Split View: 과학적 암기법 완벽 가이드 2026 — 뇌과학 기반 기억 최적화 딥다이브

The Science of Memory 2026 — A Neuroscience-Based Deep Dive into Memory Optimization

• Introduction — Why Do Some People Remember Better
• 1. The Ebbinghaus Forgetting Curve — Why We Forget
• 2. The Science of Spaced Repetition — The Distributed Practice Effect
• 3. From SM-2 to FSRS — The Evolution of Spaced-Repetition Algorithms
• 4. Retrieval Practice vs Rereading — Karpicke & Roediger 2008
• 5. The Testing Effect — A Test Is Not Assessment, It Is Learning
• 6. Optimal Anki Setup — FSRS Parameters and New-Card Count
• 7. How to Write Good Cards — The Minimum Information Principle
• 8. The Memory Palace and Method of Loci — Placing Knowledge in Space
• 9. The Brains of Memory Champions — Training, Not Talent
• 10. Chunking — Miller's 7±2 and Expert Chunking
• 11. Dual Coding — Combining Text and Images
• 12. Interleaving vs Blocked Practice — Mix It to Make It Last
• 13. Levels of Processing and Elaborative Encoding
• 14. Elaborative Interrogation — Asking "Why?"
• 15. The Feynman Technique — Learning by Teaching
• 16. Sleep and Memory Consolidation — From Hippocampus to Neocortex
• 17. The Power of Naps — A 20-Minute Reboot
• 18. Exercise and BDNF — Fertilizer for the Brain
• 19. Applying It to Language Learning — A Vocabulary System
• 20. A Spaced-Repetition System for Developers — Memorizing Commands and Concepts
• 21. Common Myths — The Learning-Styles Myth and Rereading
• 22. A 4-Week Practical Memory-Optimization System
• Conclusion — Memory Is a Skill
• References

Introduction — Why Do Some People Remember Better

"I just have a bad memory" is almost always wrong. Across 140 years of memory research, one conclusion keeps repeating: good memory is not a talent, it is a method. When scientists scanned the brains of world memory champions, they were anatomically ordinary. What was different was the strategy they used.

This article throws out intuitive-but-inefficient habits like "highlight and reread," and assembles memory techniques that have been validated by experiment. Three tools sit at the core: spaced repetition (reviewing just before you forget), active recall (retrieving from your own mind instead of rereading), and the spaced-repetition software (SRS) that automates both.

Every section cites real research. From Ebbinghaus (1885) to Karpicke & Roediger (2008), to the FSRS algorithm published in 2022 and now shipping inside Anki by default, we look at the evidence for what works and why. At the end we pull it together into a practical system you can apply directly to language learning and to memorizing developer knowledge.

1. The Ebbinghaus Forgetting Curve — Why We Forget

Memory research begins in 1885 with Hermann Ebbinghaus. Using himself as the subject, he memorized thousands of meaningless consonant-vowel-consonant syllables (e.g. "WID", "ZOF") and measured how much he forgot over time. The result is the famous forgetting curve: memory drops steeply right after learning, then flattens.

There are two key insights. First, forgetting is fastest in the first few hours to a day. Second, relearning the same material takes far less time (the "savings" effect). Even when something seems completely forgotten, a trace remains, and review strengthens that trace.

Ebbinghaus's results were long treated as anecdote, but in 2015 Murre and Dros faithfully reproduced the original experiment and published in PLoS ONE that the curve does in fact replicate. The forgetting curve is not a myth but a repeatedly verified phenomenon.

[ Forgetting curve: retention R over time ]

R = 100% ┤●
         │ ●
         │  ●●
  ~58%   │    ●●●          ← after 20 minutes
  ~44%   │       ●●●●      ← after 1 hour
         │           ●●●●●
  ~34%   │                ●●●●●●●        ← after 1 day
  ~25%   │                       ●●●●●●●● ← after 6 days
         └──────────────────────────────────► time

Ebbinghaus exponential model:   R = e^(-t/S)
  R = retrievability (0 to 1)
  t = time elapsed since last review
  S = stability of the memory — grows with each review

The variable that matters is stability S. Each review increases S and flattens the curve, so the time until the next forgetting grows longer. This principle is the mathematical basis of spaced repetition.

