Memory is trainable, and the evidence is unusually specific about how. Retrieval practice moved one-week recall from about 40 percent to about 61 percent in Roediger and Karpicke's 2006 experiment. Spaced review beat cramming across 254 studies covering more than 14,000 people. Six weeks of memory-palace training took ordinary volunteers from recalling 26 words out of 72 to recalling roughly 62. What does not work is equally well documented: rereading, highlighting, brain-training games, and ginkgo. The seven methods below are the ones with numbers behind them.
You meet someone at a party. Two minutes later their name is gone. You read something fascinating yesterday and cannot reconstruct a single sentence of it today. You study for three hours and retain almost nothing.
You do not have a bad memory. You have an untrained one, and you are almost certainly using the two study methods that research rates as least effective.
Why You Forget, and How Fast
Hermann Ebbinghaus spent the 1880s memorizing nonsense syllables and measuring what survived. The curve he produced is the most famous finding in memory research and also the most frequently misquoted.
Here is what his data actually showed, measured as savings on relearning:
| Time since learning | Roughly how much survived |
|---|---|
| 20 minutes | 58% |
| 1 hour | 44% |
| 1 day | 34% |
| 6 days | 25% |
| 31 days | 21% |
You will often read that you forget 80 percent within a day and that nearly nothing survives a month. Neither number is his. Ebbinghaus lost about two thirds of the material in the first 24 hours, and then the curve flattened hard. A month later he still held roughly a fifth of it.
That shape matters more than the individual figures. Forgetting is brutal at the start and gentle afterwards. Jaap Murre and Joeri Dros replicated the entire experiment in 2015 in PLOS ONE, running 70 hours of learning and relearning, and the curve came out close to the original. This is one of the few findings in psychology that has survived 130 years of scrutiny intact.
The practical consequence: the first day after you learn something is when you lose the most, and it is also when one short review buys you the most. Everything below follows from that.
1. Spaced Repetition: The Largest Single Lever
Spaced repetition means reviewing material at widening intervals rather than in one block. It is the single best-evidenced thing you can do to your study schedule, and it costs no extra time.
Nicholas Cepeda and colleagues synthesized 254 studies with more than 14,000 participants in 2006 and found that spaced practice beat massed practice at identical total study time. Same hours, better recall, purely from how the hours were arranged.
A follow-up in 2008 gave the finding a usable rule. The optimal gap between reviews scales with how long you need to hold the information. For a test in a week, review after about a day. For something you want in six months, the gap stretches to weeks. As a rough heuristic, space your reviews at roughly 10 to 20 percent of the retention interval you care about.
A schedule that works for most material
| Review | When | What it does |
|---|---|---|
| 1st | Same day, a few hours later | Catches the steepest part of the curve |
| 2nd | Next day | Converts a fragile trace into a durable one |
| 3rd | 3 days later | First real test of long-term storage |
| 4th | 1 week later | Gaps widen as the memory stabilizes |
| 5th | 1 month later | Maintenance, not learning |
Two things make this work. Each successful recall physically strengthens the trace, and each review that happens just before you would have forgotten is the most efficient one you can do. Reviewing too early wastes effort because nothing has decayed yet. Reviewing too late means relearning from scratch.
If you want the full comparison of how spacing differs from related methods, we broke it down in microlearning vs spaced repetition vs active recall.
2. Active Recall: Testing Beats Rereading, Every Time
Here is the finding that surprises people most: the best way to improve memory is not to study more. It is to test yourself more.
Henry Roediger and Jeffrey Karpicke ran the definitive version in 2006. Students studied a passage, then either restudied it or took a practice test. Five minutes later, the restudy group looked better. One week later the ranking flipped hard: the tested group recalled about 61 percent, the restudy group about 40 percent.
Same material. Same exposure. A 50 percent relative improvement from swapping input for output.
The mechanism is straightforward. Rereading is recognition, and recognition is easy, which is exactly why it feels productive. Retrieval is hard, and the difficulty is the point. Every time you drag something out of memory without a prompt, you make the route back to it wider.
Retrieval also does something rereading cannot: it tells you the truth about what you know. Fluency is a liar. Text you have read four times feels familiar, and familiarity gets mistaken for knowledge right up until the moment you need to produce it.
