The Effect of Baseline Performance and Age on Cognitive Training Improvements in Older Adults: A Qualitative Review

Abstract

Findings that the brain is capable of plasticity up until old age have led to interest in the use of cognitive training as a potential intervention to delay the onset of dementia. However, individuals participating in training regimens differ greatly with respect to their outcomes, demonstrating the importance of considering individual differences, in particular age and baseline performance in a cognitive domain, when evaluating the effectiveness of cognitive training. In this review, we summarized existing literature on cognitive training in adults across several cognitive domains to contrast the conditions supporting the magnification and compensation effects of cognitive training improvement. The strongest support for the magnification effect was observed in studies involving younger adults with more intact cognitive abilities in training tasks targeting episodic memory, explained in part by factors specific to episodic memory training. By contrast, support for the compensation effect was found in several studies targeting working memory in older individuals as well as in the majority of studies targeting executive functioning, suggesting the preservation of neural plasticity in these domains to very old age. Future research should more clearly stratify individuals according to their baseline cognitive abilities and assign specialized, skill-specific cognitive training regimens in order to directly answer the question of how individual differences impact training effectiveness.

1. Introduction

Cognitive training (CT) is a program of guided mental activities designed to maintain or improve cognitive functions that support accomplishment of everyday tasks and independent living (Maraver et al. 2016). CT relies on functional and structural neural plasticity to improve cognitive functioning. Interest in CT training has increased following findings that the brain is capable of significant plasticity up until very old age (de Villers-Sidani et al. 2010; Hosseini et al. 2014; Nyberg et al. 2012; Zinke et al. 2014). Research suggests that CT is helpful not only in terms of limiting cognitive decline in aging individuals but may also improve cognitive performance in individuals with mild cognitive impairment (Mendoza Laiz et al. 2018). However, past studies have administered CT regimens indiscriminately without taking into account baseline characteristics of individuals. Prior research has shown that individuals participating in similar CT regimens may differ substantially with respect to their outcomes suggesting the importance of considering individual differences when evaluating for whom CT works, and in what context (Könen et al. 2015).

Prior research suggests that the factors contributing to individual variability in response to CT include fluid intelligence, mental status, verbal ability and compliance in implementing CT strategies (Verhaeghen and Marcoen, 1996). However, two particular factors — baseline cognitive ability and age — are highly predictive of CT gains, especially with respect to memory training, and have been the primary focus of past literature (Borella et al. 2017).

Two opposing hypotheses (Figure 1) have been proposed to explain the role of age and baseline performance (BP) in a cognitive domain on the effectiveness of CT (Borella et al. 2017). The compensation effect predicts that high BP or younger individuals will benefit less from CT because they are already functioning near their optimal level and have less room for improvement. By contrast, the magnification effect postulates that high-performing individuals or younger individuals who already perform well in a specific cognitive domain will benefit most from training in that domain because they may utilize their more efficient cognitive resources (Borella et al. 2017; Lövden et al. 2012). Both theories depend on the notion that younger individuals generally have undergone less cognitive decline than older adults and have a higher baseline cognitive profile. While there is ongoing debate about which theory best accounts for the effect of age and BP on training gains, more research is needed to delineate which of these theories is more relevant in different cognitive domains to better understand how training gains can be maximized given specific training conditions and cognitive abilities.

Fig. 1. The magnification and compensation effect theories of cognitive training (CT).

The compensation effect predicts that older, lower baseline ability individuals improve most from CT, while the magnification effect hypothesizes that younger, higher baseline ability individuals improve most. Green brain regions indicate optimal utilization.

In order to discuss CT effectiveness, it is important to contextualize how ‘gains’ are typically measured (Table 1). The primary five metrics of training effect include: (1) training improvements, (2) near-transfer improvements to structurally similar tasks, (3) near-transfer improvements to structurally dissimilar tasks, (4) far-transfer improvements, and (5) improvements in daily life functioning and activities. It is important to note that while the training and near-transfer improvements will be the focus of this review, they do not necessarily lead to improvement in cognitive ability unless they enhance one’s capacity to engage beneficial neural processes not typically engaged (Curlik and Shors, 2012).