2. The Science of Spaced Repetition — The Distributed Practice Effect

For the same total study time, spreading it across several days (distributed practice) beats cramming it into one session (massed practice) by a wide margin. This is the distributed practice effect, or the spacing effect.

Cepeda et al. (2006) synthesized 254 experiments with roughly 14,000 participants and confirmed that distributed practice is on average substantially better than massed practice. Even more interesting is the follow-up, Cepeda et al. (2008): the optimal review gap is proportional to how long you want to remember (the target retention interval). Reviewing at roughly 10 to 20 percent of the target interval was most efficient.

The key point: reviewing just before you forget is most efficient. Review too often and you waste effort (re-seeing what you already know); review too late and you must relearn from scratch. Because no human can compute this "right timing" for hundreds of items, we need an algorithm.

3. From SM-2 to FSRS — The Evolution of Spaced-Repetition Algorithms

The attempt to automate spaced repetition began in the 1980s with Piotr Woźniak in Poland. The SM-2 algorithm (1987) of his SuperMemo became the root of countless later systems, including Anki and Mnemosyne. SM-2 assigns each card an "E-Factor" (ease) and adjusts the next interval based on the quality of your response.

The near-40-year standard SM-2 had limits. It grew intervals by simple multiplication and could not finely reflect individual differences or per-card difficulty. The answer is FSRS (Free Spaced Repetition Scheduler). FSRS uses the DSR model, grounded in the stochastic-optimization work of Ye et al. (2022) presented at KDD and in large-scale learning logs (MaiMemo data). DSR models each card's memory state with three variables: Difficulty, Stability, and Retrievability.

# SM-2 (SuperMemo 2, 1987) — the archetype of interval scheduling
# q = response quality (0 to 5), EF = ease factor (min 1.3)
if q >= 3:                      # correct
    if n == 1:   I = 1          # 1 day
    elif n == 2: I = 6          # 6 days
    else:        I = round(I_prev * EF)
    n += 1
else:                           # wrong -> start over
    n = 1
    I = 1
EF = EF + (0.1 - (5 - q) * (0.08 + (5 - q) * 0.02))
if EF < 1.3:
    EF = 1.3

# FSRS (2022+) — DSR model: Difficulty, Stability, Retrievability
# Power forgetting curve (FSRS-4.5 onward):
#   R(t, S) = (1 + (19/81) * t / S) ** (-0.5)
#   designed so that R = 0.9 when t = S
# 19 parameters w[0..18] are optimized from your own review logs

The strength of FSRS is personalization. It learns from your actual review history (right/wrong, response time) to fit 19 parameters to you. As a result, at the same target retention it often cuts review load 20 to 30 percent compared to SM-2. FSRS was integrated as an official feature starting with Anki 23.10 in 2023, so today you can enable it with a single checkbox.

4. Retrieval Practice vs Rereading — Karpicke & Roediger 2008

If you had to name the single most important experiment in the study of memory technique, it would be the 2008 Science paper by Karpicke and Roediger. Participants memorized 40 Swahili-English word pairs, then split into four conditions. The core contrast was "keep studying (rereading)" versus "keep testing (retrieval)."

The final test a week later was shocking. The group that repeatedly retrieved (tested) remembered about 80 percent, while the group that kept restudying already-correct words remembered about 35 percent. Study time was identical. What made the difference was not "reading again" but "pulling it out of your own head."

This phenomenon is called retrieval practice or the testing effect. The very act of retrieving a memory strengthens it. Rereading, by contrast, only inflates the fluency illusion: because the text flows easily, you feel you know it, but the moment you close the book, nothing comes.