What the research actually rates
John Dunlosky and colleagues reviewed ten common study techniques in 2013 and graded each on utility. The results are a quiet indictment of how most people study:
| Technique | Rated utility |
|---|---|
| Practice testing | High |
| Distributed (spaced) practice | High |
| Elaborative interrogation (asking "why is this true?") | Moderate |
| Self-explanation | Moderate |
| Interleaved practice | Moderate |
| Summarization | Low |
| Highlighting and underlining | Low |
| Rereading | Low |
Highlighting and rereading are the two most popular study methods in the world. Both sit at the bottom. They are not worthless, they are just input, and memory is built by output. We walk through the high-utility ones in detail in 10 study techniques that actually work.
3. The Memory Palace: 2,000 Years Old and Genuinely Effective
Roman orators had no note cards, so they built palaces. The method of loci works by placing information at specific points along a route you already know, then mentally walking that route to retrieve it.
It sounds like a parlor trick. It is not. Martin Dresler and colleagues published the test in Neuron in 2017. They scanned 23 of the world's top memory athletes, then took volunteers with ordinary memories and gave them 30 minutes of method-of-loci training a day for six weeks.
Before training, participants recalled around 26 words from a list of 72. After training, they recalled about 35 more, landing near 62 out of 72. Four months later, with no continued practice, they were still holding more than 22 words above their starting point. Their brain scans had shifted toward the connectivity patterns seen in the memory champions, which is neuroplasticity doing exactly what it is supposed to do.
How to build one
- Pick a route you know cold. Your apartment, your walk to work, your childhood home. It must be a place you can traverse in the same order every time.
- Fix ten stops along it. Front door, coat hook, kitchen counter, sofa, and so on. Always the same ten, always the same order.
- Place one item at each stop, and make it absurd. Milk exploding across the front door. A whole chicken sitting in your armchair. Bizarre, oversized, and moving beats tidy and accurate every time.
- Walk the route to retrieve. Do not try to recall the list. Recall the walk, and the list comes with it.
One honest caveat. Dunlosky's review rated the keyword mnemonic as low utility, and mnemonics in general are narrow tools. They are superb for ordered lists, names, and vocabulary. They do very little for conceptual understanding. Use a palace for the grocery list and the speech, not for grasping why interest rates move.
4. Chunking: Your Working Memory Holds Four Things, Not Seven
George Miller's "magical number seven" from 1956 is one of the most quoted numbers in psychology. It is also outdated. Nelson Cowan reviewed the evidence in 2001 and put the real limit closer to four items when you strip out rehearsal and grouping tricks.
Four. That is the size of the doorway everything has to pass through.
You cannot widen the doorway, but you can make the packages bigger. Chunking groups information into fewer, more meaningful units.
Take the digits 1 4 9 2 1 7 7 6 1 9 6 9. That is twelve items, well past the limit. Now read them as 1492, 1776, 1969: Columbus, independence, moon landing. Three chunks, trivially memorable, because each one hooks into knowledge you already have.
This is why phone numbers are hyphenated and why experts remember more than novices in their own field. A chess grandmaster does not have a better memory than you. They have bigger chunks. Where you see twenty pieces, they see four familiar formations.
When you are learning something new, stop trying to hold twenty facts. Find the four or five ideas the twenty facts hang from.
5. Elaboration: Connect It or Lose It
Isolated facts are the hardest thing your brain stores. Connected facts are the easiest, because every connection is another route back.
Elaboration means asking why something is true and how it relates to something you already know. It is the mechanism behind the well-documented benefit of explaining ideas out loud, which is often mislabeled as the "teach it to retain 90 percent" rule. That percentage is fabricated, as we traced in the origins of the learning pyramid, but the underlying effect is real. Explaining forces retrieval, exposes gaps, and demands that you organize scattered facts into a sequence someone else could follow.
Three questions do most of the work:
- Why is this true? Forces you past the surface statement.
- What does this remind me of? Attaches the new idea to an existing network.
- Where would this break? Reveals whether you understand the boundary or just the headline.
This is also the fix for the most common complaint about reading, which is finishing a book and retaining none of it. We covered that specific case in how to remember everything you read.
