Edited by:
Reviewed by:
*Correspondence:
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Recent attention has focused on the benefits of cognitive training in healthy adults. Many commercial cognitive training programs are available given the attraction of not only bettering one’s cognitive capacity, but also potentially preventing age-related declines, which is of particular interest to older adults. The issue of whether cognitive training can improve performance within cognitive domains not trained (i.e., far transfer) is controversial, with meta-analyses of cognitive training both supporting and falsifying this claim. More support is present for the near transfer (i.e., transfer in cognitive domain trained) of cognitive training; however, not in all studies. To date, no studies have compared working memory training to training higher-level processes themselves, namely logic and planning. We studied 97 healthy older adults above the age of 65. Healthy older adults completed either an 8-week web-based cognitive training program on working memory or logic and planning. An additional no-training control group completed two assessments 8-weeks apart. Participants were assessed on cognitive measures of near and far transfer, including working memory, planning, reasoning, processing speed, verbal fluency, cognitive flexibility, and creativity. Participants improved on the trained tasks from the first day to last day of training. Bayesian analyses demonstrated no near or far transfer effects after cognitive training. These results support the conclusion that performance-adaptive computerized cognitive training may not enhance cognition in healthy older adults. Our lack of findings could be due to a variety of reasons, including studying a cohort of healthy older adults that were performing near their cognitive ceiling, employing a training protocol that was not sufficient to produce a change, or that no true findings exist. Research suggests numerous study factors that can moderate the results. In addition, the role of psychological variables, such as expectations and motivation to train, are critical in understanding the effects of cognitive training.
Maintaining cognitive functioning is a hallmark of successful aging. Cognitively high-functioning older adults are more socially engaged, less lonely, less physically frail, and have higher overall quality-of-life ratings than cognitively lower-functioning older adults (
Within the cognitive training literature, it is well-established that training a specific cognitive ability results in improvements in that task (i.e., target of practice) and generally in similar tasks (i.e., near transfer). Near transfer is defined as improvement in a task that is within the same cognitive domain as the trained task (
Similarly, in healthy older adults, near transfer after cognitive training is largely reliable; however, far transfer is more contested. Improving fluid intelligence is typically the main far transfer target for studies of working memory training. Therefore, we aimed to improve fluid intelligence via near transfer by training higher-level skills themselves, namely logic and planning. We compare the effects of directly training these higher-level skills to that of training working memory.
Although numerous forms of cognitive training exist, working memory training has garnered the most attention in healthy adults. The hypothesized objective of working memory training is to enhance an individual’s core ability to temporarily store and process information. Working memory training aims to increase the core capacity and processing efficiency of working memory, factors which are important for day-to-day cognitively demanding activities such as language, reasoning, problem solving, reading comprehension, and more general aspects of knowledge-based and fluid intelligence (
Another potential approach to cognitive training is to more directly target higher-order executive functioning processes. Executive functioning is a broad term that describes higher-order cognition, including reasoning, planning, and cognitive flexibility. Working memory is also often included as a domain of executive functioning. In a latent variable analysis, task-switching had a strong and significant relationship with performance-based instrumental activities of daily living in a sample of healthy older adults aged 60–90 years old, indicating that executive functioning may impact everyday activities in the elderly (
Logic and planning is an aspect of executive functioning involved in decision making and problem solving. Only one study to date has included a planning intervention with older adults.
To our knowledge, no prior studies have examined the impact of specifically training higher-level logic and planning and compared it to the most widely used single domain targeted training, working memory training, in healthy, community dwelling, older adults (>65). The primary goal of the study was to identify whether working memory training versus logic and planning training differentially impacts cognitive performance on a variety of near and far transfer tasks. Furthermore, we aimed to discover whether either type of training benefited healthy older adults relative to usual activities (i.e., a no-contact passive control condition). Given previous findings regarding the beneficial impacts of working memory training for healthy older adults, we expected that our working memory trainees would demonstrate improvements in at least the trained domain of working memory (i.e., near transfer to working memory tasks), and potentially in measures of far transfer (i.e., executive functioning, reasoning, processing speed, and creativity tasks). We also anticipated that our logic and planning trainees would demonstrate improvements in tasks tapping logic and planning (i.e., near transfer to planning and non-verbal reasoning), and potentially in far transfer tasks (i.e., working memory, processing speed, verbal fluency, cognitive flexibility, and creativity). Although, it can be argued that working memory, processing speed, and cognitive flexibility may not be far transfer tasks for logic and planning training, we grouped these tasks under far transfer, as those cognitive domains were not specifically targeted or trained. Last, we anticipated that both training groups would demonstrate these improvements relative to the passive control group.
