Home › Psychology

📖 13 min read

Working Memory Training: What the Science Actually Shows in 2025

What we’re looking at

Working memory—the mental workspace that lets you hold and manipulate information for a few seconds—is a fundamental cognitive skill. It helps you remember a phone number long enough to dial it, follow multi-step instructions, or solve math problems in your head. For years, commercial brain-training programs have promised that targeted working memory exercises can not only boost this core capacity but also improve intelligence, academic performance, and everyday functioning. With global cognitive training markets now worth billions, the question isn’t whether these programs are popular—it’s whether they actually work.

This evidence review synthesizes the current state of research on working memory training, drawing from eight meta-analyses and systematic reviews published between 2013 and 2025. These studies collectively analyzed hundreds of randomized controlled trials involving thousands of participants, from healthy adults and typically developing children to clinical populations with ADHD, Alzheimer’s disease, mild cognitive impairment, and post-stroke cognitive impairment. The picture that emerges is more nuanced—and considerably less optimistic—than marketing materials suggest.

The evidence — what studies actually found

The most comprehensive assessment of working memory training in healthy populations comes from two major meta-analyses. A 2013 review of 23 studies with 30 group comparisons found that training programs produced reliable short-term improvements in the specific working memory tasks being trained—what researchers call “near transfer” [PMID 22612437]. Verbal working memory gains averaged a standardized mean difference (SMD) of approximately 0.5, a moderate effect. However, these improvements vanished at follow-up assessments. More critically, the review found “no convincing evidence” that training generalized to other cognitive skills including nonverbal and verbal ability, attention control, reading, or arithmetic.

A 2016 meta-analysis examining 87 publications with 145 experimental comparisons reached an even more sobering conclusion [PMID 27474138]. While immediate post-training improvements on working memory measures were reliable (intermediate transfer), there was zero convincing evidence of benefits on measures of “far transfer”—real-world cognitive skills like intelligence, reading comprehension, or mathematical reasoning—when working memory training was compared to active control conditions. Crucially, this study also found that the degree of improvement on working memory measures didn’t predict the size of any far-transfer effects, suggesting that even when people get better at the trained tasks, this doesn’t translate into broader cognitive gains.

A 2025 systematic review specifically examining Cogmed, one of the most widely used commercial working memory training programs, analyzed 11 randomized controlled trials in adults [PMID 40581145]. Results showed small but significant improvements in both verbal and visuospatial working memory immediately after training. However, sustained benefits were limited: only verbal working memory showed small continued improvement at 2-6 months post-intervention, and no significant differences were observed between groups at follow-up periods under two months.

For clinical populations, the evidence is mixed but slightly more encouraging. A 2015 meta-analysis of 16 studies involving 759 children with ADHD found that cognitive training produced moderate effects on laboratory tests of working memory (verbal: SMD = 0.52; visual: SMD = 0.47) [PMID 25721181]. However, effects on ADHD symptoms themselves were heavily dependent on who was doing the rating. When outcomes were assessed by raters who were likely blinded to treatment assignment, effects on total ADHD symptoms dropped to SMD = 0.20, and effects on inattention symptoms became non-significant (SMD = 0.32). Working memory training specifically showed no effects on ADHD symptoms, though interventions targeting multiple neuropsychological processes showed larger effects.

In older adults with cognitive impairment, combined interventions show more promise. A 2024 meta-analysis of dual-task training—which simultaneously challenges both cognitive and physical abilities—in older adults with mild cognitive impairment or dementia found significant improvements in global cognition (SMD = 0.477), executive function (SMD = -0.310), working memory (SMD = 0.714), gait, physical activity, and depression [PMID 38703695]. Similarly, a meta-analysis of non-invasive brain stimulation (repetitive transcranial magnetic stimulation or transcranial direct current stimulation) combined with cognitive training in Alzheimer’s disease and MCI patients showed improvements in global cognition, with effects maintained at follow-up for some interventions [PMID 38937842].

