Make It Stick: An In-Depth Guide to Learning That Lasts

'Make It Stick' was co-authored by Peter C. Brown, Henry L. Roediger III, and Mark A. McDaniel, and the English version was published in 2014. It doesn't sell any magical tricks; instead, it organizes decades of cognitive psychology research on memory, forgetting, practice, testing, and transfer into a practical system that everyday learners can apply.

The book isn't concerned with getting an easy, smooth learning experience in the moment, but with how to make knowledge retrievable and usable years later. Reading with this question in mind makes you reconsider many commonly accepted learning habits.

1. How learning feels doesn't equal how effective it is

The key to understanding the whole book is distinguishing between your subjective feelings during learning and the actual results afterward. Reading a passage and understanding it, recognizing words, following the teacher's explanation, or getting easier at the same type of problem—these can all give you a strong sense of mastery, but they don't prove that the knowledge has entered long-term memory.

Familiarity, fluency, and being able to independently retrieve and apply knowledge are three different things.

Reading the same thing ten times abandon = to give upThe last time it might feel effortless, mostly involving recognition. But the next day, when you see 'abandon' and have to recall meaning yourself, the task becomes retrieval. Exams, writing, and real work rely more on the latter.

2. Why rereading is tempting but often inefficient

Rereading easily creates a sense of progress. The first time feels hard, the second time you start to understand, the third you become familiar with it, and by the fourth you might mistakenly think you've mastered it. In reality, the change is sometimes just that the material feels more familiar, not that the knowledge is ready for independent use.

Underlining, highlighting, and emphasizing are not useless. Of course, first encountering material requires input, but the problem is that reading alone can't carry the entire learning process. Without retrieval and feedback, just input usually only accumulates a lot of things you've once understood.

A more complete learning structure should be:

Input → Try to retrieve → Expose errors → Targeted relearning → Space out intervals → Retrieve again

An inefficient structure often looks like:

Input → Input → Input → Input → Exam

3. The first big core: Retrieval Practice

If you only keep one method, active recall is the one worth keeping. Retrieval practice isn’t about looking at the answer again; it’s about temporarily hiding the answer and forcing your brain to regenerate the knowledge on its own. Testing here doesn’t just measure memory, it also shapes memory — that’s the testing effect.

You don’t have to wait for a formal exam to do retrieval practice. It can be simple:

• After finishing a page of a book, close it and write down the three most important points.
• After learning a concept, explain it in your own words without looking at the material.
• When studying history, independently explain the causes and consequences of an event.
• When studying psychology, distinguish between operant conditioning and classical conditioning.
• When memorizing English words, see 'inevitable' and first recall its meaning before trying to make a sentence.

Every effortful recall trains your brain to find the path to that knowledge in the future.

4. Knowledge is reinforced not just through input, but also through retrieval

A common learning instinct is to fully memorize content before starting tests. Learning science offers a different order: testing itself helps memory, so you don’t have to wait until you fully know something to recall it; starting retrieval while still unfamiliar is even more valuable.

Two people study the same article: one reads it straight through four times, while the other reads it, closes the material, tries to recall, then checks, and recalls again. The first person usually feels it’s smoother, while the second often faces moments where they can’t remember or answer fully. These uncomfortable moments actually reveal the weak spots in memory. A few days later, when tested again, the person who actively retrieved the information often retains it better.

Being comfortable during practice doesn’t mean better long-term results; constantly finding gaps during practice doesn’t mean you’re learning poorly.

5. Desirable difficulties: Performance doesn’t equal Learning

Not being able to recall is usually seen as a learning failure, but the right level of difficulty might be exactly where learning happens. Reviewing after a delay, hiding answers before recalling, mixing different types of questions, and trying first before checking all reduce fluency in the moment but can improve long-term retention. This phenomenon is often called Desirable Difficulties.

Two concepts need to be distinguished:

• Performance：Performance at the moment of completing a task.
• Learning：How much knowledge remains over time and whether it can be used in a new context.

Doing 20 identical quadratic equations in a row, with the last ones almost done instantly, only shows good immediate performance. A month later, if those are mixed into 30 different question types and you can’t even figure out which method to use, the long-term learning wasn’t as strong as it felt initially.

6. The second core: Spacing — distributed learning

The same ten hours of study spread across multiple days usually supports long-term retention better than doing it all in one day. The key change brought by spacing is forgetting.

When you immediately repeat A → B right after learning, the answer is still in working memory, and the brain hardly needs to search. Seeing A → ? after a gap requires searching, reconstructing, and trying, and successfully recalling B this way trains your brain more than mindless repetition.