Healthy adults over the age of 65 were recruited from the community in Calgary, AB, Canada. Informed consent was obtained from each participant online, as well as in-person. Study procedures were approved by and carried out in accordance with the University of Calgary Conjoint Faculties Research Ethics Board. Potential participants completed an online screening questionnaire to assess eligibility. Exclusion criteria were age less than 65, lack of English proficiency, history of head trauma, brain fever, self-reported neurological or psychiatric illness, dementia, or altered consciousness, use of benzodiazepines or illicit drugs in past 3 months, current visual, auditory, or motor impairment, cardiovascular condition, respiratory problems, and a score less than 27 on the Mini Mental State Examination (MMSE). Last, all participants needed access to a high speed internet connection.
Individuals meeting inclusion criteria were then invited to attend an in-person, individual cognitive assessment. Prior to the assessment, participants completed online questionnaires assessing demographics, mood, physical activity, and sleep quality. Study eligibility was further confirmed in-person. After the assessment, participants were quasi-randomized (accounting for sex distribution) to one of three groups: working memory training, logic and planning training, or a passive (i.e., no-training) control group. A research assistant introduced participants in the two training groups to the BrainGymmer website and games. Both training groups were instructed to train, at a time and location of their convenience, for approximately 30 min per day, 5 days a week, for 8-weeks, totaling 20 h of training. Adherence to training was monitored weekly and phone calls or emails followed if participants deviated from the protocol. All groups completed a second assessment after 8 weeks. Participants in the training groups received no remuneration, but were entered into one of several draws held throughout the study. We did not pay participants in the two training groups, as research suggests remuneration for training reduces participants’ intrinsic motivation to train, which is one necessary factor for training to work (
The working memory and logic and planning training games were provided by BrainGymmer
The three games in this domain primarily targeted maintenance and manipulation of information. In the Multi-Memory game, a square grid was presented and different tiles were placed on the grid. Participants had to remember the placement of the tiles, which then disappeared and were replaced by a distractor pattern. For each trial, participants had to recreate the original pattern of tiles. The size of the square grid and number of tiles changed as a function of performance. In the Moving Memory game, pairs of cards were shown with the same image, but with different numbers at the bottom. The cards were then flipped and scrambled with only the number on the card visible. For each trial, participants had to pick the two cards with the same image, until no pairs of cards remained. The number of pairs to be remembered changed as a function of performance. Last, in the
To investigate the relationship between the tasks, day 1 scores from each task were correlated: specifically, highest n-back achieved for the n-back task, and mean score and difficulty level for Multi-Memory and Moving Memory. Day 1 scores among the three working memory games were correlated (
The three games in this domain primarily targeted planning, reasoning, and problem solving abilities. In the Square Logic game, a grid of numbered squares was presented. The objective was to stack the squares using the rule that squares can only be stacked onto squares that are one point higher or lower in value. The number of squares to stack changed as a function of performance. In the Out of Order game, a series of squares were presented, each with different shapes, patterns within the shape, color, and number of shapes. The objective of this game was to rearrange the squares so that each square matched at least one characteristic of the square adjacent to it. The number of squares to arrange changed as a function of performance. Last, in the Patterned Logic game, a pattern with missing pieces was presented. Participants had to choose the correct piece to complete the pattern. Pattern complexity changed as a function of performance.
Correlations among the games were conducted based on day 1 mean score and difficulty level. The Square Logic game mean score was associated with Pattern Logic game mean score (
At baseline, participants completed an online self-report demographics questionnaire and inventories of state characteristics (mood, physical activity, sleep quality) commonly known to impact cognitive performance. Mood was measured with the Beck Depression Inventory-II (BDI-II;
At baseline and post-training, training group participants provided motivation ratings. Specifically, participants responded to the question, “How motivated would you say you are/were to complete the cognitive training component of this study?” by marking a 7-point scale ranging from no motivation to substantial motivation.