A 2023 review examining repetitive transcranial magnetic stimulation (rTMS) combined with cognitive training in post-stroke cognitive impairment found large effects on global cognition (SMD = 0.780), executive function (SMD = 0.769), and working memory (SMD = 0.609), plus medium improvements in activities of daily living (SMD = 0.418) [PMID 37004840]. However, no effects were found on memory or attention specifically.

Finally, a 2018 meta-analysis of physical exercise interventions in adults over 50 found that exercise improved cognitive function overall (SMD = 0.29), with benefits from aerobic exercise, resistance training, multicomponent training, and tai chi [PMID 28438770]. Sessions of 45-60 minutes at moderate intensity were associated with the strongest cognitive benefits, suggesting that physical activity may be more effective than isolated cognitive training for improving brain function.

How it might work — biological mechanisms

The theoretical rationale for working memory training rests on the concept of neuroplasticity—the brain’s ability to reorganize and form new neural connections in response to experience. Working memory relies heavily on prefrontal cortex networks, particularly the dorsolateral prefrontal cortex, which coordinates attention, manipulation of information, and executive control. Proponents argue that intensive, adaptive training (where difficulty increases as performance improves) should strengthen these neural circuits, much like physical training strengthens muscles.

Neuroimaging studies have shown that working memory training can produce changes in brain activation patterns, particularly in prefrontal and parietal regions. Some studies report increased efficiency (less activation for the same performance level) while others show expanded recruitment of neural resources. However, the critical question isn’t whether training changes the brain—all learning does—but whether these changes support broader cognitive improvements.

The dual-task and combined intervention approaches that show more promising results in clinical populations may work through different mechanisms. Dual-task training simultaneously engages multiple cognitive and motor systems, potentially promoting integration across neural networks rather than strengthening a single isolated capacity [PMID 38703695]. Non-invasive brain stimulation combined with cognitive training may enhance neuroplasticity through modulating cortical excitability, allowing training effects to consolidate more effectively [PMID 38937842][PMID 37004840].

Physical exercise likely improves cognition through multiple pathways: increasing brain-derived neurotrophic factor (BDNF) which supports neuron survival and growth, improving cerebrovascular health and oxygen delivery to the brain, reducing inflammation, and promoting neurogenesis in the hippocampus [PMID 28438770]. These systemic effects may explain why exercise produces broader cognitive benefits than isolated working memory training.

Where the evidence is strong (and weak)

The evidence is strongest on three points:

First, working memory training reliably improves performance on the trained tasks. This is the most consistent finding across all meta-analyses. People get better at the specific exercises they practice, with moderate to large effect sizes in the short term [PMID 22612437][PMID 27474138][PMID 40581145].

Second, these training effects are highly specific and don’t generalize to untrained cognitive domains in healthy populations. Multiple high-quality meta-analyses converge on this conclusion. Claims that working memory training boosts IQ, academic achievement, or general cognitive ability in typically developing individuals are not supported by the controlled research literature [PMID 22612437][PMID 27474138].

Third, combined or multimodal interventions show more promise than isolated cognitive training, especially in clinical populations. Approaches that pair cognitive training with physical exercise, brain stimulation, or multiple cognitive challenges appear more effective than working memory training alone [PMID 38937842][PMID 38703695][PMID 37004840][PMID 28438770].

The evidence is weakest regarding long-term maintenance of any benefits. Most studies show that even the specific training gains fade within months of stopping practice [PMID 22612437][PMID 40581145]. Few studies have followed participants beyond 6 months.

For clinical populations, evidence quality varies. While some meta-analyses show positive effects in ADHD, Alzheimer’s disease, MCI, and post-stroke populations, these reviews often note important limitations: small sample sizes, lack of blinding, heterogeneous intervention protocols, and publication bias [PMID 25721181][PMID 38937842][PMID 37004840]. The 2016 meta-analysis explicitly found that “there is no evidential value from the studies of working memory training using treated controls” when publication bias is considered [PMID 27474138].

Practical takeaways

For individuals considering working memory training programs, the evidence suggests tempering expectations:

If your goal is to improve performance on working memory tasks specifically—for example, if you’re a student trying to build skills for a particular type of academic test—computerized training may produce short-term gains. However, expect these benefits to be narrow and to require ongoing practice to maintain.