There is no absolute schedule that works for everyone and for all content. One day, three days, seven days, fourteen days, and thirty days can serve as starting points, but what really matters is the difficulty of retrieval, accuracy, and how long you want to retain it.

Seven, the danger of cramming lies in its short-term effectiveness being too good

Cramming isn't completely useless. Studying intensively for six hours tonight before an exam tomorrow might result in decent short-term scores. The danger is that this outcome can make people believe that concentrated study is the most effective, while in reality, it only measures performance after 24 hours.

If you don't use this knowledge after the exam, concentrated practice sometimes makes practical sense. But TOEFL vocabulary, the math foundation needed for subsequent courses, professional knowledge for graduate studies, and job skills should be optimized for usability months or even years later.

What really needs to be compared is not how fast you learn today, but how much knowledge you can still recall a year from now.

Eight, the third core: Interleaving – mixed practice

Suppose you need to learn three types of math problems: A, B, and C. Blocked practice would do all of A first, then all of B, and finally all of C; interleaved practice mixes A, B, and C together.

The value of interleaving lies in requiring the learner to first identify what type of problem it is, then choose a method. When a textbook arranges 20 Pythagorean theorem problems by chapter, the chapter titles quietly provide hints, and learners just follow the algorithm. Real exams don’t indicate which formula to use, and problems at work don’t come pre-labeled with categories.

Full problem-solving ability involves two layers:

• Identifying the problem:See what structure it belongs to.
• Choosing and executing a method:Call upon the right knowledge to complete the task.

Blocked practice often trains only the second layer, while interleaved practice trains classification, discrimination, and strategy selection simultaneously.

Nine, the difference between students and experts often lies in first identifying the problem

Beginners often ask how to solve a problem when they first see it, while experts quickly recognize what type of problem it is. Doctors don’t get a card with the disease name written on it when facing patients; researchers, facing complex phenomena, also need to figure out which theory can explain them.

Interleaving practice isn’t about randomly shuffling all content. Frequently switching between English, calculus, swimming, and Chinese history isn’t necessarily valuable. It works better to mix related, easily confused, or comparable categories, like classical conditioning, operant conditioning, and observational learning, or areas like triangle area, Pythagorean theorem, and cosine law.

The goal of interleaving isn’t to create confusion but to practice knowing when to use which concept.

Tenth, the fourth core: Generation — try first, then learn.

Generation means creating an answer first, then looking at the official explanation. When learning about hindsight bias, instead of checking the definition right away, it’s better to first guess what kind of psychological bias it might describe based on real-life situations, then verify.

Trying first activates your existing knowledge, forms predictions, and exposes gaps. Later information isn’t isolated input anymore; it corrects your previous guesses. Learning thus shifts from passively receiving information to actively solving problems.

It’s simple to do: try first, guess first, explain first, then check the answer.

Eleventh, desirable difficulties don’t mean more errors are better.

Difficulties only have value if they help you learn correctly. Letting a complete beginner independently derive quantum mechanics, or making retrieval so hard that they can only answer 'I don’t know' every time, won’t automatically improve memory.

The ideal situation is that the answer isn’t immediately in sight, requiring ten or fifteen seconds of careful search, and finally being able to recall or come close to the correct answer. Struggling in this way reinforces retrieval paths. If you can’t answer at all for a long time, reduce the difficulty, add clues, re-understand the material, and then gradually remove the prompts.

Twelfth, the fifth core: Elaboration — connecting new knowledge to old knowledge.

Memory is more like a network than a categorized warehouse. The more connections a concept forms with existing knowledge, causal relationships, concrete examples, and similar concepts, the more paths you usually have to find it in the future.

When learning about the spacing effect, don't just remember the name 'spacing effect'; you can continue to ask:

• Why is spacing usually helpful? Because it allows a certain degree of forgetting to occur.
• Why can forgetting sometimes be helpful? Because relearning requires retrieval.
• Why is retrieval valuable? Because successful retrieval itself is a learning event.

Thus, spacing, forgetting, effortful retrieval, and memory strengthening form a causal structure, no longer just an isolated slogan.

Thirteen, self-explanation is more important than just understanding the answer.

If you can't explain a concept in your own words, you probably haven't truly understood it. Just feeling familiar when seeing 'classical conditioning' doesn't mean you can explain it clearly with examples a child can understand.

After learning a concept, you can systematically answer three questions:

• What is it?Train on definition and core features.
• Why is it the way it is?Train on causal structures.
• How does it differ from similar concepts?Train discrimination ability.