Table 1.

The five unique metrics utilized to determine cognitive training (CT) effectiveness.

Training Metric	Description	External Validity	Example
Training Improvements	Measure of improvement on the training tasks administered throughout the study	Lowest	Zinke et al. (2012) assessed the two-week progress of older adults (N=20, mean age=86.8) in three verbal and two visuospatial working memory (WM) tasks
Improvements in Near-Transfer to Structurally Similar Tasks	Measure of changes in pre- and post-training scores on tasks within the same cognitive domain and with a very similar structure to the training tasks	Low	Li et al. (2008) assessed improvements in a spatial three-back task sharing processing requirements with the spatial two-back training task but at a higher level of WM load
Improvements in Near-Transfer to Structurally Dissimilar Tasks	Measure of changes in pre- and post-training scores on tasks in the same cognitive domain but with a dissimilar structure to the training tasks	Moderate	Karbach et al (2009) examined whether task-switching training in executive functioning (EF) transferred to structurally dissimilar near-transfer tasks such as the color-stroop and number-stroop tasks
Improvements in Far-Transfer	Measure of changes in the pre- and post-training scores on untrained tasks in different cognitive domains	High	Dahlin et al. (2008) assessed improvements in perceptual speed, episodic memory, and reasoning after 5 weeks of computer-based EF training
Improvements in Daily Life and Everyday Functioning Activities	Measure of improvements in tasks closely simulating everyday life activities and functioning abilities	Highest	Cantarella et al. (2017) utilized the everyday problem test, the timed instrumental activities of daily living scale and the behavioral rating inventory of executive function to assess changes in everyday life abilities following verbal WM training

Metrics of training effect range from least (training improvements) to most (improvements in everyday life and everyday functioning activities) indicative of everyday life cognitive improvements.

No past review has thoroughly investigated the impact of age and BP on CT effectiveness across multiple cognitive domains. Thus, this review will summarize the existing literature on CT to contrast the conditions supporting the magnification and compensation effects. More specifically, it will attempt to understand (1) how age and BP impact CT effectiveness, (2) how that varies across the domains of working memory (WM), episodic memory (EM) and executive functioning (EF), and (3) why domain-specific patterns of improvement emerge in order to clarify how the efficacy of CT can be maximized for each individual. Additionally, the review will discuss shortcomings of past studies and outline directions for future research.

2. Methods

A PubMed search using the keywords ‘age,’ ‘baseline cognitive ability,’ ‘baseline performance,’ ‘cognitive training,’ ‘individual differences’ and ‘training intervention’ was conducted to identify articles for review. We excluded: (1) articles that examined the effectiveness of CT without considering how individual differences moderate that effect; (2) articles that did not include healthy older adults; (3) articles that utilized CT regimens not exclusively focused on abilities in EM, WM or the task-shifting component of EF; and (4) articles that examined the impact of individual differences (i.e. motivation or genetic variants) on CT gains without significantly discussing BP or age (Figure 2).

Fig. 2. Flow diagram showing reasons for exclusion of studies and number of qualifying cognitive training (CT) studies in the domains of episodic memory (EM), executive functioning (EF) and working memory (WM).

108 studies were selected, and based on the exclusion criteria, 23 studies that examined the impact of baseline performance (BP) and age on CT effectiveness were chosen for analysis.

Once an extensive profile of articles was compiled, we selected several defining characteristics of each study to help aggregate the findings – for example, CT domain, study sample demographics, and training/transfer effects. To determine whether each study supported the occurrence of a compensation or magnification effect, we first looked at which subject body derived the most improvements in training and structurally similar near-transfer tasks. For studies in which improvements on these two metrics did not reveal which subject body benefitted most from CT, we examined structurally dissimilar near-transfer tasks and/or far-transfer tasks.

