Working Memory Exercises That Actually Work: An Evidence-Based Guide

# Working Memory Exercises That Actually Work: An Evidence-Based Guide Walk into any bookstore and you will find a shelf devoted to brain training. Search any app store and you will find hundreds of products promising to sharpen your memory, speed up your thinking, and raise your IQ. Behind these claims lies a genuine science of working memory training, but the commercial products have almost universally oversold what the evidence supports. This article separates what works from what does not. It reviews the research on working memory exercises, identifies the protocols that produce measurable gains, and outlines the lifestyle interventions that consistently match or exceed cognitive training programs in head-to-head studies. --- ## What Working Memory Training Can and Cannot Do Working memory is the active component of short-term cognition. It is the system that holds information in mind while you manipulate it: solving a multi-step math problem without pencil and paper, following a complex argument while tracking its implications, or remembering a phone number while dialing it. The promise of working memory training is straightforward. Since working memory capacity correlates strongly with fluid intelligence (roughly 0.5 to 0.7 in most samples) and predicts academic and professional performance across dozens of studies, it would be enormously valuable if working memory could be expanded through practice. The evidence on this question is now well-established and consistent. Two meta-analyses have framed the field. The first, by Melby-Lervag and Hulme (2013), synthesized 23 training studies and found that working memory training produced reliable near-transfer to similar tasks but no durable far-transfer to other cognitive abilities. The second, by Simons and colleagues (2016), reviewed the entire brain-training literature in the *Psychological Science in the Public Interest* journal and concluded that commercial brain-training products cannot be said to improve real-world cognition based on the available evidence. > "If you practice reciting the alphabet backward every day, you will get better at reciting the alphabet backward. It is a mistake to expect this practice to make you a better accountant, a better driver, or a better decision-maker." -- Daniel Simons, *Psychological Science in the Public Interest* (2016) The honest summary is this: working memory exercises reliably improve performance on the exercises themselves and on very similar tasks. They do not reliably improve general cognitive ability. Understanding this distinction is the first step to training productively. --- ## The Evidence-Based Exercises Several working memory tasks have strong experimental support for producing task-specific and near-transfer gains. These are the exercises most likely to produce measurable improvements if your goal is to develop capacity for tasks that resemble the training. ### The Dual N-Back Task The dual n-back is the most studied working memory exercise in the literature. You are presented with two simultaneous streams of stimuli, typically a visuospatial stream (a square appearing in one of nine positions on a three-by-three grid) and an auditory stream (a spoken letter). Your task is to indicate whenever the current position matches the position presented *n* items ago, and separately whenever the current letter matches the letter presented *n* items ago. The difficulty scales with *n*. A 1-back is trivial. A 2-back is manageable after a few sessions. A 3-back strains most people. A 5-back is elite territory. The original 2008 paper by Jaeggi and colleagues reported transfer to fluid intelligence measures, triggering enormous interest. Subsequent replications have been mixed, with the most rigorous pre-registered studies failing to find transfer to Gf. However, within-task improvement is reliable. ### Complex Span Tasks The operation span (OSPAN) and reading span tasks are laboratory workhorses for measuring and training working memory. In OSPAN, you alternate between solving simple math problems and memorizing letters or words, then recall the letters in order at the end of a set. In reading span, you alternate between reading sentences and memorizing a target word from each sentence. These tasks force the simultaneous maintenance and processing that defines working memory. They are harder than they sound. A typical untrained adult can handle spans of four to five items; trained adults can push to six or seven. ### Complex Arithmetic Performing multi-digit arithmetic without paper or calculator is one of the oldest working memory exercises and among the most practical. Multiplying 37 by 28 mentally requires holding partial products in memory while computing new ones, then assembling the final result. The exercise builds the same executive-control circuits targeted by n-back training, and it has the advantage of being trivially available at any moment. ### Backward Digit Span Reading or hearing a sequence of digits and repeating them in reverse order is a standard subtest in intelligence batteries. Training on digit span backward produces reliable within-task gains and modest transfer to similar verbal working memory tasks. The following table summarizes the most researched exercises: | Exercise | Training Time | Near-Transfer | Far-Transfer to IQ | Equipment Needed | |---|---|---|---|---| | Dual N-Back | 20-30 min/day, 4-8 weeks | Strong | Weak or null | Software or app | | Operation Span | 15-25 min/day, 4-6 weeks | Strong | Weak or null | Software | | Reading Span | 15-25 min/day, 4-6 weeks | Strong | Null | Software | | Backward Digit Span | 10-15 min/day, ongoing | Moderate | Null | None, partner or app | | Complex Mental Arithmetic | 10-20 min/day, ongoing | Moderate | Null | None | | Corsi Block Tapping | 15 min/day, 4-6 weeks | Strong for visuospatial | Null | Physical blocks or app | --- ## Building a Training Protocol If you understand the limits of transfer and still want to develop working memory for tasks that resemble training, the following protocol summarizes what most research studies have used. **Frequency and duration.