These three types of tasks are much closer to real use than looking at the definition three more times.

Fourteen, the sixth core: Reflection—turning experience into knowledge.

Spending a few minutes reflecting after learning allows an experience to settle into reusable knowledge. At the end of the day, you can review: what did I truly learn today, what were the hardest parts, what mistakes did I make, why did those mistakes happen, and what clues should I pay attention to next time I face similar problems.

Without reflection, a person might have ten years of work experience but still be repeating the same year's experience. By adding review sessions, experience gradually turns into principles, judgment, and strategies.

the seventh core: Calibration — accurately judging what you know

Learning ability depends not only on how much you know, but also on whether you can correctly judge how much you know. This kind of self-assessment belongs to metacognition.

The most dangerous state is not not knowing, but thinking you know when you don’t. In the former case, you know you need to keep learning; in the latter, you may stop too early because of a false sense of mastery. Truly reliable judgment comes from untimed tests, delayed tests, and varied applications, not familiarity.

Knowledge illusion: input ability doesn’t equal output ability

When looking at the answer, you think you can do it, but covering the answer you can’t start; when someone explains it, it seems easy, but when it’s your turn you can’t explain it; when reading an English article, you know most of the words, but can’t use them in writing. All these show that input ability is not the same as output ability.

Another value of retrieval practice is that it helps measure learning and form a feedback loop:

Learning → self-test → identify gaps → targeted relearning → test again → adjust judgment

This loop not only accumulates knowledge, but also trains how to manage your own learning.

Mistakes are not failures, they are information

In an effective learning framework, tests aren’t meant to prove intelligence, but to find knowledge gaps. If you get 8 questions wrong out of 30, the real value isn’t just copying the correct answers, but judging which type each mistake belongs to:

• The knowledge point was never understood.
• A formula or fact was not recalled.
• The question conditions were misread.
• Two similar concepts were confused.
• A calculation or operational mistake was made.
• Mistaken question-type judgment caused the wrong strategy to be chosen.

Different mistakes require different ways to correct them, so classifying errors is usually more important than just organizing wrong questions.

18. Five Levels of Mastering Knowledge

Seeing something doesn’t mean you can use it. Based on the depth of learning, mastery can be divided into five levels:

The highest learning goal is the fifth level. Real-world problems don’t appear neatly chapter by chapter; true ability is seeing unfamiliar problems, recognizing patterns, recalling relevant knowledge from long-term memory, choosing methods, and solving them.

19. Why Doing Tons of Similar Questions Can Create Fake Experts

When continuously doing the same type of probability questions, chapter titles, the previous question, and just-used formulas all give environmental cues. Speeding up later doesn’t necessarily mean independent judgment has improved.

In real exams, probability questions might appear between calculus and statistics questions, without any reminders to use Bayes’ theorem. So in later learning stages, you should gradually remove the “training wheels”:

• Start by understanding methods with complete example questions.
• Reduce step-by-step hints and complete them independently.
• Mix similar question types and judge which method to use.
• Finally, tackle superficially unfamiliar but structurally similar problems.

This gradual removal of hints is closer to real ability than mechanically doing a hundred of the same type of questions.

20. People who grasp things quickly also tend to overestimate their learning outcomes

People who understand things quickly often feel a strong sense of mastery after just one look, then quickly move on to the next part. But speed of understanding and long-term memory are two different issues. Being able to follow the author’s argument at the moment only shows that your working memory can handle it; it doesn’t mean you’ll be able to reconstruct it independently three months later.

The easier the material is to understand, the more you need to use delayed retrieval to check if it’s really stuck. Mature learners focus not on looking smart, but on whether their learning system can consistently produce knowledge that can be used in the long term.

21. Eight counterintuitive learning comparisons

These comparisons collectively train a new learning intuition: a little difficulty now can sometimes make things easier later, but the key is that the difficulty helps with retrieval, discrimination, or transfer.

22. Turn the whole book into an executable seven-step loop

Step 1: Understand

Read your textbook normally, watch courses, or listen to explanations to first build a basic model of the content without needing to memorize every detail right away.

Step 2: Close the material

Close your book, pause the video, or cover the answers. Without this step, it’s easy to fall back on recognition instead of retrieval.

Step 3: Active Reconstruction

Recall what was just discussed, what the core points were, the reasons behind them, and try to come up with your own examples.

Step 4: Check

Reopen the materials and compare what was correct, what was missed, and what you misunderstood.

Step 5: Correct

Focus on relearning the parts you didn’t get right; there’s no need to reread the whole chapter indiscriminately.