Ultimately, 23 studies (Table 2) were selected for analysis in the review based on the inclusion criteria outlined above. Of the 23 CT studies chosen, 6 involved EM training, 11 involved WM training, and 6 involved EF training with a focus on task-shifting. Additionally, one of the studies examined only the effect of BP on CT effectiveness, 16 examined only the effect of age on CT effectiveness, and 6 studies examined both age and BP as moderators of CT effectiveness.

Table 2.

Information on the 23 cognitive training (CT) studies chosen for review in the domains of working memory (WM), episodic memory (EM) and task-switching in executive functioning (EF).

Authors	Sample Means	CT Domain	Length of Training	Training Improvements	Transfer effects	Magnification or Compensation Effect?	Ceiling Effect?
*Borella et al. 2014	N=20/20 (young-old/old-old) Age=69.9/79.6 years Education= 10.7/8.8 years	WM (visuospatial matrix task)	3 sessions, 60 min	- Young-old=old-old	- Young-old > old-old (pattern comparison test, forward and backward corsi span test)	Magnification	No
*Brehmer et al. 2012	N=29/26 (young/old) Age=26.2/63.9 years Education= 15.0/15.3 years	WM (spatial and verbal WM tasks)	>= 20 sessions, 26 min	- Young > old	- Young > old (span board forward, digit span backward, and span board backward)	Magnification	No
*Bürki et al. 2014	N=22/22 (young/old) Age=24.7/67.6 years Education= 14.9/15.0 years	WM (verbal n-back task)	10 sessions, 30 min	- Young > old - High BP > low BP	- Old = young	Magnification	Yes
*Heinzel et al. 2014	N=15/15 (young/old) Age=25.9/66.1 years Education= 18.2/16.1 years	WM (n-back task)	9 sessions, 45 min	- Young > old	- Young > old (verbal fluency task) - Old > young (digit span forward and CERAD delayed recall)	Magnification	Yes, strong ceiling effects in CERAD Recall and the Digit Span tasks
*Li et al. 2008	N=19/21 (young/old) Age=25.3/74.5 years	WM (spatial two-back task)	45 sessions, 15 min	- Old > young (accuracy) - Old = young (reaction time)	- Old = young (spatial three-back, numerical two- and three-back tasks) - Old > young (two-back plus processing test)	Compensation	Yes
*Richmond et al. 2011	N=21 Age=66 years Education=17 years	WM (spatial and verbal WM tasks)	20 sessions, 30 min	- Old = young	- Old = young (reading span WM task)	Neither	- Yes for both groups on Test of Everyday Attention
*Stamenova et al. 2014	N=30 Age=67.6 years Education=15.6 years	WM (word remembering task)	3 sessions, 20–30 min	- Young > old - High BP > low BP	- Old = young (forward digit span and spatial source memory task)	Magnification	No
*Salminem et al. 2015	N=20/26 (young/old) Age=24.5/64.8 years	WM (dual n-back task)	14 sessions,	- Young > old	- Old = young (WM updating task)	Magnification	No
*Von Bastian et al. 2013	N=34/27 (young/old) Age=23/68 years Education=5±1/6±2 years	WM (tower of fame, numerical complex span and figural task switching tasks)	20 sessions, 30 min	- Young > old	- Young > old (work-position binding task)	Magnification	No but floor effect
*Zinke et al. 2012	N=20 Age=86.8 years Education=11.7 years	WM (verbal and visuospatial WM tasks)	10 sessions, 25–30 min	- Low BP > high BP	- No transfer effects	Compensation	Yes
*Zinke et al. 2014	N=40 Age=77.2 years Education=14.4 years	WM (verbal and visuospatial WM tasks)	9 sessions, 30 min	- Low BP > high BP	- Young > old (visuospatial WM task) - Old > young (fluid intelligence task)	Compensation	No
*Baltes et al. 1992	N=16/19 (young/old) Age=20–30/66–80 years	EM (method of loci)	2 instruction and 17 training sessions, 60 min	- Young > old	- N/A	Magnification	No
*Brehmer et al. 2007	N=29/29 (young/old) Age= 22.5/66.9 years	EM (method of loci)	2 instruction and 2–6 training sessions, 60 min	- Young > old	-Young > old (timed recall task)	Magnification	Yes for instruction gains in young adults
*Kliegl et al. 1990	N=18/19 (young/old) Age=23.9/71.7 years Education=College/12.5 years	EM (method of loci)	2 instruction and 6 training sessions, 60–90 min	- Young > old	-Young > old (cognitive assessments)	Magnification	Yes but young adults still improved more
*Lövden et al. 2012	N=29/29 (young/old) Age=22.5/66.9 years	EM (method of loci)	2 instruction and 3–7 training sessions	- High BP > Low BP (instruction gains) - Young > Old (training gains)		Magnification	Yes for instruction gains
Rebok et al. 2013	N=629 Age=73.5 years Education=13.7 years	EM (method of loci, visualization strategies, and categorization methods)	4 instruction and 5 training sessions, 60–75 minutes	- Old = young		Neither	No
Verhaeghen et al. 1996	N=63/76 (young-old/old-old) Age=18.7/66.5 years Education=College/12.8 years	EM (method of loci)	3 sessions, 75–120 min	- Young > old (speed of mental operations, number of list rehearsals, and associative memory)		Magnification	Yes at certain training tasks (findings not included)
Cepeda et al. 2001	N=109 Age=21–82 years	EF (two nonswitch blocks and eight switch blocks)	2 sessions	- Old > young (reduction in switch costs) - Old = young (cue-target interval)		Compensation	No
Dorbath et al. 2011	N=85/91 (young/old) Age=24.1/66.3 years	EF (four continuous counting tasks)	4 sessions, 45 min	- Old > young (error rate) - Low BP > high BP (error rate)	- N/A	Compensation	Yes for training gains accuracy
*Karbach et al. 2009	N=56/56 (young/old) Age= 22.4/68.7 years	EF (two single-task and two mixed-task blocks)	4 sessions	- Old = young (switching costs) - Old = young (far-transfer tasks)	- Old > young (reduction in mixing costs)	Compensation	No
Karbach et al. 2017	N=42/42 (young/old) Age=22.0/68.7 years	EF (two mixed-task blocks)	4 sessions	- Old > young - Low BP > high BP	- Old > young (task-switching task) - Low BP > high BP (task-switching task)	Compensation	No
Kray et al. 2017	N=81/82 (young/old) Age=21.9/70.8 years No difference in education (p=0.11)	EF (4 tasks with different switching conditions)	4 sessions, 30–40 min	- Young > old (reduction in switch costs)	- Young > old (reduction in mixing costs)	Magnification	No
*Sandberg et al. 2014	N=29/30 (young/old)	EF (updating, shifting, and inhibition tasks)	15 sessions, 45 min	- Young = old	- Young = old (updating and inhibition tasks) - Young > old (two complex working memory tasks)	Neither	No

