Working Memory: How It Works and How to Improve It
By Peter Benson, Cognitive Enhancement Researcher · Updated October 2026 · Citations re-verified
Short answer: Working memory can hold only a few chunks of information at once, often estimated at three to five. You can’t simply expand it, but you can use it better: write things down, cut interruptions, and protect sleep and stress levels. Brain-training apps don’t carry over to everyday tasks, and supplement evidence is weak.
⚕️ Educational information, not medical advice — full disclaimer. This page contains two affiliate links, both in the buying section.
Working memory is the part of your mind that holds a few things in view while you work with them: the start of a sentence while you finish it, the steps of a calculation, the structure of a document you are editing. When it overloads, you lose your place, re-read the same paragraph or forget what you were about to say.
The useful news is that most working-memory failures come from overload, not from a weak memory. This guide explains the limit, what pushes you past it, and which fixes have evidence, from free habits to supplements. For the wider picture, see the Memory & Learning hub.
How much can working memory hold?
Less than most people think. The old figure was “seven, plus or minus two” (Miller, 1956). When Cowan (2001) reviewed the evidence and controlled for tricks such as grouping and rehearsal, he concluded that the real limit is three to five chunks, averaging about four.
That figure is an estimate from controlled experiments, not a fixed four-slot container. How much you can hold in practice depends on the task, your attention, what you already know and how the information is grouped. It is still a useful rule of thumb: assume you can juggle only a handful of things, and plan accordingly.
A chunk is whatever you hold as one unit. That is why experts seem to hold more: a chess master sees a familiar pattern of pieces as one chunk, where a beginner sees several separate pieces. The capacity is the same; the chunks are richer.
Working memory also matters beyond the moment. A meta-analysis of 86 samples estimated the correlation between working memory and general intelligence at about 0.48: closely related, but not the same thing (Ackerman et al., 2005).
Why stress and poor sleep hit it so hard
Working memory relies on networks across the brain, with the prefrontal cortex, behind the forehead, playing a central role: where nerve cells keep firing to hold information “in mind”. That activity is easily disrupted. Amy Arnsten’s review of the stress research describes how stress chemicals weaken prefrontal networks; her Yale lab summarises it as “exposure to even mild, uncontrollable stress causes a rapid loss of PFC function” (Arnsten, 2009). Much of that work is in animals, but it explains why thinking feels harder when you are stressed and out of control.
Sleep loss has measured effects in people. A meta-analysis of 70 studies of short-term total sleep deprivation found a moderate impairment in working memory, both in accuracy and in speed. Simple attention, such as staying alert, was hit even harder (Lim & Dinges, 2010). Those studies kept people awake for 24 to 48 hours; regular short nights are less studied but point the same way. The full picture is in the sleep architecture guide.
What actually helps
Write it down. Psychologists call it cognitive offloading: using a note, a list or a reminder so your head doesn’t have to hold the information. A review by Risko and Gilbert describes how people use it to reduce mental demand, and why it works best when you trust the external record (Risko & Gilbert, 2016). Anything that does not need to be in your head right now is better on paper, leaving your limited capacity for the task in front of you.
Stop switching tasks. Switching between tasks has a cost: each switch means rebuilding what you were holding before. Protected blocks of work, with notifications off and email closed, reduce how often you pay it.
Chunk the material. Group related items under one heading so they take up the space of one item instead of several. This is the same thing experts do automatically.
Protect sleep and manage stress. These have clearer human evidence for working memory than any supplement in this article.
Do brain-training apps work?
They make you better at the games. They don’t make you better at much else. A meta-analysis of 87 publications and 145 comparisons found that working-memory training improved performance on similar working-memory tasks, but effects on intelligence, reading, arithmetic and other “far transfer” measures were close to zero and did not last (Melby-Lervåg et al., 2016). If you enjoy them, play them. Don’t expect them to help you hold a meeting in your head.