The lesson is clear. Instead of underlining the textbook, close it and explain to yourself what you just read. Before flipping a flashcard, always retrieve the answer first. Feeling difficulty is normal, and that difficulty is the learning.

5. The Testing Effect — A Test Is Not Assessment, It Is Learning

We think of tests only as "a tool to check what we already learned." But research shows the test itself is a powerful learning tool. Roediger and Karpicke (2006) named this "test-enhanced learning."

Karpicke and Blunt (2011) went further. They compared retrieval practice with building elaborate concept maps. Students expected concept mapping to be more effective, but in fact the retrieval-practice group learned significantly more. Even the students' own predictions were the exact reverse. Our intuitions about what works are systematically wrong.

Dunlosky et al. (2013) reviewed ten learning techniques and graded their utility. Only practice testing and distributed practice earned a "high utility" rating. The rereading and highlighting that students use most were rated "low utility." This contrast is the theme running through this entire article.

6. Optimal Anki Setup — FSRS Parameters and New-Card Count

The most popular tool for turning theory into practice is Anki — free (except the iOS app), open source, with built-in FSRS support. The most common reason beginners fail is misconfiguring it so that reviews pile up beyond what they can handle. Below is a sustainable starting point.

# Anki deck options (recommended starting point)
FSRS: on (officially supported in Anki 23.10+)
Desired retention: 0.90         # target retention rate 90%
New cards/day: 10-20            # a sustainable range
Maximum reviews/day: 200-9999   # generous so you don't fall behind
Learning steps: 1m 10m          # new-card learning steps
Optimize FSRS parameters: once a month (after 1000+ reviews accrue)

# Warning: raising Desired retention to 0.97
#   explodes review load by 2-3x. 0.85-0.92 is realistic.

The most common mistake is setting the target retention too high. Going from 0.90 to 0.97 improves memory by only a few percentage points while review load doubles or triples. For most learning, 0.85 to 0.90 is the optimal balance. Only raise it temporarily when you need certainty, such as right before an exam.

New-card count also demands care. One card means dozens of future reviews. Add 20 cards a day and in a few months daily reviews can exceed 200. A volume you can steadily digest every day beats cramming in 100. FSRS parameter optimization is fine about once a month, after at least 1,000 reviews have accumulated.

7. How to Write Good Cards — The Minimum Information Principle

Even with the same algorithm, card quality drives the outcome. The heart of Woźniak's "Twenty Rules of Formulating Knowledge" is the minimum information principle: one card should hold one fact. The simpler the item, the more reliably the brain can schedule it.

A bad example is a card like "What are the five main causes of World War II?" Miss one of the five and it counts as "wrong," resetting the whole thing. The better approach is to split it into five small cards, or turn each item into a cloze (fill-in-the-blank) card.

In short, the four principles of a good card are: split it atomically, demand active retrieval, minimize context without being ambiguous, and add visuals when possible. Time spent making good cards is an investment that greatly reduces future review time.

8. The Memory Palace and Method of Loci — Placing Knowledge in Space

The memory palace, known academically as the method of loci, is a technique over 2,000 years old. Legend traces it to the ancient Greek poet Simonides, who identified bodies in a collapsed banquet hall by recalling where each guest had been seated. Roman orators systematized it to memorize long speeches (Cicero, and the treatise Rhetorica ad Herennium).

The principle works like this. Choose a deeply familiar space such as your home or your commute, and fix an ordered set of spots within it (front door, shoe rack, living-room sofa...). Place each item to memorize at a spot as a vivid, bizarre image. Later you walk through that space in your mind and collect the images. Because the human brain evolved to remember space and imagery far better than abstract lists, this method is powerful.

The memory palace is especially strong for ordered information (speeches, lists, a deck of cards, digits of pi). For understanding an interconnected system of concepts, retrieval practice and elaboration are better. The two are complementary, not competitors.