6. Sleep: The Part Where Memory Is Actually Built
Learning does not finish when you close the book. It finishes overnight.
During slow-wave sleep, the hippocampus replays the day's traces and transfers them to the cortex for long-term storage. Jan Born and Bjorn Rasch laid out the evidence in Physiological Reviews in 2013, and the direction is consistent across decades of work: sleep after learning improves retention, and sleep loss before learning impairs the ability to encode at all.
This makes the all-nighter the worst possible trade. You buy a few hours of study and pay for them with the consolidation window that would have made the study stick. If you have to choose between one more hour of review and one more hour of sleep the night before an exam, the evidence favors sleep.
If falling asleep is the actual bottleneck, we covered what genuinely moves it in why melatonin is not fixing your sleep.
7. Aerobic Exercise: The Only "Brain Booster" With Structural Evidence
Most things sold as memory enhancers do nothing. Aerobic exercise is the exception, and the evidence is anatomical rather than anecdotal.
Kirk Erickson and colleagues published a randomized trial in PNAS in 2011. Older adults who walked regularly for a year increased hippocampal volume by roughly 2 percent, effectively reversing one to two years of age-related shrinkage. The control group's hippocampus continued to decline. Larger volume tracked with better spatial memory performance.
The hippocampus is the structure that files new memories. You can grow it by walking. That remains one of the more remarkable findings in cognitive neuroscience, and it costs nothing.
What Does Not Work (And Why It Sells So Well)
Three popular categories have been tested properly and come up short.
Brain-training games. Daniel Simons and a team of researchers reviewed the entire literature in 2016. Their conclusion: brain training reliably improves performance on the trained task, shows limited transfer to closely related tasks, and shows little evidence of improving everyday cognition. You get better at the game. That is where it stops. We tested the current crop in our review of brain training apps.
Ginkgo biloba. The Ginkgo Evaluation of Memory study followed 3,069 adults aged 75 and over for roughly six years and found no reduction in dementia incidence. It is the largest trial of the most popular memory supplement in the world, and it was null.
Rereading your notes. Covered above, and worth repeating because it is the default behavior of almost every student on earth.
The New Risk: Outsourcing Recall to AI
Something changed in the last two years that older memory advice does not account for. When an assistant can retrieve, summarize, and phrase anything on demand, the retrieval step that builds memory quietly disappears.
An MIT Media Lab team ran an early test of this in 2025. They tracked 54 participants writing essays with an LLM, with a search engine, or unaided, recording EEG throughout. The LLM group showed the weakest neural connectivity, reported the lowest sense of ownership over their work, and in the first session roughly 83 percent could not quote a single line from the essay they had just written. The researchers called the effect "cognitive debt."
Treat this as a signal rather than a settled result. It is a preprint, the sample is small, and the task is narrow. But the mechanism is exactly what fifty years of retrieval research would predict: if you never retrieve, you never encode.
The fix is not avoiding AI. It is deciding what you want to own. Use AI to find and structure information, then close it and reconstruct the idea from memory before moving on. That single step converts consumption back into learning.
How to Improve Your Memory in 30 Days
Techniques fail when you try to adopt seven of them on a Monday. Add one layer a week instead.
Week 1: Learn less, but retrieve it
Cut your material to one concept a day. After learning it, close everything and say it out loud from memory. Thirty seconds. That is the whole exercise, and it does more for retention than anything else on this list.
Week 2: Add spacing
Keep the daily retrieval, and add a review of yesterday's concept before you start today's. You are now running a two-day loop. At the end of the week, do one pass over all seven.
Week 3: Add elaboration and a palace
For each concept, answer "why is this true?" and "what does this remind me of?" in one sentence each. Separately, build one ten-stop memory palace and use it for something concrete, like a list or a set of names.
Week 4: Fix the infrastructure
Protect sleep on the nights after you learn something that matters. Add three or four sessions of brisk walking or equivalent. These are not study techniques, they are the conditions under which study techniques work.
By day 30 you are running the full loop: small unit, immediate retrieval, spaced review, deliberate connection, adequate sleep. Nothing here requires more hours than you already spend.
