This is a how-to for building a mistake notebook as a "retest device" rather than a fair copy. It is written for exam students whose wrong answers pile up across problem sets, past exams, mock exams, and class handouts. In seven steps, it covers which mistakes to record, how to write them, when to re-solve them, when to retire them, and how much time to spend. The claim is simple: the prettier you make your mistake notebook, the less you improve.
"Copy the problems you got wrong into a notebook, neatly." It is a staple of study advice, and you have almost certainly heard it. Headings in colored pen, the model answer carefully transcribed, the key points underlined. Finishing a notebook like that feels satisfying. But the prettiness itself is not the harm. The harm is that the time spent making it pretty contains not a single attempt to remember, and trying to remember is what strengthens memory.
The reason is simple. Memory consolidates most strongly when you try to recall something, not when you summarize it. Rewriting, transcribing, and rereading keep your hands busy and use up time, so they feel like studying. Yet research in educational psychology has repeatedly reported that this kind of work sits among the less efficient methods for retention. Two methods are reported to work: retrieval practice, which means recalling the answer without looking, and spaced repetition, which means trying again right around the time you start to forget.
So we replace the design of the mistake notebook from the ground up. Spend as little time as possible making it, and re-solve from it as many times as possible. Each of the seven steps below comes with one line on why it works.
How is this different from cycling a problem set or a next-day retest?
Cycling a problem set handles wrong answers inside one book. This notebook collects wrong answers from many materials into one place, and the next-day retest becomes one of its parts.
I previously wrote about how to cycle a problem set: go rough on the first pass, and from the second pass on, re-solve only what you got wrong. That column stays inside a single problem set. This one handles the outside. It collects the wrong answers that come up across problem sets, past exams, mock exams, and class handouts into one device and turns them into an asset. It also serves as the final catch basin for questions that survive your second and third passes.
In the column on restarting a stalled study routine with systems, I described putting only yesterday's wrong answers in front of you again the next day. That is the same motion as the first retest in Step 4 below. There, it is the smallest first step for getting moving again. Here, it is one part of a device you keep running until exam day.
This column also shares its spirit with redoing your self-assessment by counting records instead of trusting memory: deal with yourself based on records, not feelings. The target differs, though. That one is about self-assessment. This one is about managing wrong answers.
Why doesn't "summarizing it neatly" work?
Because rereading, transcribing, and summarizing rate poorly for retention in the research, while feeling highly productive. That gap creates an illusion of understanding.
In 2013, a large review paper comparing the effectiveness of major learning techniques was published (Dunlosky et al.). Methods that revisit the material, such as rereading, highlighting, and summarizing, received low ratings across the board. Only two earned the top rating: testing yourself (retrieval practice) and spacing your study out over time (distributed practice).
There is also a famous experiment (Roediger and Karpicke, 2006). One group repeatedly reread a passage; another group was repeatedly tested on it. Immediately afterward, the rereaders scored higher. One week later, the tested group had clearly pulled ahead.
What makes rewriting dangerous is not only its low efficiency. The felt payoff is large. Look at a beautifully written-up notebook and a sense of "I understand this" arises. Psychology has a name for a closely related phenomenon: the "illusion of fluency," which means mistaking the smoothness of processing (it reads easily, it looks familiar) for understanding or memory. Being able to read something does not guarantee being able to recall it. And in the exam room, recall is the only thing you are ever asked for.
The seven steps below all exist to embed those two principles, retrieval practice and spaced repetition, into the physical device that is your notebook.
Step 1: Which mistakes go in? Diagnose the cause in one line, from three types
Only mistakes whose cause you can diagnose in one line go in. Careless errors stay out.
Do not record every mistake. First, diagnose the cause of each wrong answer in one line, as one of these three:
- Knowledge gap — you did not know it, or had not memorized it
- Misreading — you misread the problem statement or its conditions
- Procedural slip — you knew the method, but it broke down partway through
Only mistakes carrying this diagnosis, meaning reproducible mistakes, go into the notebook. Pure careless errors, such as a slip of the pen or a copying mistake, stay out. A careless error is not a gap in knowledge, so re-solving is unlikely to fix it. You prevent that kind of mistake with mechanisms instead, such as physically checking your answer-sheet rows and fixing a priority order for your final review. I covered those in the column on exam-day time allocation.
→ Why it works: putting the cause into one line changes the work from "memorize the right answer" to "pinpoint how I go wrong," which is what lets you dodge the same trap next time. Narrowing the entry criteria also turns the notebook into a device holding only questions worth re-solving, which raises the value of every retest.
Step 2: How do you write it? Problem on the left, "how I got it wrong" in one line on the right, never the model answer
Left: the problem or its location. Right: one line on how you got it wrong. Nothing else, and never the model answer or the explanation.
This is all the format you need.
- Left side: the problem itself. Write it out if it is short. If it is long, write only the number and location, such as "workbook p. 34, problem 3." Pasting in a photocopy is fine too.
- Right side: one line on what you got wrong and how. Use the three types from Step 1 plus one concrete note (for example, "Procedural slip. Dropped the sign when moving the term").
- What you never write: the model answer or the explanation. No transcribing.
Leaving out the model answer is not laziness. It is the heart of this device. If the answer is written in the notebook, your eyes catch it when you re-solve. "Seeing" happens before "recalling," and the retrieval practice collapses. The answer and the explanation already exist in the original problem set or the past-exam solutions, so there is no need to hold them twice. As long as the location is written down, you can go back there to check your work.
→ Why it works: what strengthens memory is not seeing but the effort of trying to recall, and physically keeping the answer out of the notebook forces retrieval practice every time you open it.