9. The Brains of Memory Champions — Training, Not Talent

Are world memory champions born with special brains? Maguire et al. (2003) tested exactly this. Scanning superior memorizers with fMRI, they found the brains were structurally no different from ordinary people, and general intelligence was unremarkable. What differed was which regions activated. While memorizing, the champions heavily used areas involved in spatial memory and navigation (the hippocampus, retrosplenial cortex, and others). Nearly all of them were using the method of loci.

Dresler et al. (2017) published even more decisive evidence in Neuron. When ordinary people with no memory training practiced the method of loci for six weeks, their memory scores approached champion levels, and their brain connectivity patterns grew to resemble those of champions. Even four months after training stopped, much of the improved memory persisted.

The conclusion supports this article's premise. Exceptional memory is not innate but a skill built through training. Learn the method and practice, and anyone can improve dramatically.

10. Chunking — Miller's 7±2 and Expert Chunking

George Miller's 1956 paper "The Magical Number Seven, Plus or Minus Two" is one of the most cited pieces in psychology. He proposed that human working memory can hold only about 7 (±2) items at once. (Later Cowan (2001) revised the true capacity closer to 4.) This is why phone numbers are broken into groups of three or four digits.

The technique that gets around this limit is chunking: grouping several items into meaningful units. The eight digits "1-9-4-5-2-0-2-6" are a burden, but grouped into "1945" and "2026" (say, a historical year and this year), they become easy. The chunk itself is processed as a single item, saving working memory.

Much of expert skill comes from chunking, as shown in the famous chess study by Chase and Simon (1973). Chess masters could memorize the piece positions of a real game in seconds, but for randomly placed pieces they were no better than novices. The master's edge was not photographic memory but the ability to recognize meaningful patterns (chunks). Ericsson et al. (1980) trained an ordinary student, "SF," to recall 79 digits in order; the trick was chunking the digits into running times he knew.

11. Dual Coding — Combining Text and Images

Allan Paivio's dual coding theory holds that our brains process verbal and visual information through two separate channels. Encode the same concept both in words and as an image and you get two retrieval routes, making the memory sturdier. Forget one and you can recall via the other.

The practical implication is large. Studying with pure text alone yields lower retention than adding diagrams, pictures, mind maps, and hand-drawn sketches. But the visuals must be directly connected to the content, not decorative. Irrelevant images actually distract (repeatedly confirmed in Richard Mayer's multimedia-learning research).

Putting an image on a flashcard, drawing a concept by hand, reconstructing a process as a flowchart — these are all applications of dual coding. The question "if I drew this concept as a single picture, what would it be?" goes beyond mere visualization; it is a powerful learning act that forces deep processing of the content.

12. Interleaving vs Blocked Practice — Mix It to Make It Last

Doing many problems of the same type together is blocked practice; mixing several types is interleaving. Intuitively, finishing one topic completely before moving on (blocked) seems better, and during practice blocked practice does feel like it goes more smoothly. But for long-term retention and transfer, interleaving wins.

Rohrer and Taylor (2007) showed this in learning math problems. The group that mixed problem types had a lower accuracy during practice, but far outperformed the blocked group on a test a week later. The follow-up by Taylor and Rohrer (2010) replicated the result. Interleaving works because you must judge for yourself "which type is this?" each time. That discrimination is itself retrieval and deeper learning.

This connects to Robert Bjork's concept of "desirable difficulties." Conditions that make learning briefly harder (interleaving, spacing, retrieval) reduce performance in the moment but strengthen long-term memory. Conversely, conditions that make learning smooth (blocking, cramming) seem to go well now but are quickly forgotten. Comfort and learning often point in opposite directions.

13. Levels of Processing and Elaborative Encoding

Craik and Lockhart's (1972) levels of processing theory holds that memory strength depends on how deeply you processed the information. Semantic (deep) processing that engages a word's meaning lasts far longer than shallow processing of its surface, such as its typeface.

The key tool of deep processing is elaborative encoding: connecting new information to what you already know, generating examples, and asking "why is this so." An isolated fact has only one retrieval cue, but a fact woven into many concepts has many retrieval routes and is easier to recall. Memory is less a warehouse of information than a web of connections.