Why Small Daily Sessions Beat Long Ones
Suppose you need to learn 100 facts. Cram them in one night and you may pass a test tomorrow. Learn five a day for twenty days, retrieving each one on a widening schedule, and most of them are still there next year.
Identical total effort. Radically different outcome. This is the spacing effect and the testing effect stacked on top of each other, and it is the reason short daily sessions outperform weekend marathons.
It is the design principle behind NerdSip, and it is worth being precise about which parts of that principle a tool can actually give you.
A NerdSip course is three, five, or ten lessons of roughly 200 words each, so a short one runs about five minutes end to end. Every single lesson closes with a multiple-choice question rather than a summary, which is the retrieval step from section two built into the format. Finished courses can be pulled back later into a review round, and on paid plans a weekly recap re-asks questions from the previous week's material. Streaks, XP, and weekly leagues exist to solve the part almost everyone fails at, which is showing up again tomorrow.
What it does not do is schedule your reviews by algorithm. There is no due-date engine deciding that a particular fact is ripe for recall on day nine. If you want strict interval scheduling, a dedicated tool like Anki is the honest recommendation, and everything in this article works with paper index cards too. The method matters more than the tool. What matters most is that the loop runs daily.
The Mistakes That Quietly Kill Retention
- Studying until it feels easy. Fluency is the feeling of recognition, not the state of knowing. If it feels effortless, you are probably rereading.
- Reviewing too soon. Reviewing something you have not begun to forget is wasted effort. Let a little decay happen first.
- Learning in one long block. Two hours on Sunday loses to twenty minutes six times a week, at less total cost.
- Skipping the first-day review. This is the steepest part of the curve and the cheapest place to intervene.
- Treating sleep as optional. You cannot consolidate what you did not sleep on.
Bottom Line
Memory is not a fixed capacity you were issued at birth. It is a set of processes, and the processes respond to how you use them. Ordinary volunteers doubled their word recall in six weeks with one technique. Students improved one-week retention by half by swapping rereading for self-testing. Older adults grew hippocampal tissue by walking.
None of that requires talent. It requires learning less at a time, retrieving it instead of rereading it, spacing the reviews, sleeping properly, and moving your body. Do those five things and the question stops being how to improve your memory and starts being what you want to fill it with.
References & Further Reading
- Murre & Dros (2015), Replication and Analysis of Ebbinghaus' Forgetting Curve, PLOS ONE
- Cepeda et al. (2006), Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis, Psychological Bulletin
- Cepeda et al. (2008), Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention, Psychological Science
- Roediger & Karpicke (2006), Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention, Psychological Science
- Dunlosky et al. (2013), Improving Students' Learning With Effective Learning Techniques, Psychological Science in the Public Interest
- Dresler et al. (2017), Mnemonic Training Reshapes Brain Networks to Support Superior Memory, Neuron
- Cowan (2001), The Magical Number 4 in Short-Term Memory, Behavioral and Brain Sciences
- Rasch & Born (2013), About Sleep's Role in Memory, Physiological Reviews
- Erickson et al. (2011), Exercise Training Increases Size of Hippocampus and Improves Memory, PNAS
- Simons et al. (2016), Do "Brain-Training" Programs Work?, Psychological Science in the Public Interest
- DeKosky et al. (2008), Ginkgo Biloba for Prevention of Dementia (GEM Study), JAMA
- Kosmyna et al. (2025), Your Brain on ChatGPT: Accumulation of Cognitive Debt, arXiv preprint
Frequently Asked Questions
Can you actually improve your memory, or is it fixed?
You can improve it, and the effect sizes are large. In a 2017 study published in Neuron, volunteers with ordinary memories trained with the method of loci for 30 minutes a day over six weeks and went from recalling about 26 words out of 72 to about 62. Four months later, with no continued practice, they still recalled more than 22 words above their starting point. Memory capacity is not a fixed trait; it responds to technique.
What is the fastest way to improve your memory?
Replace rereading with self-testing. It is a single change, it costs no extra time, and it has the largest immediate payoff of anything on the list. In Roediger and Karpicke's 2006 experiment, students who took a practice test recalled about 61 percent of a passage a week later, compared with about 40 percent for students who simply restudied it. Close the book and reconstruct the idea from memory before you move on.
How long does it take to improve your memory?