The 23 studies chosen for review, each of which examines the impact of baseline performance (BP) and/or age on cognitive training (CT) improvements. Studies are organized by cognitive domain, then alphabetically.

^*denotes studies that adopted an adaptive CT regimen.

3. Results

3.1. Magnification vs. Compensation Effect Across All Cognitive Domains

Overall, the magnification effect has been observed more frequently in 13 out of 23 CT studies (Table 3). In studies examining the impact of age on CT effectiveness, 10 out of 16 studies found a magnification effect. On the other hand, 1 out of 1 study examining the impact of BP on CT effectiveness observed a compensation effect. Finally, for studies examining the effect of age and BP on improvements in CT, 3 out of 6 studies observed a compensation effect.

Table 3.

Past findings of the compensation and magnification effects across the cognitive domains of episodic memory (EM), executive functioning (EF) and working memory (WM).

	Magnification Effect			Compensation Effect
	EM	WM	EF	EM	WM	EF
Age	4/5	5/7	1/4	0/5	1/7	2/4
Baseline Performance	0/0	0/1	0/0	0/0	1/1	0/0
Age and Baseline Performance	1/1	2/3	0/2	0/1	1/3	2/2
Total	5/6	7/11	1/6	0/6	3/11	4/6

The strongest support for the magnification effect has been found in EM CT studies, while the compensation effect has been observed across several cognitive training studies in WM and EF.

