four objects on the screen (TDL4), they would earn an FSL of three on that trial. In this case, they could make an error by selecting the first object presented on the trial (n-back 3; primacy), the second object presented on the trial (n-back 2), or the third object presented on the trial (n-back 1; perseverative). 

The distribution of n-back errors made in the Expert Phase was quantified for each TDL and compared with chance performance using x2 Goodness of Fit Tests (Fig. 4H). These tests revealed that, for all TDLs, the observed n-back error distributions were significantly different from those expected by chance (TDL3: x2 = 154.17, p = 2.13
1035; TDL4: x2 = 110.98, p = 7.94
1025; TDL5: x2 = 116.52, p = 4.34
1025; TDL6: x2 = 74.06, p = 3.15
1015; TDL7: x2 = 50.44, p = 1.13
10
109; TDL8: x2 = 38.49, p = 9.01
107; TDL9: x2 = 14.74, p = 0.04). 

Further, on each TDL, marmosets made significantly more primacy errors than perseverative errors (Wilcoxon signed-rank tests for paired samples; TDL3: Z = 3.38, p = 6.1
105; TDL4: Z = 3.05, p = 0.001; TDL5: Z = 3.24, p = 0.0002; TDL6: Z = 3.27, p = 0.0001; TDL7: Z = 3.15, p = 0.00,024; TDL8: Z = 2.52, p = 0.0081; TDL9: Z = 2.10, p = 0.034). These results show that marmosets experienced retroactive interference, wherein newly acquired information disrupts temporary memory storage, leading to errors made by identification of more remotely presented stimuli as novel.

One advantage of using infrared touch screens to assess cognitive performance is the ability to measure choice latencies accurately and reliably. This measure is widely considered a robust readout of processing speed and is correlated with cognitive load and task difficulty (Bopp and Verhaeghen, 2018; Gray et al., 2018; De Boeck and Jeon, 2019). 

To determine whether this pattern held with marmosets performing the DRST, we assessed correct and incorrect choice latencies for each marmoset on each Phase of the DRST, and for each TDL in the Expert Phase. A Scheirer Ray Hare test revealed significant main effects of latency type (correct choice, incorrect choice) and Phase of the DRST (Novice, Learner, Expert), and also a significant interaction between these factors (Fig. 5A; latency type: H(1) = 12.58, p = 0.0004; Phase: H (2) = 15.13, p = 0.0005; interaction: H(2) = 8.69, p = 0.01). Correct choice latencies, but not incorrect choice latencies, significantly decreased across the Phases of the DRST (Friedman's tests; correct: x2 (2) = 22.93, p = 1.05
105; incorrect: x2 (2) = 0.12, p = 0.94). 

Correct choice latencies were shorter in the Learner and Expert Phases than in the Novice Phase, with no difference between the Learner and Expert Phases (pairwise post hoc Nemenyi tests; Novice vs Learner: p = 0.01; Novice vs Expert: p = 6.55
105; Learner vs Expert: p = 0.16). These results show that, with improvement on the task, marmosets were more quickly able to identify the novel stimulus. Next, to determine whether incorrect choices could be because of impulsiveness, we compared correct to incorrect choice latencies within each of the Phases. 

During the Novice Phase, correct and incorrect choice latencies were similar; however, during the Learner and Expert Phases, incorrect choice latencies were significantly longer than correct choice latencies (Fig. 5A; Wilcoxon signed-rank tests for paired samples; Novice: Z = 0.71, p = 0.49; Learner and Expert: Z = 3.38, p = 6.10
105 ). This demonstrates that, when marmosets made an error, it was likely not because of impulsivity since they took considerably longer to respond in those cases.

To investigate choice latency patterns more thoroughly, we quantified the Expert Phase latency data by TDL. Since there was no significant relationship between average age during the Expert Phase and choice latencies (Spearman's rank-order correlations; correct: rs(13) = 0.40, p = 0.14; incorrect: rs(13) = 0.30, p = 0.27), further analyses focused on assessing whether the latency data revealed information about cognitive load and task difficulty. First, we examined whether the increased cognitive load required by more difficult TDLs was reflected in the choice latency data. 