Step 6: Spacing

Temporarily step away from this content, study something else, or let sleep help consolidate your memory.

Step 7: Re-recall

The next day, a few days later, and a few weeks later, recall everything again without prompts, gradually adding modified tasks.

23. Specific Example: Learning Psychology

When learning Classical Conditioning, an inefficient approach might be reading the textbook for twenty minutes, highlighting, copying definitions, and reading it again, then stopping after feeling like you understood it.

Using retrieval-based learning, you could read for fifteen minutes, then close the book and write down from memory the meaning of classical conditioning, and draw it out:

US → UR

CS + US

CS → CR

Then explain why Pavlov’s experiment demonstrates this process, compare it with the difference in operant conditioning, give a different real-life example other than dogs salivating, and judge whether advertisements might use this mechanism.

The next day, redraw it without looking at notes; three days later, mix Classical Conditioning and Operant Conditioning for judgment; two weeks later, distinguish theories when facing a new scenario. Training should gradually move from defining to recalling, explaining, distinguishing, applying, and transferring, so knowledge starts becoming skill.

24. Specific Example: Memorizing English Words

Take 'mitigate' as an example. Looking at 'mitigate = lessen' twenty times in a row only gives you a strong sense of familiarity. A better starting point is to use flashcards: write 'mitigate' on the front and 'lessen; ease' on the back, then actively recall the meaning before turning the card over.

When training further, you can add four types of tasks:

• 1. Say the meaning and make your own sentence.
• 2. Compare the differences between mitigate, reduce, alleviate, and relieve.
• 3. Complete the sentence: The new policy may ______ the impact of inflation.
• 4. Reappear after a few days and judge whether it fits in a new context.

This way, what you’re building is not just the correspondence between English words and Chinese meanings, but also context, collocation, discrimination, and multiple recall paths, getting closer to real language ability.

25. A concrete example: how to really finish a book

Reading from page one to page three hundred only shows that you completed the action of reading. A month later, if you cannot state the most important points of the book, the knowledge hasn’t automatically stuck just because you turned the last page.

A more effective approach is to pause and close the book after finishing a meaningful section, and answer these questions:

• - What problem is this section addressing?
• - What is the author’s conclusion?
• - What evidence supports the conclusion?
• - Are there possible counterexamples?
• - How does it relate to the previous content?

After finishing the whole book, write out the structure from memory without looking at the table of contents, compress the 300 pages into 20 key concepts, then compress them into 5 core principles, and finally form a knowledge map. Experts’ advantage usually lies not just in remembering more facts, but in having better-organized knowledge structures.

26. Why teaching others is very effective

Explaining things to others naturally involves extracting, organizing, interpreting, and giving examples. When the listener keeps asking why, gaps in knowledge will quickly become obvious. A lot of stuff seems clear when reading, but only when you actually try to explain it out loud do you realize the structure isn’t complete.

Even without a real audience, you can simulate teaching: talk to the air, record yourself, retell it to a study partner, or write it out as an article. The key isn’t the act of teaching itself, but reorganizing and producing knowledge without relying on materials.

27. Whether notes are useful depends on what your brain is doing

Taking notes isn’t inherently effective or ineffective; it depends on the cognitive activities happening during the process. If the teacher writes a sentence and the student copies it, it’s mostly transporting information; summarizing it in your own words adds a layer of processing; writing from memory without looking at materials and then mapping out the connections between concepts truly combines retrieval and organization.

For example, if you connect Retrieval, Spacing, Forgetting, Desirable Difficulty, and Long-term Retention in a concept map, notes stop being just a place to store knowledge and become a product of thinking.

28. Why flashcards work, and why they’re so easily wasted

Spaced repetition tools like Anki combine retrieval and spacing, making them great for practicing memory. But if cards are too vague, like just writing "Psychology chapter 3" on the front and cramming two pages of content on the back, they turn into mere answer-checking machines.

Good retrieval questions should be specific, for instance:

• What is retrieval practice?
• What is the core difference between classical conditioning and operant conditioning?
• Why does massed practice tend to create an illusion of mastery?

For content that needs to be understood, the cards should gradually upgrade from 'What' to 'Why and When'. When learning confirmation bias, you not only answer what it is, but also explain why it persists and be able to determine whether it falls under this bias in the case. This way, the cards train concepts, not just definitions.

29. The three stages of learning: Build, Strengthen, and Generalize

Phase One: Build

Knowledge is built through reading, attending classes, observing demonstrations, and understanding, with the goal of forming an initial model.