3.2. Episodic Memory Training

Of the 6 EM studies examining the impact of BP and/or age on CT improvement, 5 out of 6 observed a magnification effect. Of these 6 studies, 5 specifically analyzed the impact of age on CT effectiveness, while 1 analyzed the impact of BP and age.

3.3. Working Memory Training

Of the 11 WM CT studies examining the impact of BP and/or age on CT improvement, 7 out of 11 studies supported the magnification effect while 3 out of 11 studies supported the compensation effect. Evidence for a compensation effect in WM CT was most strongly supported by studies with WM interventions targeting only older adults (2 out of 5).

3.4. Executive Function (Task-Switching) Training

Of the 6 EF studies with an emphasis on task-switching examining the impact of BP and/or age on CT improvement, 4 supported a compensation effect, 1 supported a magnification effect, and 1 produced inconclusive results.

4. Discussion

The current review suggests that EM CT is most beneficial for younger, higher BP individuals, while EF (task switching) CT is most beneficial for older, lower BP individuals. On the other hand, WM CT seems to provide more varied results requiring further consideration of characteristics of the CT regimen and individual profile. Moreover, a more critical evaluation of why these patterns have emerged is critical to understanding how we can best maximize CT-related benefits in each individual.

4.1. Episodic Memory Requires Intact Cognitive Abilities to Improve

A common training technique in EM is instruction and practice in the method of loci, a mnemonic device that depends on spatial relationships between “loci” (i.e. rooms in a familiar building) to organize and recall memory content (Qureshi et al. 2014). Literature in this topic has consistently observed the magnification effect in the context of age-group differences between subjects (see Table 2), suggesting that acquiring and utilizing the method of loci technique to induce EM improvement depends upon efficient existing cognitive resources.

Additionally, neurocognitive factors specific to the EM domain may help explain this phenomenon. EM is one of the first cognitive domains to suffer with aging, and thus is less easily recovered in older individuals who have less neuroplasticity (Levine et al. 2002). In fact, a study by Levine et al. (2002) found that while recall of episodic details was impaired in older adults, production of semantic details remained intact, supporting the notion that EM is particularly sensitive to cognitive aging. Thus, while literature has provided more evidence of plasticity in cognitive domains such as WM, it seems that the process of cognitive aging specifically impacts EM in a manner such that it cannot be easily recovered with practice on training tasks.

4.2. Cognitive Disuse as a Moderator for Improvements in WM Training

One of the more notable findings of WM training studies is support for the compensation effect hypothesis, particularly in low BP individuals among an older subject body. The magnitude of training gains observed in lower baseline-ability adults in WM CT is promising because it indicates the extent to which areas of the brain in charge of WM (e.g. the prefrontal cortex) are capable of plasticity up until very old age. To explain this phenomenon, Zinke et al. (2012) discussed the disuse hypothesis, which suggests that lack of practice in WM tasks leads to suboptimal use of cognitive resources that results in cognitive decline; thus, low-performing individuals were able to reactivate the neural resources underlying WM through practice during the adaptive training regimen. On the other hand, participants with high WM baseline capacity had maintained their WM ability prior to training through frequent everyday practice, and thus, did not profit as much from the CT regimen that focused on repetitive practice of these processes.

4.3. Task Difficulty Helps Explain Differential Training Gains

The supply-demand mismatch hypothesis, explained by Lövden et al. (2010), also helps to explain the occurrence of the compensation effect in WM training. In this model, a mismatch between available cognitive resources and task demands promotes changes in cognitive performance. When the amount of mismatch is ideal, repeated performance in training tasks induces neural changes that promote cognitive plasticity. Most of the WM CT regimens examined in this paper were adaptive, meaning that they began easy and increased in difficulty as subjects progressed. We speculate that several of the WM training tasks favored the ideal amount of supply-demand mismatch (i.e. demands exceeding available capacity) for low baseline-ability participants. As WM performance improved for these individuals, so too did task difficulty, inducing an ideal amount of mismatch throughout the study to promote continual improvement. On the other hand, high baseline-ability participants in WM began closer to the maximum difficulty level of the adaptive WM CT regimens. Thus, these individuals may have increased their available cognitive resources more quickly than the task increased in difficulty, producing insufficient mismatch between task demands and available neural capacity and ultimately inducing less improvement.