Strong, positive Spearman's rank-order correlations between both types of choice latency and trial difficulty level reflect longer processing time necessary for more difficult trials (Fig. 5B; correct: rs(7) = 0.93, p = 0.0002; incorrect: rs(6) = 0.95, p = 0.0003). Finally, we assessed whether the overall pattern of longer incorrect than correct choice latencies found in the Expert Phase was consistent for each of the TDLs separately. A Scheirer Ray Hare test revealed significant main effects of latency type (correct choice, incorrect choice; H (1) = 27.97, p = 1.2
107 ) and trial difficulty level (TDL1–TDL9; H(7) = 94.12, p = 1.8
1017), but no significant interaction (H(7) = 0.97, p = 1.00). 

Wilcoxon signed-rank tests for paired samples revealed incorrect choice latencies were longer than correct choice latencies at all TDLs, demonstrating that the marmosets took more time to choose if they made an error, even on the most difficult trials, and therefore were not choosing impulsively when incorrect (Fig. 5C; TDL2: Z = 3.26, p = 0.0001; TDL3: Z = 3.21, p = 0.0002; TDL4: Z = 3.26, p = 0.0001; TDL5: Z = 3.26, p = 0.0001; TDL6: Z = 3.20, p = 0.0002; TDL7: Z = 3.26, p = 0.0001; TDL8: Z = 3.14, p = 0.0004; TDL9: Z = 3.26, p = 0.0001).

To assess marmosets' engagement on the DRST, we quantified the proportion of initiated trials that were completed versus omitted (no response within the 12-s response window). There was no significant difference in trial completion rate across the Novice, Learner, and Expert Phases (Fig. 6A; Friedman's test; x2 (2) = 1.73, p = 0.42), indicating that marmosets were equivalently engaged with the task throughout all Phases. Further, there was no significant relationship between age and proportion of trials completed for any of the Phases (Novice: rs(14) = 0.044, p = 0.08; Learner: rs(14) = 0.10, p = 0.73; Expert: rs(14) = 0.30, p = 0.27). 

Finally, to assess whether marmosets were equally engaged on portions of trials that were more difficult (i.e., higher TDLs), we quantified completion rates for each of the TDLs independently. During the Expert phase, marmosets completed .90% of each of the TDLs they encountered, demonstrating continued engagement even when trials became challenging, despite not being water restricted, testing in the home cage in a room that housed many other animals, and being free to leave the testing chamber at any time (Fig. 6B).

Reaction Time Task (RTT)

To assess potential noncognitive confounds for the interpretation of our DRST results, we measured motor speed using a Reaction Time Task. In this task, marmosets initiated a trial and then were rewarded for touching a single target stimulus placed pseudo-randomly on the screen (Fig. 7A). A Spearman's rank-order correlation, used to determine the relationship between age and reaction time, revealed a strong, positive correlation between these variables, which reached statistical significance (Fig. 7C; rs(10) = 0.87, p = 0.0002). 

To determine whether age-related increased reaction times were observed at each of the nine locations, or whether, for some locations, there was no age-related association, we correlated age with reaction time for each of the nine possible stimulus locations separately. We found strong associations between advancing age and longer reaction times when the target stimulus appeared in the most difficult to reach locations on the screen: all three locations in the top row, and the two outer positions in the middle row (top row, location 1: rs(10) = 0.72, p = 0.008; location 2: rs(10) = 0.68, p = 0.015; location 3: rs(10) = 0.70, p = 0.012; middle row, location 4: rs(10) = 0.80, p = 0.002; location 5: rs(10) = 0.54, p = 0.07; location 6: rs(10) = 0.60, p = 0.04; bottom row, location 7: rs(10) = 0.57, p = 0.054; location 8: rs(10) = 0.44, p = 0.15; location 9: rs(10) = 0.31, p = 0.32). These results demonstrate that with aging, marmosets have slowed motor speed.