If you’re hoping for broader cognitive enhancement—better intelligence, problem-solving, academic performance, or professional productivity—current evidence doesn’t support working memory training as an effective intervention for healthy adults or children. The hoped-for transfer to real-world skills has not materialized in well-controlled studies.

For clinical populations, the picture is more nuanced. Combined interventions show the most promise:

  • Older adults with MCI or early dementia may benefit from dual-task training that pairs cognitive challenges with physical activity [PMID 38703695]
  • Post-stroke cognitive impairment may respond to brain stimulation combined with cognitive training [PMID 37004840]
  • Children with ADHD may see modest symptom improvements from cognitive training programs that target multiple neuropsychological processes, though effects are smaller when assessed by blinded raters [PMID 25721181]

Physical exercise appears more reliably beneficial for cognitive function than isolated cognitive training. Adults over 50 should aim for 45-60 minutes of moderate-intensity aerobic and resistance exercise most days of the week, which shows consistent cognitive benefits [PMID 28438770].

Beware of marketing claims. The disconnect between commercial advertising for brain-training programs and the peer-reviewed evidence is substantial. Programs backed by testimonials, celebrity endorsements, or selective citation of positive studies may overstate what the complete body of evidence actually shows.

Limitations of this review

Several factors limit the conclusions we can draw from this synthesis:

Heterogeneity of interventions: “Working memory training” encompasses diverse protocols—different tasks, intensities, durations, and adaptive algorithms. What works (or doesn’t) for one program may not apply to others. The Cogmed-specific review provides more consistency but still shows variability across studies [PMID 40581145].

Variability in outcome measures: Studies use different tests to assess cognitive function, making cross-study comparisons challenging. What one study calls “executive function” may be measured quite differently than in another.

Age and population differences: This review synthesized findings from children, healthy adults, older adults, and various clinical populations. Developmental stage and baseline cognitive status likely moderate training effects, but sample sizes in individual meta-analyses often precluded detailed subgroup analyses.

Publication bias: Multiple reviews identified evidence of publication bias, where studies with positive findings are more likely to be published than those with null results [PMID 27474138]. This means the true effects may be even smaller than meta-analyses suggest.

Short follow-up periods: Most studies assess outcomes immediately after training or within a few months. We know little about whether any benefits persist beyond six months or translate into long-term functional outcomes.

Control group challenges: The choice of control condition matters enormously. Studies using passive control groups (no intervention) often show larger effects than those using active controls (alternative activities). Many studies fail to adequately blind participants and assessors, introducing expectation effects.

Individual differences: Most meta-analyses report average effects, but some individuals may respond better than others to training. Predictors of individual response (baseline working memory capacity, motivation, genetics) remain poorly understood.

FAQ: Working Memory Training Effectiveness

Does working memory training actually improve your memory?

Working memory training produces modest, short-term improvements on the specific memory tasks you practice—what researchers call “near transfer.” Meta-analyses show people typically achieve moderate gains (around 0.5 standardized mean difference) on trained tasks immediately after completing programs [PMID 22612437]. However, these improvements usually disappear at follow-up assessments weeks or months later. More importantly, getting better at the training exercises doesn’t mean your overall memory capacity has fundamentally increased; you’ve essentially just gotten better at those particular tasks through practice.

Why doesn’t brain training transfer to real-world skills?

The scientific evidence consistently shows that improving performance on working memory exercises doesn’t translate into better performance on everyday cognitive tasks like reading comprehension, math problem-solving, or general intelligence. A comprehensive 2016 meta-analysis of 87 publications found zero convincing evidence of “far transfer” to real-world skills when compared to active control groups [PMID 27474138]. The reason appears to be that our brains become skilled at the specific strategies needed for training tasks rather than expanding underlying cognitive capacity. Essentially, you learn to excel at the game without improving the fundamental mental abilities the game claims to target.

What about working memory training for ADHD—does it help?