At baseline and post-training, all participants underwent testing of working memory, processing speed, executive functioning (logic and planning, verbal fluency, cognitive flexibility), creativity, and reasoning. Measures were chosen to tap a variety of near and far transfer cognitive processes, for their use in previous investigations of working memory training, their sensitivity to age-related differences in cognitive performance, and their reliability, validity, and utility in the cognitive assessment of healthy older adults. Cognitive tasks were grouped by cognitive domain based on conceptual relationships among specific measures.
Working memory was assessed with the Digit Span subtest of the Wechsler Adult Intelligence Scale-IV (WAIS-IV;
Working memory was also measured with the Automated Operation Span (Aospan) task (
The Tower Test (TT) was utilized to examine spatial planning, rule learning, and inhibition of impulsive responding (
The Raven’s Advanced Progressive Matrices (RAPM;
Processing speed was measured using multiple tasks. Raw scores from the Symbol Search subtest of the WAIS-IV (
Processing speed was also assessed with Trail Making Test (TMT) Items 1, 2, and 3 and Color-Word Interference Test (CWIT) Items 1 and 2, all of which are subtests of the Delis Kaplan Executive Function System (D-KEFS;
From the verbal fluency subtest of the DKEFS, we utilized the letter fluency task (LF) to measure speeded verbal generation of words belonging to a particular phonemic category.
Cognitive flexibility was assessed with four tasks from the DKEFS. Specifically, we used Color-Word Interference Test Item 3 (CWIT 3) which measures inhibition, and Color-Word Interference Test Item 4 (CWIT 4) which measures task-switching performance. Trail Making Test 4 (TMT 4) and Design Fluency Test 3 (DF 3) were also included to assess task-switching and inhibition.
The first 2-items from the Design Fluency (DF) subtest of the DKEFS were used (DF 1 and DF 2) to assess initiation of problem-solving behavior, visual pattern generativity, creativity in drawing new designs, and inhibition of previously drawn responses.
To analyze the demographic, mood, sleep, physical activity, and baseline cognitive data, analyses of variance (ANOVAs) and chi-squared tests were conducted. To analyze the cognitive training data, correlations and Bayesian Repeated Measures Analyses of Variance (Bayes RM-ANOVAs) were conducted using the JASP statistics package, version 0.7, available online at
Within conceptually related cognitive domains, correlations (two-tailed) were conducted among baseline cognitive outcome measures. Based on moderate to strong correlations among some outcome measures within cognitive domains, composites were created by adding z-scores of tasks which significantly correlated (α < 0.05) within a domain. Specifically, within the working memory domain, Digit Span Total scores and Aospan scores were not significantly correlated so were analyzed individually. A processing speed composite was created using TMT 1, TMT 2, TMT 3, CWIT 1, and CWIT 2 scores, whereas Symbol Search scores were analyzed independently. A cognitive flexibility composite was created from CWIT 3 and CWIT 4 scores, and remaining tasks within the cognitive flexibility domain were analyzed separately. A DF composite was created using both tasks within that domain (DF 1 and DF 2). The logic and planning, verbal fluency, and reasoning domains were composed of single tasks analyzed individually.
Bayesian Repeated Measures Analyses of Variance were utilized to compare group differences across time for the motivation data and each cognitive composite or individual task. For the cognitive data, first, a Bayes RM-ANOVA including all three groups was conducted. If that Bayes RM-ANOVA was significant, we followed up to investigate if the training groups improved compared to the passive control group, which assessed test-retest fluctuations. If this was significant, we followed up by comparing the two training groups.