To assess whether the age-related motor speed deficits were associated with DRST performance, we correlated reaction times on the RTT with several primary dependent variables from the DRST. There were no significant Spearman's correlations between reaction time on the RTT and any of the DRST measures assessed including maximum learning rate (rs(10) = 0.46, p = 0.14), maximal performance measured by FSL (rs(10) = 0.46, p = 0.13), and Expert Phase trial completion rate (rs(10) = 0.44, p = 0.15). These results show that although marmosets have slowed motor speed with age, this did not negatively impact DRST performance.

Progressive Ratio Task

To assess another potential noncognitive confound for the interpretation of our DRST results, we measured motivation using a Progressive Ratio Task. In this task, marmosets were required to expend increasing amounts of effort (screen touches) to earn a same-sized reward (Fig. 7B). A Spearman's rank-order correlation was used to determine the relationship between age and the number of rewards earned during the PRT (Fig. 7D). There was a strong, negative correlation between these variables, which was statistically significant, indicating that with advancing age, marmosets are less motivated to expend energy in exchange for a reward (rs(10) = 0.82, p = 0.001). There was a similarly strong, negative correlation between age and break point, showing that the final response requirement completed within the 1-h session was lower with age (rs(10) = 0.80, p = 0.002).

To assess the relationship between decreased motivation with age on the PRT and cognitive performance on the DRST, Spearman's rank-order correlations were run between rewards earned on the PRT and the primary dependent variables from the DRST. There were no significant correlations between rewards earned on the PRT and any of the DRST measures assessed including maximum learning rate (rs(10) = 0.21, p= 0.52), Expert Phase FSL (rs(10) = 0.31, p= 0.33), and Expert Phase trial completion rate (rs(10) = 0.45, p = 0.14).

Discussion

This work firmly establishes the marmoset as a key model of age-related cognitive impairment with several advantages. This is important because, as a NHP, the marmoset offers translational relevance to humans beyond that found in rodents. As compared with the long-lived macaque, the short lifespan of marmosets enables longitudinal aging studies within a reasonably short timeframe. 

In this study, we performed comprehensive behavioral characterization of a large cohort of marmosets. These marmosets performed thousands of trials to assess learning and working memory capacity continuously for up to four years. This duration of continuous assessment accounts for up to 50% of the adult marmoset lifespan. 

Further, marmosets began training at ages that conjointly covered the lifespan, enabling assessment of age-related differences in the acquisition of the working memory task. To our knowledge, these monkeys have undergone the most thorough cognitive profiling of any marmoset aging study conducted to date, enabling us to identify which specific aspects of task performance are impaired with age.

Comparison with prior studies of age-related cognitive changes in macaque and marmoset

Earlier DRST studies concluded that aged macaque monkeys have poorer working memory than young macaques (Herndon et al., 1997; Moss et al., 1997). The present results show that marmoset aging is associated with both learning and working memory impairments. Specifically, we found that aging is associated with delayed onset of learning, slowed learning rate after onset, and decreased asymptotic working memory performance. These impairments are evident in our study and may have been undetected in the macaque because of key differences in task implementation. 

In the macaque studies, monkeys were trained to a high level of proficiency on a delayed nonmatch-to-sample (DNMS) task before administration of a relatively small number of DRST trials (one hundred). This experimental design assumes that macaques can transfer the rule learned through DNMS training to DRST, with no additional training, and that they immediately perform the DRST maximally. It is unclear that they can, given the greater complexity of the DRST task. In contrast, the monkeys used in the present study trained on the DRST for thousands of trials over an extended period. This enabled us to analyze both learning and maximal performance separately, using the same task. It is an open question whether the impaired performance of aged macaques reflects a memory impairment, a delayed onset of learning, a slower rate of learning, or some combination of these factors.

To date, there is one prior study of cognitive aging in marmosets where data were collected over several years (Rothwell et al., 2022). This prior study assessed visual discrimination and cognitive flexibility over a substantial portion of the adult marmoset lifespan. However, there are several key experimental design differences between this prior study and the work reported here, which merit discussion. First, Rothwell and colleagues implemented longitudinal analyses and tested marmosets once per year, whereas we used cross-sectional analyses and measured cognitive performance continuously. The present approach densely characterized individual cognitive aging trajectories, facilitating analysis of multiple aspects of task performance. While we identified age-related impairment across all Phases of testing, Rothwell and colleagues reported impairment only on the last annual test. 