You will notice a difference in retention within one to two weeks of consistent retrieval practice, because the effect is about how you encode rather than about slow physical change. Larger gains take longer: the memory-palace study ran six weeks at 30 minutes a day, and the hippocampal growth seen in exercise trials took about a year of regular aerobic activity.
Does spaced repetition really work?
Yes, and it is one of the best-replicated findings in learning research. A 2006 synthesis of 254 studies covering more than 14,000 participants found that spreading study sessions out beat massing them together at identical total study time. The practical rule from later work is to space reviews at roughly 10 to 20 percent of the interval you need to remember for: about a day's gap for a test next week, weeks for something you want to keep for months.
How many things can your working memory hold?
About four, not seven. George Miller's famous 'magical number seven' from 1956 has been revised: Nelson Cowan's 2001 review put the real limit closer to four items once you remove rehearsal and grouping strategies. You cannot raise that ceiling, but chunking lets you make each of the four items bigger and more meaningful, which is why experts appear to remember more in their own field.
Do brain training apps and games improve memory?
Largely no. A 2016 review by Simons and colleagues in Psychological Science in the Public Interest found that brain-training programs reliably improve performance on the trained task, show limited transfer to closely related tasks, and show little evidence of improving everyday cognition. You get better at the game, not at remembering names or facts. Techniques such as retrieval practice and spacing transfer to real material because they work on real material.
Do memory supplements like ginkgo biloba work?
The evidence for ginkgo is negative. The Ginkgo Evaluation of Memory study followed 3,069 adults aged 75 and over for roughly six years and found no reduction in dementia incidence. It remains the largest trial of the most popular memory supplement on the market. Supplements can help when they correct an actual deficiency, such as B12, but no pill has evidence approaching that of sleep, exercise, and retrieval practice.
Why do I forget things so quickly?
Because forgetting is front-loaded. Ebbinghaus's data, replicated in PLOS ONE in 2015, shows roughly 58 percent of new material surviving after 20 minutes, 44 percent after an hour, and 34 percent after a day, after which the curve flattens and about a fifth is still there a month later. The first 24 hours are when you lose the most, which is also why a single short review on day one buys you more than any other review you will do.
Does sleep affect memory?
Substantially. During slow-wave sleep the hippocampus replays the day's memory traces and transfers them to the cortex for long-term storage. Sleep after learning improves retention, and sleep loss before learning impairs your ability to encode at all. This makes the all-nighter the worst trade in studying: you buy study hours by spending the consolidation window that would have made those hours stick.
Is using AI making my memory worse?
It can, if it removes the retrieval step. A 2025 MIT Media Lab preprint tracked 54 people writing essays with an LLM, with a search engine, or unaided. The LLM group showed the weakest neural connectivity, the lowest sense of ownership over their work, and in the first session about 83 percent could not quote a single line from the essay they had just written. It is a small, unreviewed study, but the mechanism matches decades of retrieval research. Use AI to find and structure information, then close it and reconstruct the idea from memory.
What is the best app for improving memory?
Judge apps by mechanics rather than branding: short units, a forced retrieval step at the end of each one, and some way of bringing material back later. Anki remains the standard for strict algorithmic spaced repetition, provided you are willing to build your own cards. NerdSip suits you better if you want the material generated for you: type any topic and it becomes a short course where every lesson ends in a multiple-choice question, so retrieval is built into the format, and completed courses can be pulled back into a review round afterwards. It does not schedule reviews by algorithm, so add your own calendar reminders if strict intervals matter to you.
Every NerdSip course clears a four-layer fact-checking pipeline before it reaches the app: grounded in live sources, scored nightly, gated on failure, and reviewable by a person at one tap. Articles on this blog are drafted with AI assistance, then researched, verified, and edited by our team.
Editorial responsibility: Philip S., ai51 UG (haftungsbeschränkt) — contact via imprint.
📚 Keep Learning
Explore the NerdSip learning hubs
The blog is our editorial layer. When you want to move from reading into doing, these hubs organize the rest of NerdSip by format, including the app itself and the press that has written about it.
Turn Reading Into Remembering
NerdSip turns any topic into a short course where every lesson ends in a question, not a summary. Retrieval built in, plus a reason to come back tomorrow.