The supply-demand mismatch hypothesis can also be extended to the EM domain, in which research shows that low baseline ability participants did not benefit as much from CT as did high baseline ability participants. As previously noted, EM is known to be extremely sensitive to cerebral aging (Shing et al. 2010; Nyberg et al. 2012). EM training in the method of loci may induce too large a difference between task difficulty and cognitive resources for older, lower baseline profile adults, preventing these individuals from applying resources suited to the task. Several studies indeed speculated that the high difficulty of the method of loci training techniques explained the low training gains in older adults (Kliegl et al. 1990; Verhaeghen and Marcoen, 1996). Furthermore, Borella et al. (2017), in a review examining the effects of four WM training regimens, substantiated the importance of considering task difficulty when determining who improves most in CT. They found that in near-transfer to structurally dissimilar tasks, younger individuals improved more for active information processing (more demanding) tasks (e.g. backward digit span), while older individuals and lower baseline-ability individuals improved more for more passive (less demanding) tasks (e.g. forward digit span).

The supply-demand hypothesis offers a lens through which to better understand the training conditions that support optimal gains, stressing the importance of task difficulty as a moderating variable for improvement in individuals with different starting profiles. The difficulty of the training task is especially worth considering given that it may relate to the amount of transfer that is observed (Stamenova et al. 2014). Finding the optimal task difficulty for each individual, if accomplished, has the potential to maximize positive individual outcomes following CT.

4.4. Process-Based vs. Strategy-Based Training Regimens

It is important to consider the consistent finding across the majority of studies that implemented task-switching (EF) CT regimens; older and lower BP individuals tended to derive the most training and near-transfer improvements. To account for the consistent occurrence of the compensation effect in EF CT, Karbach et al. (2017) distinguished between strategy-based and process-based CT regimens. Strategy-based interventions (i.e. mnemonic training) tend to be more complex CT regimens that target specific processing capacities, whereas process-based interventions (i.e. task-switching paradigm) target more general processing capacities, such as WM or EF, typically resulting in more robust transfer effects. It is plausible that process-based interventions might be more beneficial to lower BP individuals than strategy-based interventions because they can tap into existing cognitive resources in other domains that may be stronger than the cognitive ability for which they perform at low BP. The idea that low-functioning individuals in task-switching can draw on existing efficient cognitive resources in other cognitive domains in order to improve in EF-ability is substantiated by findings that age-related decline in task-switching is independent of WM and perceptual speed abilities (Cepeda et al. 2001). Neuroimaging evidence suggests that EF is subserved by distributed neural networks including the frontal, striatal and parietal regions – as opposed to EM that is supported by a more localized network involving medial temporal structure. Therefore, older individuals may compensate for baseline deficiencies in task-switching via other stronger processing capacities within the greater EF network in order to induce functional plasticity.

4.5. The Ceiling and Floor Effects as Modifiers of Training Gains

While the supply-demand mismatch hypothesis offers a feasible explanation for past findings, it is important to recognize that lack of improvement in WM CT regimens by high baseline-ability individuals in several studies may also be attributed to the characteristics of the WM cognitive domain. Miller (1956) found that there was a limit to the capacity of short-term WM; adults could only store between 5 to 9 items in their short-term memory (the 7 +/− 2 rule). While this rule does not necessarily apply to WM CT tasks, which vary in duration and form, it does suggest a limit in plasticity in some spheres of WM. In fact, the majority of studies reporting higher improvement in WM training in low-capacity individuals observed a ceiling effect; that is, individuals with a high baseline profile in WM could not increase their test score throughout the whole CT regimen because they had already reached the highest attainable score on a task (Li et al. 2008; Zinke et al. 2012). The occurrence of a ceiling effect was also observed in near-transfer tasks for younger adults following a multi-domain CT regimen (Schmiedek et al. 2010). Moreover, in the same way that the ceiling effect can explain compensation effects, the floor effect — the opposite of the ceiling effect — can also be responsible for several observed magnification effects (von Bastian et al. 2013; Dorbath et al. 2011).