The JASP statistical analysis program generates Bayes factors using default prior probabilities; however, rather than producing a probability estimate in support of the null hypothesis based on an arbitrarily determined cut-off of statistical significance, the Bayes approach compares likelihood estimates of the obtained data occurring under the null (01) versus alternative (10) hypothesis. Advantages and specific procedures of the Bayesian RM-ANOVA approach, including the use of default priors, are extensively discussed elsewhere (e.g.,
Screening, eligibility, consent, and completion rates for the working memory training, logic and planning training, and no-training control groups are presented in
Participant characteristics are presented in
Demographics, mood, sleep, physical activity, cognition, and training characteristics.
Working memory training | Logic and planning training | Passive control | ||
---|---|---|---|---|
N | 36 | 32 | 29 | |
Age | 70.39 (4.54) | 70.81 (4.98) | 70.24 (4.48) | |
Range | 65–86 | 84–65 | 65–78 | |
Sex (% female) | 64 | 69 | 69 | |
Ethnicity (% Caucasian: Asian: Other) | 94: 6: 0 | 88: 13: 0 | 86: 7: 7 | |
Marital status (% coupled) | 72 | 69 | 62 | |
Education (years completed) | 15.43 (3.48) | 15.44 (2.86) | 15.52 (2.86) | |
Range | 7–23 | 10–21 | 9–22 | |
Employment (% retired) | 86 | 81 | 83 | |
Income (% <$50,000: $50,000–$95,000: >$95,000) | 31: 47: 22 | 55: 23: 23 | 32: 50: 18 | |
Beck Depression Inventory | 5.61 (6.55) | 3.66 (4.29) | 4.97 (6.12) | |
Range | 0–24 | 0–18 | 0–26 | |
Beck Anxiety Inventory | 3.33 (4.42) | 2.25 (3.22) | 2.48 (3.00) | |
Range | 0–18 | 0–15 | 0–10 | |
PSQI Total | 4.86 (3.03) | 4.47 (3.07) | 4.41 (3.09) | |
Range | 2–12 | 0–12 | 0–15 | |
RAPA Aerobics | 4.83 (1.75) | 5.56 (1.27) | 5.14 (1.60) | |
Range | 0–7 | 4–7 | 2–7 | |
RAPA Strength | 1.56 (1.30) | 1.69 (1.26) | 1.62 (1.24) | |
Range | 0–3 | 0–3 | 0–3 | |
MMSE | 28.89 (0.95) | 28.67 (1.00) | 29.03 (0.94) | |
Range | 27–30 | 27–30 | 27–30 | |
WASI-II 4-item composite | 111.75 (13.24) | 113.94 (10.28) | 112.52 (11.88) | |
Range | 66–133 | 91–135 | 96–148 | |
Training time (hours) | 19.01 (2.14) | 19.44 (2.42) | ||
Range | 14.23–22.68 | 12.32–24.87 |
Importantly, groups had similar MMSE scores [
At baseline, the null hypothesis suggesting groups were equally motivated to complete the training was supported,
Participants completed the cognitive pre-assessment and cognitive post-assessment close to beginning and ending their training [mean = 1.25 days before training,
Means, standard deviations, and Bayes factors indicating support for the null versus alternative hypotheses are presented in
Means and Standard Deviations before (T1) and after (T2) training period, and Bayes factors1 of time and interaction effects.