The practice is simple. When you learn a new concept, ask yourself: "What that I already know is this similar to?" "What is a concrete example?" "If this were false, what would change?" Such questions weave the information tightly into your existing knowledge network. The elaborative interrogation and Feynman technique in the next two sections are concrete ways to apply this principle.

14. Elaborative Interrogation — Asking "Why?"

Elaborative interrogation is the technique of persistently asking "why is this so?" about a new fact. Instead of simply memorizing "polar bears are white," you ask "why are polar bears white?" and generate an answer (camouflage in the snow is advantageous). The causal explanation you build connects the fact to prior knowledge and strengthens the memory.

Pressley et al. (1987) showed that a group using elaborative interrogation remembered facts significantly better than a rote-memorization group. In the Dunlosky et al. (2013) review, elaborative interrogation earned a "moderate utility" rating. It is not as powerful as retrieval practice, but it needs almost no preparation and is especially useful where understanding is shallow.

One caveat: the answer must be accurate. Build a plausible-but-wrong explanation and you instead cement a misconception. So elaborative interrogation should be used alongside reliable sources, verifying the explanations you generate.

15. The Feynman Technique — Learning by Teaching

Named after physicist Richard Feynman, the Feynman technique is a method of "explaining something simply enough that a child could understand." It has four steps. (1) Pick a concept. (2) Explain it in very plain words with no jargon. (3) Find where the explanation stalls — the parts you do not truly know. (4) Fill those gaps with sources and explain again.

The technique is powerful because it destroys the "illusion of knowing." You may feel you understand in your head, but the moment you try to explain out loud, the gaps in your logic surface. Rendering it in plain words deepens the level of processing, and the framing of explaining to someone else becomes powerful retrieval practice. It aligns with the "protégé effect" in learning science: study while expecting to teach someone, and you learn better.

It is especially useful for developers. Explain the algorithm or system design you just learned to a colleague (or a rubber duck). If you can answer "why does this part behave this way?" in a code review without stumbling, you truly understand it. If you cannot explain it, you do not yet know it.

16. Sleep and Memory Consolidation — From Hippocampus to Neocortex

Pulling an all-nighter to cram is, from the perspective of memory science, the worst possible choice. This is because consolidation — the settling of what you learned into long-term memory — happens mainly during sleep. As the sweeping review by Rasch and Born (2013) summarizes, sleep is not wasted memory time but an essential stage that completes memory.

The mechanism is this. During the day, new experiences are stored temporarily in the hippocampus. During deep (slow-wave) sleep, the brain replays the day's experiences and gradually transfers them to long-term storage in the neocortex. This is called systems consolidation and is summarized as the hippocampus-neocortex dialogue. Diekelmann and Born (2010) noted that slow-wave sleep is especially important for declarative memory such as facts and knowledge, and REM sleep for procedural and emotional memory.

The practical implications are strong. First, cutting sleep to add study time is a net loss, because it breaks consolidation. Second, reviewing right before sleep maximizes the consolidation benefit. Third, a full night's sleep before an exam beats an all-nighter fueled by caffeine. Sleep is not the opposite of studying but a part of it.

17. The Power of Naps — A 20-Minute Reboot

Nighttime sleep is not the only thing that matters. In a Nature Neuroscience study, Mednick et al. (2003) showed that "a nap is as good as a night." Repeating a visual task, performance gets worse across the day (perceptual fatigue); a nap reverses this decline, and a sufficiently long nap (including slow-wave and REM sleep) produced learning gains rivaling a full night's sleep.

The effect of a nap depends on its length. A short nap restores alertness and focus, while a nap that reaches slow-wave sleep aids declarative memory consolidation. Sleep too long, however, and you can wake from deep sleep into a groggy state (sleep inertia) that lingers.