Supplements: creatine, alpha-GPC and citicoline
Creatine. The trial most often quoted for working memory is Rae et al. (2003): 45 young vegetarians took 5g a day for six weeks, and backward digit span and a reasoning test improved compared with placebo. The largest trial since then, Sandkühler et al. (2023), gave 123 adults 5g a day for six weeks and found no statistically significant effect on working memory or reasoning, and no extra benefit for vegetarians. Meta-analyses reporting a small memory benefit have been challenged for counting the same participants more than once (Citherlet, 2026), and the European Food Safety Authority concluded in 2024 that cause and effect had not been established.
Creatine may matter more when the brain is short of energy. In a small crossover trial, 15 adults kept awake for about 21 hours took a single large dose (0.35g per kg) or placebo; word-memory accuracy and processing speed were better on creatine (Gordji-Nejad et al., 2024). That dose is far higher than the usual 5g and was given once, under supervision. More detail is in the creatine guide.
Alpha-GPC. In the one trial in healthy adults, 20 men took 315mg or 630mg, and a Stroop attention test improved an hour later; two other attention tests did not change. The study was funded by an alpha-GPC supplier (Kerksick, 2024). A large Korean study linked alpha-GPC use to higher stroke risk in people over 50, while a later study in people with mild cognitive impairment found lower risk. Both are observational. See the alpha-GPC guide.
Citicoline is another choline source, sold alone and in multi-ingredient products. I have not reviewed its trials for this article, so I make no claim for it here.
In practice: no supplement has strong evidence for improving working memory in healthy, well-rested adults. Creatine is cheap and well studied for safety; any benefit for thinking is uncertain and may be largest when sleep is short.
Working-memory claims ranked by evidence
🟢 Stronger or replicated human evidence · 🟡 Limited, mixed or preliminary human evidence · 🔴 Not established
Worked example — reading a dense report without overloading working memory
You need to understand a report’s argument: a main claim, three supporting points, a caveat and your own objection. That is more than most people can hold at once, so pieces drop out and you re-read.
Write as you go. Put each supporting point on its own line as you meet it. It is now on paper, not in your head.
Group. Put the three points under one heading, “evidence for the claim”, so they travel as one chunk.
Cut interruptions. Close email and chat for the length of the task, so nothing new pushes out what you are holding.
The point: you didn’t try to concentrate harder, which doesn’t add capacity. You moved load onto paper, packed more into each chunk, and stopped new input pushing things out.
The NeuroEdge Formula Working Memory Self-Testing Framework
Free habits first, supplements last, one change at a time. A way of testing, not a prescription.
1. Offload
Use one capture system, paper or app, for every task and idea. Give it two weeks before judging.
2. Protect blocks
Work in uninterrupted blocks with notifications off. Notice errors and how tired you feel afterwards.
3. Fix sleep
If you are regularly short of sleep, fix that before buying anything.
4. Then one supplement
Creatine at 5g a day, judged over 8–12 weeks; or alpha-GPC on some days and not others, over two weeks. Check the cautions first.
For how to set a baseline and read your own results, see our self-testing method; for doses and how long to wait, see the dosing and timing guide.
Peter’s Testing Notes
A personal impression, not tracked data
Sleep. This is the one thing where I saw repeatable cause and effect. After nights under six hours I lost my train of thought mid-conversation, had to re-read paragraphs, and second-guessed decisions.
Habits. A strict capture system, a paper notebook plus a task app, made a striking difference. Before it, I rehearsed to-dos in my head and forgot half of them; afterwards my head felt like it had room to breathe. Ninety-minute blocks with my phone on airplane mode and email closed led to fewer errors on complex work, and I felt less drained afterwards. Both habits did more for me than any supplement.
Creatine. I have taken creatine for years, usually 5g a day in my morning decaf with no loading phase. After about two weeks the first time, I noticed it seemed to buffer fatigue: after a poor night, the fog was less severe, and I felt I could hold more steps in my head when debugging code or editing long documents. It was subtle, not dramatic.