Phase Two: Strengthen

Knowledge is reinforced through active recall, spaced repetition, testing, and feedback, with the goal of enabling long-term extraction.

Stage Three: Generalize

Generalize knowledge through interleaved exercises, variant tasks, unfamiliar questions, and real problems, aiming to make it still usable after leaving the textbook.

Many studies stay at the Build stage, constantly watching videos, reading, and attending classes, but rarely get into Strengthen, and even less in Generalize. The result is accumulating a large amount of knowledge once understood, rather than the abilities you currently possess.

30. How to judge whether you have truly learned it?

Reliable mastery of standards should not be just about whether you understand them now, but should be checked layer by layer:

• After 24 hours, can you explain it without looking at the materials?
• Can it still be explained after a week?
• Can the problem be solved after slight changes?
• Can you still distinguish them when mixed with similar concepts?
• When there are no alerts in real environments, can you think of using it?

Only by being able to do all five is you getting closer to Mastery. Delay, change, and lack of prompts are all necessary conditions for testing long-term knowledge.

31. The boundaries of this method

Active recall, spacing, and interleaving can't solve all learning problems. Their effectiveness also depends on the nature of the material, existing knowledge, feedback quality, and task type, so you need to understand four boundaries when using them.

First, retrieval can't replace understanding.

If you don't understand Newton's second law and just memorize F = ma every day, you might just be reinforcing the string of characters. Retrieval practice is built on preliminary understanding and accurate representation.

Second, remembering knowledge isn't the same as creating knowledge.

Research, writing papers, proposing new theories, and solving open problems also require reasoning, creativity, critical thinking, and problem definition—they can't be fully reduced to flashcards.

Third, harder isn't always better.

Making learning unnecessarily complicated isn't beneficial difficulty. Difficulty is only worth keeping if it helps retrieval, discrimination, transfer, or building better knowledge structures.

Fourth, the theoretically optimal method isn't necessarily the best in practice.

No matter how scientific a method is, if you can't stick to it daily, its actual effect is near zero. A good learning system must fit into your life and be sustainable long-term.

32. Effortless doesn't equal high-quality learning.

The Chinese subtitle of "The Nature of Cognition" might make readers expect learning to be easy, but the book actually conveys the opposite: high-quality learning often comes with not recalling things, making mistakes, waiting, retrying, mixed problem types, difficulty, and uncertainty.

These experiences may not feel pleasant subjectively, but they might be closer to long-term learning than smooth sailing. The accurate summary is: learning that's slightly harder but in the right direction is often remembered longer.

33. The habit most worth keeping long-term: close the answers.

After reading, close the book, shut the PDF, pause the video, cover the answers, and then see what remains when there's no hint. This small action can quickly turn a recognition task into a retrieval task.

The parts you can reorganize after closing the book begin to become knowledge that truly belongs to you; the parts you can't recall accurately show exactly where the next round of learning should focus.

Thirty-four: The Ten Core Principles of "Make It Stick"

1. Don't judge whether you've learned something by familiarity; judge it by whether you can recall it independently.
2. Start actively recalling as soon as possible after learning; don't endlessly repeat input.
3. Space out your reviews; don't cram all your practice together.
4. Allow a moderate amount of forgetting because retrieving information itself is part of the training.
5. Mix similar types of problems to practice when to use which knowledge.
6. Try first, then look at the answer. This usually has more learning value than just receiving the answer.
7. Making mistakes isn't a failure of the learning system; it's information produced by the learning system.
8. Continue explaining why things happen, what the differences are, and give new examples for concepts.
9. Don't judge long-term mastery by how smooth practice feels.
10. The ultimate goal isn't just memorization, but correctly transferring knowledge in unfamiliar situations.

Conclusion: Learning is a continuous process of reconstructing knowledge.

Ordinary views of learning often see it as simply putting knowledge into the brain. A more complete process should be:

Build knowledge → Retrieve knowledge → Forget some → Retrieve again → Correct errors → Reorganize → Apply in different contexts

Learning isn't just storage; it's closer to construction + retrieval + reconstruction. Some people read a lot of material but have little knowledge they can readily access; others encounter less material but build solid, well-organized knowledge. The difference is often not the absolute amount of time invested, but the cognitive activities the brain engages in during learning.

If any link in the chain is close to zero for a long time, the overall effect will drop noticeably. Learning smoothly doesn't necessarily mean you're learning well; feeling temporarily challenged doesn't necessarily mean you're learning poorly. The real test is whether you can still pull the knowledge out a week, a month, or even a year later, and know when to use it.