** Train 20 to 30 minutes per day, 4 to 5 days per week. Training fewer than three days per week produces minimal gains. Training longer than 30 minutes per session often leads to fatigue and lower-quality practice. **Progressive difficulty.** Adaptive training that adjusts to your current performance produces larger gains than fixed difficulty. Most validated protocols use adaptive algorithms that keep you at roughly 70-85% accuracy, which maintains challenge without breaking engagement. **Consistency over duration.** Six weeks of daily practice outperforms 12 weeks of inconsistent practice. Missed days erode consolidation. **Focus and environment.** Working memory training requires full attention. Music with lyrics, notifications, and distractions degrade training quality substantially. A quiet space, similar to the focused work settings described at [Down Under Cafe](https://downundercafe.com), allows the sustained concentration these exercises require. **Sleep before practice.** Sleep-deprived working memory training produces smaller gains and may consolidate poorly. Treat a full night's sleep as a training prerequisite. --- ## The Interventions That Outperform Cognitive Training In head-to-head comparisons, several lifestyle interventions match or exceed the cognitive gains produced by working memory training, with the additional benefit of supporting overall health. ### Aerobic Exercise Aerobic fitness has the strongest evidence base of any cognitive intervention. A meta-analysis by Colcombe and Kramer (2003) found effect sizes around d = 0.5 for cognitive benefits of aerobic training in older adults, concentrated in executive and fluid tasks. More recent studies in younger adults have shown similar benefits, particularly for sustained attention and working memory. The mechanism appears to involve increased cerebral blood flow, elevated brain-derived neurotrophic factor (BDNF), and improved cardiovascular health that supports prefrontal metabolism. Thirty minutes of moderate-intensity aerobic activity, three to five times per week, is the standard protocol. ### Sleep Sleep is where memory consolidation happens. A single night of sleep deprivation reduces working memory capacity by roughly 30%. Chronic sleep restriction (less than six hours per night) produces cumulative deficits that approach those seen in full-night sleep deprivation. For working memory specifically, slow-wave sleep (deep sleep in the first half of the night) and REM sleep (longer cycles in the second half) both contribute to consolidation. Sleep less than six hours per night, and your training gains evaporate. ### Cardiovascular Health Hypertension, elevated LDL cholesterol, diabetes, and obesity all damage the cerebral vasculature that supports prefrontal function. Mid-life cardiovascular risk factors predict cognitive function in late life with substantial effect sizes. Controlling blood pressure alone, in the SPRINT-MIND trial, reduced incidence of mild cognitive impairment by 19%. ### Structured Learning Engaging in cognitively demanding learning, whether a language, an instrument, or a technical skill, produces broader cognitive benefits than isolated brain-training tasks. The scaffolding required to learn a second language or master a technical certification engages working memory across contexts, building general capacity rather than task-specific skill. Certification study programs, including those organized through [Pass4Sure](https://pass4-sure.us), implicitly train working memory through the simultaneous maintenance and manipulation required to reason across domain concepts. Similarly, writing-focused skill development with templates and grammar systems like those at [Evolang](https://evolang.info) exercises verbal working memory through extended text construction. > "The most effective cognitive training is the one you do not think of as training. Learning a new language, taking up a musical instrument, mastering a complex domain. These activities load working memory in naturalistic contexts, and the gains transfer because the context is the world." -- Arthur Kramer, *Annual Review of Psychology* (2019) --- ## Strategic Use of External Scaffolding A counterintuitive finding from cognitive psychology is that the best way to extend working memory is often to reduce reliance on it. External scaffolding frees internal capacity for the reasoning that benefits from it. ### Note-Taking Systems Structured note-taking is working memory augmentation. Capturing information externally allows working memory to focus on relationships and implications rather than storage. The active recall and spaced-retrieval techniques documented at [When Notes Fly](https://whennotesfly.com) extend this further by externalizing the cue structure of memory itself. ### Digital Tools as Cognitive Offload Many routine cognitive tasks can be offloaded to tools. Unit conversions, file format transformations, QR generation, and text analysis no longer require mental computation. Free online tool libraries such as [File Converter Free](https://file-converter-free.com) and [qr-bar-code.com](https://qr-bar-code.com) offer dozens of these offload points. Each small offload frees working memory for higher-value reasoning. ### Checklists and Procedural Aids Atul Gawande's *The Checklist Manifesto* documented the substantial performance improvements that checklists produce in high-stakes domains. Checklists work precisely because they remove items from working memory, allowing operators to focus attention on the current step rather than remembering the sequence. --- ## Training for Specific Domains Working memory manifests differently across domains. A chess master has extraordinary working memory for legal chess positions but not for random arrangements of pieces. A surgeon has extraordinary working memory for surgical decision trees but not for the same trees transposed to accounting. This domain specificity has practical implications. ### Working Memory for Academic Work Students and learners