Second, marmosets enrolled in the Rothwell and colleagues study began testing at approximately the same age (four to six years old), and impairment was identified when all animals were between 8 and 10 years of age. In contrast, marmosets in our study began testing at ages varying from young adult to geriatric, and we found that cognitive impairment occurs throughout the adult marmoset lifespan, in line with what has been observed in humans (Wu et al., 2021). Since aging is inherently a heterogeneous process, the fact that all marmosets in the Rothwell and colleagues study showed cognitive impairment simultaneously is unexpected. 

One possible explanation is that stimulus-dependent variation in task difficulty contributed to this pattern. For example, some stimulus pairs used in the first three years of the study were discriminable by shape and color, whereas all stimuli used in year four differed only by shape. Therefore, marmosets may have found the stimuli used in the terminal year more challenging, accounting for the decline in performance. Further, Rothwell and colleagues reported that impairment on the visual discrimination task was more severe than on the cognitive flexibility task. This is the opposite of what would be expected from published work demonstrating that cognitive flexibility often declines with aging, while visual discrimination typically does not (Bartus et al., 1979; Lai et al., 1995). Our results align with decades of work in macaques and humans, demonstrating the translational relevance of the marmoset as a model of aging.

Patterns of errors reveal Marmosets' approach

Since a detailed analysis of the types of errors made on neurocognitive assessments can, in humans, predict the emergence of future cognitive impairment (Bondi et al., 1999; Thomas et al., 2018), we investigated the patterns of errors marmosets made on the DRST as they learned the task and when they were performing maximally. Previous work in macaques has shown that aging is associated with increased perseverative errors on the DRST (Moss et al., 1997). Interestingly, our results only recapitulated this finding during the Learner Phase, and not during the Novice or Expert Phases, supporting the idea that aged macaque DRST impairment reflects learning rather than working memory deficits. In addition, older marmosets required more experience before they switched from mostly perseverative to mostly primacy errors. We hypothesize that both findings are associated with the learning impairment we identified in marmosets and support learning rather than working memory impairment in the aged macaques. 

Finally, we used the pattern of error choices to determine whether marmoset performance was affected more significantly by proactive interference (i.e., full working memory prohibits adding novel objects), or retroactive interference (i.e., newly encountered information obscures more remote memories). We found that once marmosets were performing maximally, they more frequently made an error by selecting stimuli that appeared earlier in the trial. This supports the idea that marmosets had more difficulty remembering more remote stimuli, and that performance was affected most significantly by retroactive interference (Amit et al., 2003). Together, these findings highlight the importance of analyzing process scores in addition to performance scores to understand behavior more thoroughly (Kaplan, 1988).

Age-related motor speed deficits

Age-related motor speed deficits, measured by increased reaction times, are well-documented in rodents, NHPs, and humans (Bachevalier et al., 1991; Woods et al., 2015). To understand the impact of age on reaction time in our marmosets, and to assist us in controlling for this noncognitive confound when interpreting our DRST results, we employed the RTT in a subset of the marmosets that performed the DRST. We found a strong association between increased reaction times and increased age. Critically, we found no associations between reaction time on the RTT and any of the DRST performance scores, demonstrating that the age-related motor speed impairments do not account for impairments on the DRST.

Unlike our findings from the RTT, we did not find any associations between aging and choice latencies on the DRST. We did, however, find that choice latencies were longer for more difficult TDLs, in all animals. This is likely because choice latencies on cognitive tasks are approximations of task difficulty and cognitive load (Bopp and Verhaeghen, 2018; Gray et al., 2018; De Boeck and Jeon, 2019), and this supersedes the effects of age-related motor deficits. Additionally, for all TDLs, marmosets responded more slowly if their answer was incorrect than correct, demonstrating that, even when trials became more difficult, marmosets did not respond impulsively.