Placing a short nap between study sessions can prevent the afternoon dip in focus and aid the consolidation of what you learned in the morning. "Push through the drowsiness and study more" is generally worse than "nap briefly and retrieve refreshed."

18. Exercise and BDNF — Fertilizer for the Brain

There is a neuroscientific reason not to sit at your desk all day for the sake of memory. Aerobic exercise raises the secretion of a protein called BDNF (brain-derived neurotrophic factor). BDNF is a kind of "fertilizer for the brain" that supports neuron survival and growth and strengthens synapses. Cotman et al. (2007) laid out the pathway by which exercise promotes brain health through BDNF.

Animal studies are even more direct. van Praag et al. (1999) showed that running mice had increased neurogenesis in the hippocampus along with improved learning and long-term potentiation (LTP). As we saw, the hippocampus is the key region where new memories are first stored. In human studies too, Erickson et al. (2011) reported in PNAS that a year of aerobic exercise increased hippocampal volume and improved memory.

The practice is not hard. Simply placing light aerobic exercise (brisk walking, jogging 20 to 30 minutes) before or after studying helps encoding and consolidation. Along with sleep and retrieval practice, exercise belongs to the "biological foundation" that holds memory up. No matter how good your SRS is, efficiency drops if you do not sleep and do not move.

19. Applying It to Language Learning — A Vocabulary System

The classic application that ties all these principles together is foreign-language vocabulary learning. A language is a task of getting thousands of individual items (words) into long-term memory — a perfect fit for SRS. Here is an evidence-based vocabulary system.

• Learn in frequency order. Start with the most frequent words and the same effort lets you understand more sentences. The top 2,000 words cover a large share of everyday conversation.
• Make cards from example sentences, not bare words. A word in context lasts longer than an isolated word (elaborative encoding). Cloze (fill-in-the-blank) sentence cards work well.
• Attach images. Use dual coding — link to a picture rather than a translation so retrieval happens directly, without routing through your native language.
• Retrieve first, then check. Before looking at the back of the card, always retrieve the pronunciation and meaning yourself. This is the decisive difference from rereading.
• Encode with sound. Say it out loud and listen to native audio so the auditory channel is engaged too.

The memory palace applies to language learning as well. Place abstract grammar rules or irregular conjugations in a space, or link etymology to imagery (for example, build a bizarre scene from a native-language word that sounds similar) to create retrieval cues. The key is to review briefly every day, retrieval-first, just before you forget.

20. A Spaced-Repetition System for Developers — Memorizing Commands and Concepts

Developers need memory too. People say "why memorize when you can just search," but there is a large difference in productivity and flow between instantly recalling a frequently used command or core concept and looking it up every time. Searching breaks flow; retrieval keeps it. SRS is ideal for getting knowledge like shell commands, shortcuts, algorithmic complexity, language syntax, and API signatures into long-term memory.

# Example developer cards (split them atomically)

Q: Which command edits only the last commit message in git?
A: git commit --amend

Q: Which Linux command follows a file's contents in real time?
A: tail -f filename

# Cloze card examples — Anki syntax (double curly braces)
{{c1::O(log n)}} — time complexity of binary search on a sorted array
SELECT col FROM t {{c1::WHERE}} cond {{c2::GROUP BY}} col

# Principle: one command = one card. No "list of 10 options."

But avoid memorization without understanding. Much of developer knowledge follows naturally once you grasp the principle. What to memorize with SRS is "things you understood but forget from rare use" (uncommon flags, regex syntax, shortcuts) and "things useful only when retrieved instantly even though you know the principle" (complexity, data-structure properties). A two-tier structure works well: understand the concept itself with the Feynman technique, then layer the details on top with SRS.

It is also useful for technical-interview prep. Turn the core properties of data structures and algorithms, system-design patterns, and language-specific pitfalls into cards and spread the learning over several months; it sticks far more solidly than cramming.