Alpha-GPC. I took 300mg from Nootropics Depot in the morning, sometimes again around 2pm on heavy days, four or five times a week for about three months. The first hour of deep work felt cleaner. My caffeine intake also varied, so I can’t credit it fully. I had occasional mild headaches after three or four days in a row, which faded after 48 hours off.
Citicoline, Mind Lab Pro and others. Citicoline (250mg, three or four times a week) felt like a milder alpha-GPC, with no headaches. Mind Lab Pro did little for two weeks; by week three attention felt smoother, but I was also sleeping better, and I didn’t reorder. Caffeine alone gave sharp focus and then a crash. Rhodiola helped with endurance during heavy stretches but did nothing for memory, and Panax ginseng gave a slight mood lift but no thinking boost I could separate from caffeine.
Brain training. I used Lumosity for six months, with dual n-back about three times a week. My game scores went up; reading, multitasking and remembering names did not. I stopped.
None of this was blinded or measured, and several changes overlapped. If I started again: sleep first, offload second, and supplements only as fine-tuning.
Choosing products
The habits above cost nothing and did more in the research and in my own experience. If you want to test a supplement, creatine is the cheapest place to start; sourcing is covered in the creatine guide. For a choline source:
Nootropics Depot Alpha GPC — standalone alpha-GPC
150mg of alpha-GPC per capsule, third-party tested, with certificates of analysis by lot. Two capsules give 300mg, close to the lower dose in the healthy-adult trial. Read the stroke-risk caution first.
Mind Lab Pro — multi-ingredient
Per two capsules: 250mg citicoline, 150mg Bacopa (24% bacosides), 500mg Lion’s Mane, 100mg phosphatidylserine, 175mg N-acetyl L-tyrosine, 100mg L-theanine, 50mg Rhodiola, 75mg pine bark extract and B vitamins. It contains no alpha-GPC. Several doses are below those used in single-ingredient trials, and a blend makes it impossible to tell which ingredient is doing anything.
These are affiliate links. If you buy through them, NeuroEdge Formula may earn a commission at no extra cost to you. My conclusion above is that habits matter more than either product.
Side Effects & Interactions
Alpha-GPC and stroke. One large observational study linked it to higher stroke risk in people aged 50 and over; another found lower risk in people with mild cognitive impairment. If you are over 50 or have cardiovascular risk factors, speak to your doctor first.
Choline sources. Headaches are a commonly reported side effect of alpha-GPC. Tell your prescriber if you take anticholinergic medicines or a cholinesterase inhibitor such as donepezil.
Creatine and blood tests. Creatine raises blood creatinine, the marker used to estimate kidney function, without evidence of kidney harm in healthy adults. Tell your doctor before a blood test. If you have kidney disease, don’t take it without your doctor’s agreement.
Multi-ingredient products. Blends such as Mind Lab Pro contain several active ingredients, including Bacopa and Rhodiola, each with its own interactions. Check every ingredient against your medicines.
Pregnancy and breastfeeding. Most of these supplements have not been studied in pregnancy or breastfeeding. Avoid them unless a doctor advises otherwise.
This is not a complete list. See the alpha-GPC and creatine guides.
Key takeaways
| → | Working memory holds only a few chunks, often three to five. You can’t simply expand it, but you can reduce the load on it. |
| → | Offloading, single-tasking and good sleep are the best-supported ways to work within it, and they are free. |
| → | Brain-training apps improve the games, not everyday thinking. |
| → | Creatine’s evidence for working memory is mixed; the largest trial found no effect. It may help more when you are short of sleep. |
| → | Alpha-GPC has one small, supplier-funded trial and an open stroke-risk question for people over 50. |
Frequently asked questions
Can you increase working memory capacity?
Not by much, as far as the evidence shows. Training raises scores on the trained tasks and sometimes on closely related ones, without convincing carry-over to intelligence or everyday thinking. What you can change is how much fits into each chunk, which comes with expertise, and how much load you put on it.
What’s the difference between working memory and short-term memory?