The evidence for working memory training in ADHD is mixed and controversial. While some studies show improvements on trained working memory tasks, the critical question is whether this helps with actual ADHD symptoms like inattention, impulsivity, or academic performance in daily life. Meta-analyses reveal that when studies use active control groups and blinded assessments—the gold standard for avoiding placebo effects—the benefits on real-world ADHD symptoms largely disappear. Any observed improvements may reflect expectation effects from parents and teachers who know the child received training rather than genuine symptom reduction.

How long do the benefits of working memory training last?

One of the most disappointing findings from rigorous research is that gains from working memory training fade quickly. The 2013 meta-analysis found that improvements measured immediately after training vanished at follow-up assessments [PMID 22612437]. Even for the modest benefits observed on trained tasks, there’s little evidence of durable long-term enhancement. This temporal pattern is inconsistent with the idea that training produces lasting structural changes in cognitive capacity and instead suggests temporary practice effects. If benefits don’t persist beyond the training period, the practical value for real-world cognitive function becomes questionable.

Are commercial brain training programs worth the money?

Based on current scientific evidence, most researchers would say no. While commercial programs like Cogmed and Lumosity are professionally designed and may provide some mental stimulation, rigorous meta-analyses show they don’t deliver on their core promise of improving general cognitive abilities or real-world functioning [PMID 27474138, PMID 40581145]. The improvements you see are typically limited to getting better at the specific games you practice, with minimal transfer to everyday tasks. For the cost of these programs—often hundreds of dollars—you could invest in activities with stronger evidence bases for cognitive health, such as physical exercise, social engagement, learning new complex skills, or quality education.

Studies referenced

  • PMID 38937842 — Systematic review and meta-analysis (15 studies, 685 patients) finding that non-invasive brain stimulation combined with cognitive training improved global cognition in Alzheimer’s disease and mild cognitive impairment, with effects maintained at follow-up for rTMS combinations.

  • PMID 22612437 — Meta-analysis of 23 studies examining working memory training in clinical and typically developing populations, finding short-term improvements on trained tasks but no convincing evidence of transfer to other cognitive skills or real-world abilities.

  • PMID 38703695 — Meta-analysis and trial sequential analysis of 20 randomized controlled trials (1,477 participants) showing dual-task training improved global cognition, executive function, working memory, gait, physical activity, and depression in older adults with cognitive impairment.

  • PMID 28438770 — Systematic review with meta-analysis of 39 studies demonstrating that physical exercise (aerobic, resistance, multicomponent, tai chi) improved cognitive function in adults over 50, with 45-60 minute sessions at moderate intensity showing the strongest benefits.

  • PMID 25721181 — Meta-analysis of 16 randomized controlled trials (759 children with ADHD) finding cognitive training improved working memory performance and parent-rated executive function but showed minimal effects on ADHD symptoms when assessed by blinded raters.

  • PMID 27474138 — Comprehensive meta-analysis of 87 publications with 145 experimental comparisons concluding that working memory training produces short-term specific training effects but no reliable improvements in intelligence, academic skills, or other measures of far transfer.

  • PMID 37004840 — Systematic review and meta-analysis of 8 trials (336 participants) showing repetitive transcranial magnetic stimulation combined with cognitive training produced large effects on global cognition, executive function, and working memory in post-stroke cognitive impairment patients.

  • PMID 40581145 — Systematic review and meta-analysis of 11 randomized controlled trials evaluating Cogmed Working Memory Training in adults, finding small significant improvements in verbal and visuospatial working memory immediately after training but limited evidence for sustained benefits beyond 2-6 months.


Source

  • PMID: 38937842 (read full paper on PubMed)
  • Journal: Alzheimer's research & therapy (2024)

Articles on this site are adapted from PubMed abstracts as general-interest explainers. They are not intended as medical advice.

📝 This article was adapted by the PubMed Trivia editorial team using Claude (Anthropic) as an adaptation assistant, based on the PubMed abstract cited above. Our editorial workflow includes automated NG-term checks (medical-claims law), careful "correlation vs causation" distinction, and mandatory PMID + source URL disclosure. See the full pipeline on our editorial policy and editor pages. Please report errors to sciencepubmedjp@gmail.com (we respond within 48 hours).