21. Common Myths — The Learning-Styles Myth and Rereading

Discarding methods that do not work matters as much as adopting ones that do. The most stubborn myth is "learning styles" — the belief that each person is a visual, auditory, or kinesthetic learner and should study in their own type. After reviewing it, Pashler et al. (2008) concluded that there is essentially no evidence that matching instruction to your type (the meshing hypothesis) helps. A learning-styles quiz may be fun but does not improve learning efficiency. What matters is not the type but the method that fits the content (maps by sight, pronunciation by ear).

The second myth is repeated rereading and highlighting. As we saw, these only inflate the fluency illusion and yield low retention. In the Dunlosky et al. (2013) grading, rereading, highlighting, summarizing, and keyword mnemonics all landed at "low" or "moderate" utility. The problem is that these are the easiest and most common methods.

The lesson is singular. If studying feels easy and comfortable, it is usually a low-efficiency method. Methods that feel moderately hard — retrieval, spacing, interleaving — are what build long-term memory.

22. A 4-Week Practical Memory-Optimization System

Let us tie the theory into a single routine. Below is a four-week system you can use when learning a new domain of knowledge (a language, a certification, a tech stack).

Week 1 — Understand and build cards. First, grasp the big picture with the Feynman technique. Turn what you understood into atomic cards following the minimum information principle. Invest time in card quality. Add 10 to 20 new cards a day.

Week 2 — Make retrieval a habit. Review in Anki at the same time every day. Always retrieve the answer first before flipping the card. Learn by mixing in new types (interleaving). Reinforce weakly understood parts with elaborative interrogation.

Week 3 — Deepen and apply. Actually use what you learned (for a language, write and speak sentences; for tech, write code). Reinforce hard concepts with the memory palace or dual coding. If the target retention is overwhelming, cut new-card count to digest the reviews.

Week 4 — Review and optimize. Optimize FSRS parameters (if enough review history has accrued). Find weak spots with self-testing and split failing cards into finer pieces. Repeat this cycle, gradually expanding the intervals.

Do not forget the biological foundation of all this. Sleep 7 to 9 hours daily, do aerobic exercise, and review lightly right before sleep. When method and biology go together, memory is maximized.

Conclusion — Memory Is a Skill

The conclusion that 140 years of memory research converges on is simple. Good memory is not innate but a skill you train. And the core of that skill reduces to a few counterintuitive principles: review just before you forget (spaced repetition), do not read with your eyes but pull it out yourself (retrieval practice), and choose the moderately hard method over the comfortable one (desirable difficulties).

Add the memory palace and chunking to expand capacity, dual coding and elaboration to deepen processing, and sleep and exercise to strengthen the biological foundation, and you end up learning in tune with the brain's natural design. The familiar habits of underlining and rereading, by contrast, are comfortable but futile.

Start with just three things today. First, swap rereading for self-testing. Second, review briefly every day with an SRS such as Anki. Third, do not cut sleep to study. Change the method and you can remember several times more in the same time. Memory is not a matter of talent but a matter of method.