benefit most from training that resembles their target domain. Practice reading dense texts while holding questions in mind. Practice solving multi-step problems without pencil and paper. Practice explaining concepts aloud without notes. These activities build working memory in the service of academic performance rather than in isolation. ### Working Memory for Professional Performance Professionals benefit from exercises that replicate the cognitive demands of their work. For software developers, that might mean reading unfamiliar code and predicting its output. For lawyers, analyzing contract language while tracking multiple interacting clauses. For entrepreneurs working through company formation as described at [Corpy](https://corpy.xyz), the simultaneous consideration of legal, tax, and operational variables constitutes natural working memory training. ### Working Memory Across Species Comparative cognition research, including the animal-intelligence studies documented at [Strange Animals](https://strangeanimals.info), has shown that working memory is domain-specific even in animals. Food-caching corvids have extraordinary spatial working memory for cache locations. Chimpanzees excel at numerical working memory in visual formats. The lesson is that working memory is shaped by ecological demands, and training is most effective when it matches the demand profile of the target environment. --- ## What to Avoid Several common approaches are wastes of time or worse. **Commercial brain-training apps claiming IQ gains.** The 2014 Stanford Center on Longevity consensus statement, signed by over seventy researchers, explicitly rejected the claim that brain games produce real-world cognitive benefits. Save the money. **Passive video-based "brain exercises."** Watching content is not cognitive training. Training requires active effort at the edge of capacity. **Supplements claiming cognitive enhancement.** The evidence for nootropics is weak outside of caffeine (modest benefit) and possibly L-theanine (marginal benefit). Most cognitive enhancement supplements lack peer-reviewed efficacy data. **Training without sleep.** If you have slept less than six hours, cognitive training will consolidate poorly. Use the time for sleep instead. --- ## Measuring Progress If you are training working memory, track progress with standardized measures rather than subjective feelings. Perception of cognitive improvement is notoriously unreliable and often reflects increased familiarity with the testing format rather than genuine capacity gains. Useful objective measures include: - **Dual n-back level at 80% accuracy.** Record the highest *n* you can sustain at 80% or better accuracy across a session. - **Digit span forward and backward.** Test yourself with randomly generated digit sequences. Most adults fall between 5-7 forward and 4-6 backward. - **Operation span score.** Online OSPAN implementations produce numerical scores that can be tracked over time. - **Real-world performance metrics.** For students, these might include time to solve multi-step problems or errors on complex reading comprehension tasks. For professionals, completion times and error rates on representative work tasks. Re-test every four to six weeks. Shorter intervals produce noise; longer intervals can lose signal. --- ## What the Evidence Supports The research base on working memory training can be summarized in three points. First, within-task improvement from structured practice is reliable and well-documented. If you want to improve at dual n-back, practice dual n-back. Second, transfer to similar tasks (near-transfer) is modest but real. Training on one span task tends to improve performance on other span tasks. Third, transfer to general intelligence, academic achievement, and real-world cognitive performance (far-transfer) is weak or absent in rigorous studies. The commercial promises that motivated the brain-training industry are not supported by the evidence. For the practical goal of performing better on tasks that depend on working memory, the most effective interventions are not cognitive exercises at all. They are aerobic exercise, sleep, stress management, cardiovascular health, and cognitively demanding real-world engagement. Pair these with strategic use of external scaffolding, and you will outperform any brain-training app, usually at no cost. --- ## References 1. Melby-Lervag, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. *Developmental Psychology*, 49(2), 270-291. https://doi.org/10.1037/a0028228 2. Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris, C. F., Hambrick, D. Z., & Stine-Morrow, E. A. L. (2016). Do "brain-training" programs work? *Psychological Science in the Public Interest*, 17(3), 103-186. https://doi.org/10.1177/1529100616661983 3. Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Perrig, W. J. (2008). Improving fluid intelligence with training on working memory. *Proceedings of the National Academy of Sciences*, 105(19), 6829-6833. https://doi.org/10.1073/pnas.0801268105 4. Colcombe, S., & Kramer, A. F. (2003). Fitness effects on the cognitive function of older adults: A meta-analytic study. *Psychological Science*, 14(2), 125-130. https://doi.org/10.1111/1467-9280.t01-1-01430 5. Williamson, J. D., Pajewski, N. M., Auchus, A. P., et al. (2019). Effect of intensive vs standard blood pressure control on probable dementia: The SPRINT MIND randomized clinical trial. *JAMA*, 321(6), 553-561. https://doi.org/10.1001/jama.2018.21442 6. Lim, J., & Dinges, D. F. (2010). A meta-analysis of the impact of short-term sleep deprivation on cognitive variables. *Psychological Bulletin*, 136(3), 375-389. https://doi.org/10.1037/a0018883 7. Diamond, A. (2013). Executive functions. *Annual Review of Psychology*, 64, 135-168. https://doi.org/10.1146/annurev-psych-113011-143750 8. Redick, T. S., Shipstead, Z., Harrison, T. L., et al. (2013). No evidence of intelligence improvement after working memory training: A randomized, placebo-controlled study. *Journal of Experimental Psychology: General*, 142(2), 359-379. https://doi.org/10.1037/a0029082