Age-related motivation deficits

It is well-known that older rodents, NHPs, and humans have decreased motivation to earn rewards (Lanctôt et al., 2017; Lizarraga et al., 2020; Jackson et al., 2021). To evaluate the potential impact of age-related changes in motivation on DRST performance, we employed the well-established PRT to quantify motivation in our marmosets (Spinelli et al., 2004). We found a strong association between increasing age and decreasing motivation to earn rewards. We did not, however, identify any relationship between motivation on the PRT and any of the DRST performance scores, demonstrating that age-related decreases in motivation did not account for impairments on the DRST.

To directly assess motivation to perform the DRST, we quantified the proportion of trials that monkeys completed versus omitted. We found no significant relationship between age and proportion of trials omitted, unlike what we would have predicted from marmosets' performance on the PRT. Furthermore, marmosets were equally likely to complete trials of each TDL, demonstrating engagement on the DRST even when trials were difficult. We hypothesize that when marmosets failed to complete a trial, it was likely because of inattention or distraction rather than low motivation.

In summary, here we have established the marmoset as a key model of age-related cognitive impairment. Our approach densely characterized cognitive aging trajectories in a large cohort of marmosets performing a translationally relevant working memory task. Through rigorous evaluation of process scores, we show that cognitive impairment can be detected long before it can be measured by overall performance scores. This framework will lead to a better understanding of the aging process itself and may reveal why it is the biggest risk factor for neurodegenerative diseases such as Alzheimer's disease.

References

Abbott DH, Barnett DK, Colman RJ, Yamamoto ME, Schultz-Darken NJ (2003) Aspects of common marmoset basic biology and life history important for biomedical research. Comp Med 53:339–350.

Amit DJ, Bernacchia A, Yakovlev V (2003) Multiple-object working memory–a model for behavioral performance. Cereb Cortex 13:435–443.

Bachevalier J, Landis LS, Walker LC, Brickson M, Mishkin M, Price DL, Cork LC (1991) Aged monkeys exhibit behavioral deficits indicative of widespread cerebral dysfunction. Neurobiol Aging 12:99–111.

Bartus RT, Dean RL, Fleming DL (1979) Aging in the rhesus monkey: effects on visual discrimination learning and reversal learning. J Gerontol 34:209–219.

Beason-Held LL, Rosene DL, Killiany RJ, Moss MB (1999) Hippocampal formation lesions produce memory impairment in the rhesus monkey. Hippocampus 9:562–574.

Belham FS, Satler C, Garcia A, Tomaz C, Gasbarri A, Rego A, Tavares MCH (2013) Age-related differences in cortical activity during a visuospatial working memory task with facial stimuli. PLoS One 8:e75778.

Belleville S, Rouleau N, Caza N (1998) Effect of normal aging on the manipulation of information in working memory. Mem Cognit 26:572–583.

Bondi MW, Salmon DP, Galasko D, Thomas RG, Thal LJ (1999) Neuropsychological function and apolipoprotein E genotype in the preclinical detection of Alzheimer's disease. Psychol Aging 14:295–303.

Bopp KL, Verhaeghen P (2018) Aging and n-back performance: a meta-analysis. J Gerontol B Psychol Sci Soc Sci 75:229–240.

Bor D, Duncan J, Lee ACH, Parr A, Owen AM (2006) Frontal lobe involvement in spatial span: converging studies of normal and impaired function. Neuropsychologia 44:229–237.

Cappell KA, Gmeindl L, Reuter-Lorenz PA (2010) Age differences in prefrontal recruitment during verbal working memory maintenance depend on memory load. Cortex 46:462–473.

Comrie AE, Gray DT, Smith AC, Barnes CA (2018) Different macaque models of cognitive aging exhibit task-dependent behavioral disparities. Behav Brain Res 344:110–119.

De Boeck P, Jeon M (2019) An overview of models for response times and processes in cognitive tests. Front Psychol 10:102.

De Castro V, Girard P (2021) Location and temporal memory of objects declines in aged marmosets (Callithrix jacchus). Sci Rep 11:9138.

Gazzaley A, Cooney JW, Rissman J, D'Esposito M (2005) Top-down suppression deficit underlies working memory impairment in normal aging. Nat Neurosci 8:1298–1300.

For more information:1950477648nn@gmail.com