References

• Ebbinghaus, H. (1885/2013). Memory: A Contribution to Experimental Psychology (reprint). Annals of Neurosciences, 20(4), 155–156. DOI:10.5214/ans.0972.7531.200408
• Murre, J. M. J., & Dros, J. (2015). Replication and Analysis of Ebbinghaus' Forgetting Curve. PLoS ONE, 10(7), e0120644. DOI:10.1371/journal.pone.0120644
• Roediger, H. L., & Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3), 249–255. DOI:10.1111/j.1467-9280.2006.01693.x
• Karpicke, J. D., & Roediger, H. L. (2008). The Critical Importance of Retrieval for Learning. Science, 319(5865), 966–968. DOI:10.1126/science.1152408
• Karpicke, J. D., & Blunt, J. R. (2011). Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping. Science, 331(6018), 772–775. DOI:10.1126/science.1199327
• Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis. Psychological Bulletin, 132(3), 354–380. DOI:10.1037/0033-2909.132.3.354
• Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T., & Pashler, H. (2008). Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention. Psychological Science, 19(11), 1095–1102. DOI:10.1111/j.1467-9280.2008.02209.x
• Miller, G. A. (1956). The Magical Number Seven, Plus or Minus Two. Psychological Review, 63(2), 81–97. DOI:10.1037/h0043158
• Cowan, N. (2001). The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87–114. DOI:10.1017/S0140525X01003922
• Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55–81. DOI:10.1016/0010-0285(73)90004-2
• Ericsson, K. A., Chase, W. G., & Faloon, S. (1980). Acquisition of a memory skill. Science, 208(4448), 1181–1182. DOI:10.1126/science.7375930
• Maguire, E. A., Valentine, E. R., Wilding, J. M., & Kapur, N. (2003). Routes to remembering: the brains behind superior memory. Nature Neuroscience, 6(1), 90–95. DOI:10.1038/nn988
• Dresler, M., et al. (2017). Mnemonic Training Reshapes Brain Networks to Support Superior Memory. Neuron, 93(5), 1227–1235. DOI:10.1016/j.neuron.2017.02.003
• Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(6), 671–684. DOI:10.1016/S0022-5371(72)80001-X
• Paivio, A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of Psychology, 45(3), 255–287. DOI:10.1037/h0084295
• Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498. DOI:10.1007/s11251-007-9015-8
• Taylor, K., & Rohrer, D. (2010). The effects of interleaved practice. Applied Cognitive Psychology, 24(6), 837–848. DOI:10.1002/acp.1598
• Diekelmann, S., & Born, J. (2010). The memory function of sleep. Nature Reviews Neuroscience, 11(2), 114–126. DOI:10.1038/nrn2762
• Rasch, B., & Born, J. (2013). About Sleep's Role in Memory. Physiological Reviews, 93(2), 681–766. DOI:10.1152/physrev.00032.2012
• Mednick, S., Nakayama, K., & Stickgold, R. (2003). Sleep-dependent learning: a nap is as good as a night. Nature Neuroscience, 6(7), 697–698. DOI:10.1038/nn1078
• Cotman, C. W., Berchtold, N. C., & Christie, L.-A. (2007). Exercise builds brain health: key roles of growth factor cascades and inflammation. Trends in Neurosciences, 30(9), 464–472. DOI:10.1016/j.tins.2007.06.011
• van Praag, H., Christie, B. R., Sejnowski, T. J., & Gage, F. H. (1999). Running enhances neurogenesis, learning, and long-term potentiation in mice. PNAS, 96(23), 13427–13431. DOI:10.1073/pnas.96.23.13427
• Erickson, K. I., et al. (2011). Exercise training increases size of hippocampus and improves memory. PNAS, 108(7), 3017–3022. DOI:10.1073/pnas.1015950108
• Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning Styles: Concepts and Evidence. Psychological Science in the Public Interest, 9(3), 105–119. DOI:10.1111/j.1539-6053.2009.01038.x
• Pressley, M., McDaniel, M. A., Turnure, J. E., Wood, E., & Ahmad, M. (1987). Generation and precision of elaboration effects on intentional and incidental learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13(2), 291–300. DOI:10.1037/0278-7393.13.2.291
• Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving Students' Learning With Effective Learning Techniques. Psychological Science in the Public Interest, 14(1), 4–58. DOI:10.1177/1529100612453266
• Ye, J., Su, J., & Cao, Y. (2022). A Stochastic Shortest Path Algorithm for Optimizing Spaced Repetition Scheduling. KDD '22, 4381–4390. DOI:10.1145/3534678.3539081
• Woźniak, P. A., & Gorzelańczyk, E. J. (1994). Optimization of repetition spacing in the practice of learning. Acta Neurobiologiae Experimentalis, 54(1), 59–62. (PMID:8023714)
• Open Spaced Repetition — FSRS4Anki scheduler (open source). github.com/open-spaced-repetition/fsrs4anki
• Woźniak, P. A. Twenty rules of formulating knowledge. supermemo.com