A recent paper by Smoleń et al. (2018) also casts doubt on compensation account findings, especially those that look at the correlation between BP and CT gains, explaining they are subject to the widely known methodological error called “regression to the mean.” The authors conclude that studies observing a negative correlation between BP on a task and improvement on that task as evidence for the compensation effect (i.e. Zinke et al. 2012 & Zinke et al. 2014) should be taken with caution. Despite the conclusions of Smoleń et al. (2018), it is important to consider that the majority of studies observing a compensation effect not only found enhanced training improvements in low functioning individuals but also found enhanced transfer benefits in low functioning individuals. For example, in the domain of WM, Li et al. (2008) found that older individuals improved more in not only training tasks but also in a two-back plus processing test, while Karbach et al. (2017) found both enhanced training and transfer benefits in low BP and older individuals following task-switching CT. These findings suggest that near- and far-transfer gains can help to distinguish compensation effects derived from the ceiling effect and/or regression to the mean statistical artifacts from compensation effects that represent true enhanced benefits in lower baseline profile individuals.

4.6. Limitations

One limitation of the current review is the grouping of the variables of age and BP when distinguishing between the magnification and compensation effects. While these two variables are typically correlated — that is, that younger individuals and higher BP individuals show similar patterns of training improvement compared to older and lower BP individuals — this is not always the case. For example, in an evaluation of three WM interventions, Guye et al. (2017) found a positive association of age and BP with training improvements. Similarly, Zinke et al. (2012) observed that while low BP individuals improved more than high BP individuals on all training tasks, older individuals did not improve more than younger individuals on these same tasks. These findings suggest that differences in cognitive resources between high BP and low BP individuals in a cognitive domain may be different than those between old and young individuals. Additionally, the lack of quantitative analysis between improvements in different CT regiments is a limitation that should be explored in future research.

5. Conclusions

Our review suggests that factors specific to the cognitive domain being tested can help to explain which individuals improve most from training. EM is typically the first cognitive domain to suffer with aging, and thus is less easily recovered in older individuals. On the other hand, the disuse hypothesis explains how lack of practice in WM tasks leads to suboptimal use of neural resources underlying WM that can be reactivated through an adaptive training regimen, accounting for increased improvement in older individuals in several studies. Moreover, in accordance with the supply-demand mismatch hypothesis, it seems that the difficulty of the training tasks may play a role in determining who improves most from CT, illustrating the importance of considering not only the individual characteristics of the subject body completing the training activities but also the characteristics of the training activities themselves in an attempt to maximize training effectiveness. Finally, our review suggests that process-based CT regimens — such as EF CT regimens focused on task-switching — are particularly amenable to inducing plasticity in older, lower BP individuals. Given that the findings of the studies discussed in this review were mainly post-hoc, future research should more clearly stratify individuals according to their baseline cognitive abilities and assign skill-specific CT regimens in order to directly answer the question of how individual differences impact CT effectiveness. These studies would stratify individuals into training groups according to their baseline cognitive profile and compare training-related improvements. We are finalizing a pilot study that stratified individuals into groups based on their BP in the domains of EM, inhibitory control, and processing speed to better understand how CT-related improvements can be maximized in each individual.

Highlights.

• Individuals participating in the same CT regimen differ greatly with respect to their outcomes.

• The impact of age and BP on CT improvement is domain specific. While EM training is particularly suited to younger, higher baseline ability individuals, WM and EF training are more amenable to improvements in older, lower baseline ability individuals.

• Our findings inform the design of future interventions that assign specialized, skill-specific cognitive training regimens to maximize gains in each individual