Short-term memory is holding information briefly, such as a phone number. Working memory is holding it and doing something with it, such as adding up several numbers in your head. The terms overlap, and researchers use them in different ways.
Why do I forget what I was going to say mid-sentence?
Often because something new arrived while you were holding the point and interfered with it. Attention lapses, tiredness and stress can also play a part, and you can’t tell which from the experience alone. Occasional episodes are common. In a busy conversation, jot your point down. If it is new, frequent or getting worse, see a doctor.
Does working memory decline with age?
Yes, on average, though people vary a lot. If you notice a sudden or marked change, speak to a doctor rather than reaching for supplements.
Get “7 Days to a Sharper Brain” — Free
Peter Benson’s evidence-first starting framework
The habits with the strongest evidence first — sleep, movement, stress and learning — then how to add a single supplement and test whether it actually works for you.
No spam. Unsubscribe anytime.
⚕️ Educational Information, Not Medical Advice
This article is educational and is not medical advice. I’m a cognitive enhancement researcher, not a medical doctor. If working-memory problems interfere with daily life, or you notice a sudden change in thinking, see a doctor. Alpha-GPC has been linked to stroke risk in one large observational study, and several supplements here interact with medicines. Consult a qualified healthcare provider before use, particularly if you are over 50, take prescription medicines, or are pregnant or breastfeeding.
Scientific references
1. Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. Link
2. Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review, 63(2), 81–97.
3. Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2005). Working memory and intelligence: the same or different constructs? Psychological Bulletin, 131(1), 30–60. Link
4. Arnsten, A. F. T. (2009). Stress signalling pathways that impair prefrontal cortex structure and function. Nature Reviews Neuroscience, 10(6), 410–422. PMID: 19455173
5. 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. Link
6. Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. Link
7. Melby-Lervåg, M., Redick, T. S., & Hulme, C. (2016). Working memory training does not improve performance on measures of intelligence or other measures of “far transfer”: evidence from a meta-analytic review. Perspectives on Psychological Science, 11(4), 512–534. Link
8. Rae, C., Digney, A. L., McEwan, S. R., & Bates, T. C. (2003). Oral creatine monohydrate supplementation improves brain performance: a double-blind, placebo-controlled, cross-over trial. Proceedings of the Royal Society B, 270(1529), 2147–2150. Link
9. Sandkühler, J. F., Kersting, X., Faust, A., et al. (2023). The effects of creatine supplementation on cognitive performance: a randomised controlled study. BMC Medicine, 21, 440. Link
10. Citherlet, T. (2026). Commentary: The effects of creatine supplementation on cognitive function in adults: a systematic review and meta-analysis. Frontiers in Nutrition. Link
11. EFSA Panel on Nutrition, Novel Foods and Food Allergens (2024). Creatine and improvement in cognitive function: evaluation of a health claim. EFSA Journal. DOI: 10.2903/j.efsa.2024.9100
12. Gordji-Nejad, A., et al. (2024). Single dose creatine improves cognitive performance and induces changes in cerebral high energy phosphates during sleep deprivation. Scientific Reports, 14, 4937. Link
13. Kerksick, C. M. (2024). Acute alpha-glycerylphosphorylcholine supplementation enhances cognitive performance in healthy men. Nutrients, 16(23), 4240. Link (Funded by an alpha-GPC supplier.)
14. Lee, G., et al. (2021). Association of L-α glycerylphosphorylcholine with subsequent stroke risk after 10 years. JAMA Network Open. Link
15. Kim, H.-K., et al. (2025). Association between L-α glycerylphosphorylcholine use and delayed dementia conversion: a nationwide longitudinal study in South Korea. The Journal of Prevention of Alzheimer’s Disease. Link
Cognitive Enhancement Researcher | 18+ Years Independent Research
Peter is not a clinician; this guide is educational, not medical advice. It was rewritten in October 2026 to add the largest creatine trial and the brain-training evidence, and to correct the alpha-GPC findings. Read about how NeuroEdge Formula evaluates research.
Last reviewed: October 2026 | Educational content only. Not medical advice.