Domain | Task | WMT T1 | WMT T2 | LPT T1 | LPT T2 | PC T1 | PC T2 | Time | Group × Time |
---|---|---|---|---|---|---|---|---|---|
mean ( |
mean ( |
mean ( |
mean ( |
mean ( |
mean ( |
( |
( |
||
Motivation | 5.57 (1.37) | 5.91 (1.63) | 5.91 (1.63) | 5.72 (1.49) | - | - | 1.55 | 8.50 | |
Working memory | Aospan | 24.72 (14.06) | 29.36 (13.92) | 26.45 (14.56) | 31.53 (18.12) | 26.41 (15.48) | 32.31 (16.36) | 0.01 | 0.43 |
DST | 27.83 (4.66) | 28.67 (4.67) | 27.47 (6.51) | 28.88 (5.93) | 26.32 (5.04) | 28.41 (4.98) | 4.30 | 116.25 | |
Planning | TT | 16.34 (3.50) | 18.25 (2.49) | 16.69 (4.88) | 18.28 (4.63) | 16.17 (5.09) | 18.97 (3.52) | <0.001 | 0.07 |
Reasoning | RAPM | 8.22 (2.81) | 8.33 (2.92) | 8.28 (2.98) | 8.55 (2.73) | 7.41 (1.90) | 8.21 (2.73) | 3.72 | 94.92 |
Processing speed | Composite | -0.35 (3.23) | -0.40 (3.56) | -0.09 (4.58) | -0.16 (4.26) | 0.52 (2.55) | 0.67 (3.16) | 6.48 | 161.83 |
SS | 28.11 (5.50) | 29.56 (6.73) | 29.78 (6.63) | 29.97 (7.73) | 26.69 (5.93) | 27.45 (5.57) | 1.5 | 14.88 | |
Verbal fluency | LF | 40.06 (10.82) | 43.34 (10.85) | 39.75 (10.20) | 43.25 (11.79) | 38.79 (10.94) | 43.83 (13.89) | 0.002 | 0.07 |
Flexibility | Composite | -0.05 (1.70) | -0.15 (1.88) | -0.01 (2.01) | 0.05 (1.83) | 0.08 (1.70) | 0.14 (1.69) | 6.44 | 142.79 |
DF 3 | 7.31 (2.14) | 7.92 (2.38) | 8.47 (2.87) | 9.00 (2.34) | 7.14 (2.34) | 8.08 (2.47) | 1.23 | 2.98 | |
TMT 4 | 90.26 (26.54) | 81.81 (27.65) | 86.22 (40.30) | 82.88 (41.49) | 92.31 (37.94) | 87.10 (29.39) | 1.58 | 39.40 | |
Creativity | Composite | -0.25 (1.97) | 0.09 (1.77) | 0.45 (2.09) | 0.33 (2.14) | -0.20 (1.50) | -0.14 (1.68) | 6.58 | 6.58 |
Within the working memory domain, Aospan and Digit Span Total scores were not significantly correlated (
The TT assessed visual-spatial planning. For this task, evidence was provided for an effect of time,
The RAPM task was used to measure reasoning. Data supported the null hypothesis for an effect of time,
Within the processing speed domain, outcome scores among TMT Conditions 1, 2, and 3, and CWIT Conditions 1 and 2 were positively correlated (
The verbal fluency task data revealed evidence for an effect of time,
Cognitive flexibility represents processes reliant on inhibition and task-switching. CWIT3 and CWIT4 were significantly correlated (
Creativity tasks were DF conditions 1 and 2, both which require drawing new images with the main restriction being to not repeat drawings. These two tasks were significantly correlated (
Although some level of cognitive decline is a natural part of aging, slowing, preventing, or ameliorating cognitive decline is a key goal of healthy living. In this study, we investigated near and far transfer of two active conditions, working memory training and logic and planning training, to a passive control group, in healthy community dwelling seniors (age 65 plus). Working memory training was chosen as it is the most widely studied process-specific training and is conceptually related to fluid intelligence (the most common far transfer goal). As a comparison condition, we also, for the first time, trained logic and planning to investigate its potential to transfer to fluid intelligence as near transfer. Given the positive benefits of greater fluid intelligence to healthy living, this is an important conceptual question. In this sample of 97 healthy older adults, we found only evidence for improvements on the trained tasks, and none for near or far transfer after cognitive training.
After training for an average of 19 h, we found both groups performed substantially better on all the training tasks compared from the first day of training to the last day of training. However, surprisingly, we found no evidence for near transfer. Neither maintenance nor manipulation, as measured by the Digit Span task, or working memory capacity, as measured by the Aospan task, improved relative to the control group. Similarly, the logic and planning training group did not improve on the Tower task, which measures planning. Furthermore, scores on the Raven’s Progressive Matrices task, a common measure of reasoning, did not improve after logic and planning training.
In addition to near transfer, we also investigated far transfer for both training conditions. Far transfer for working memory training was assessed by measures of processing speed, verbal fluency, cognitive flexibility, planning, creativity, and reasoning. No-training effects were found for far transfer for this group. Far transfer for the logic and planning training group was assessed by measures of processing speed, working memory, verbal fluency, cognitive flexibility, and creativity. Although whether all of those cognitive processes are far transfer for logic and planning training can be debated; nevertheless, no-training effects were found for these tasks for this group either.
Our null findings were not due to lack of motivation, as we measured pre-post motivation for training and found no differences between our two training groups at baseline or over time. Additionally, motivation to train was high (pre-post means above 5.5) on a scale ranging from one (extremely unmotivated) to seven (extremely motivated).
Our null findings fit into a mixed literature on the efficacy of cognitive training to improve cognition in healthy adults. Moreover, meta-analyses reveal conflicting results regarding cognitive gains after training, and study inclusion and analytic techniques have been debated. Three meta-analyses have specifically investigated cognitive training in healthy older adults. Of particular relevance, a meta-analysis of working memory and executive functioning training in older adults (mean age > 60) found that the two types of training did not reliably differ in their ability to change cognition (
A second meta-analysis in healthy older adults (mean age > 50) found that compared to the active control training condition, cognitive training improved working memory and processing speed, and a composite measure of cognitive function (
Furthermore, in the area of cognitive training, different statistical analyses have resulted in different results for meta-analyses in younger healthy adults. Focusing on individuals 18–50 years old, a meta-analysis of fluid intelligence gain after dual n-back working memory training reported a small but statistically significant effect of active training compared to control conditions (
Moreover, complicating the mixed literature,
As discussed above, of key theoretical interest is whether cognitive training transfers to fluid intelligence, an idea which is highly contested and debated. In this study, we used working memory training, which is theoretically linked to fluid intelligence, as well as training on higher-level cognitive processes themselves, such as planning, reasoning, and problem solving. In this study, we found that neither training on a cognitive process (i.e., working memory), which is highly correlated with fluid intelligence, nor training on tasks conceptually similar to fluid intelligence, produced gains in Raven’s Progressive Matrices, the most widely used measure of fluid intelligence.
There are a number of potential reasons why we failed to find near and far transfer effects after cognitive training, other than this being a true finding. First, the literature suggests that for cognitive training to be effective, the training has to be challenging but not frustrating (
Although the meta-analyses described above reveal a complex, controversial research area, these studies also suggest possible avenues to reconcile the different findings in the literature. Potential factors that influence the effects of training on transfer include the type of cognitive training, age of the sample, amount of training, presence or absence of randomization, type of control group, geographic location of study, remuneration for participation, and publication type (
Limitations of this study include sample size. Although we recruited approximately 30 participants per group, which is larger than many studies in the field, effect sizes for cognitive training studies are generally small; therefore, we were likely underpowered. Nevertheless, well-conducted small studies are still an asset to their field, and can also be entered into meta-analyses. Another limitation of this study included the difficulty in having training programs and cognitive tasks that isolate very specific processes. This is particularly difficult in cognitive training programs assessing working memory and higher-level cognitive processes. The logic and planning training involved working memory processes and the working memory training involved some level of reasoning and learning. However, the extent to which the different processes were emphasized was substantial in the two types of training. Last, we created cognitive composite scores by analyzing correlation patterns between baseline task scores for tasks within a domain. However, the pattern of correlation and relationship between tasks could change after training. Despite these limitations, this study has a number of strengths including using two active training groups plus a passive control, a broad array of near and far transfer measures, and ensuring thorough measurement of demographic, cognitive, health, lifestyle, and other factors that could be related to group differences in cognitive training.
In summary, the cognitive training literature in healthy adults is full of conflicting findings and methodological debates. This study found only practice effects and no near or far transfer after cognitive training in healthy older adults. Whether cognitive training leads to cognitive gains in a consistent manner is yet to be shown, and if gains are to be found they are likely to be small. Future research evaluating mediators and moderators of change will help determine if there are sub-populations of individuals for whom cognitive training may be helpful. Given that better cognitive function in older adults is associated with better physical health and social outcomes (
VG wrote project protocol, supervised data collection and data analysis and wrote the majority of the manuscript. LL-S conducted the analysis and wrote sections of the paper under the supervision of VG. VG and LL-S made an intellectual contribution.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We would like to thank May Luu, Aiko Dolatre, and Danielle Lefebvre for their help with study initiation and coordination.