Step 3: How do you use it? Never reread. Open it only to re-solve with the answers hidden
The notebook has exactly one use: retesting. Open it, solve by hand, then check against the original material.
Fix this notebook's use to one thing: retesting. Forbid yourself from flipping through it or gazing at it for reassurance. When you open it, look at the problem on the left and actually solve it, in the margin or on scrap paper. Do not settle for "ah, this one, I've got this" in your head. Move your hand, produce an answer, and then check it against the explanation in the original material.
→ Why it works: there is a wide gulf between "I understand it when I see it" and "I can produce it on my own"; rereading only verifies the former, while re-solving directly trains the latter, which is the only thing exams ask for.
Step 4: When do you re-solve? The next day, 3 days later, 1 week later, and right before the exam
Fix the schedule in advance instead of leaving it to your mood. Protect the rhythm, not the exact numbers.
- The next day — first retest
- Three days later — second
- One week later — third
- Right before the exam — final check
Let me be honest here. There is no rigorous basis for these exact numbers. What the research shows is the principle that repetitions spaced further apart make memory last longer. The optimal number of days shifts with the material and with how far away your exam is. So treat these numbers as a practical template for turning the principle into a daily routine. Drifting a little is fine. What you protect is the rhythm: once more, right when you are starting to forget.
→ Why it works: effortfully dragging a memory back from the edge of fading is said to consolidate it most strongly (the idea known as "desirable difficulties"), while re-solving something you still fully remember is easy and yields correspondingly little.
Step 5: When does a problem graduate? After two consecutive correct answers on separate days
Add a check column to each problem. Once it is answered correctly twice in a row on retests separated by days, strike it through.
Make a small check column at the edge of each problem and mark it right or wrong every time you re-solve. A problem answered correctly twice in a row, on retests separated by days, gets a horizontal line through it. It has graduated. Do not re-solve it again until the final pre-exam check.
Both conditions, "in a row" and "days apart," matter. Solving something twice on the same day may only mean the memory was still fresh. If you can produce it twice with days in between, you can judge it as retained. And as with the schedule in Step 4, this "twice" is not a threshold from the research. Once cannot be told apart from recent memory; three times makes the cycle too heavy to sustain. Twice sits between them as a design value for daily use.
→ Why it works: without a graduation rule the notebook keeps swelling until you can no longer cycle through it, whereas striking problems out keeps your time on the ones you still cannot solve, and the growing number of struck lines makes your progress visible as the notebook gets "thinner."
This notebook is finished not when it is written, but when every line has been struck out.
Step 6: How much time should you spend making it? Cap it at 3 minutes per problem and 10 minutes per day
Three minutes per problem, ten minutes per day. Past the cap, write only the location, or re-select what goes in.
Put a physical ceiling on the making. As a guide: 3 minutes per problem, 10 minutes total per day. If you cannot write an entry in 3 minutes, stop copying the problem and write only its location. If you are about to exceed 10 minutes, re-select what deserves to go in. Care beyond the cap has no value in this device.
→ Why it works: the failure I consider most dangerous is that making the notebook becomes the goal and studying ends with the satisfaction of having made it, and a time cap seals off that trap with physics rather than willpower.
Step 7: Paper or digital? Choose whichever has less friction before you start re-solving
One criterion only: which one gets you from "I should re-solve" to actually solving with less friction. Splitting by subject is fine.
A paper notebook, or photos plus an app? Plenty of people agonize here, but you need only one criterion: which one, for you, has less friction between "I should re-solve" and actually starting? Photographing problems into a flashcard-style app is strong at collecting across multiple materials and automating the spacing. For subjects like math, where you write your answer out by hand, opening a paper notebook and solving in the margin right away is often faster. Splitting by subject is fine too.
Here is one concrete example from our own shelf. passed.jp's English problem set is built so that, right after you finish a set, you check your answers against the correct ones in a list and go back to the explanations of only the questions you missed. You can re-solve the same grade's or the same unit's questions as many times as you like. For English drilling, in other words, the site already plays something close to the role of a digital retest device. The paper device in this column takes on everything outside that: math and science answers written by hand, and the wrong answers scattered across multiple materials.
→ Why it works: I believe the results this device produces are determined almost entirely by the number of re-solves, so the choice that reduces the friction to start is the choice that maximizes that count.
What not to do
Most of the prohibitions are the flip side of the steps above.
- Decorating with colored pens (two colors at most, if any)
- Transcribing model answers (Step 2)
- Opening it just to reread (Step 3)
- Recording every careless error (Step 1)
Folded into one line: never spend this device's time on work that produces zero attempts to remember. That is all.
Beyond those, two failures remain that have not been named yet. Both are tempting, so let's name them in advance and seal them off.
- Don't wait to start until it is perfectly built. Starting with today's one mistake is fine. This device grows while being used.
- Don't rewrite a page you have already made. A messy page still works. The thirty minutes a rewrite costs would buy you several re-solves.
Closing: the thinner and messier the notebook gets, the closer you are to passing
A beautiful notebook is something you make to show other people. A mistake notebook is shown to no one, so its polish earns you not a single point. Only two metrics measure this device: the quality of your selection of what goes in, and the number of times you re-solved from it.
This notebook has one more fine property. Right before the exam, when little time remains, it has become the only textbook in the world made entirely of the places where you, specifically, go wrong. Store-bought materials do not know your weaknesses. This notebook contains nothing else. The value of a single hour in the final stretch transforms.
The meticulousness it takes to keep a beautiful notebook is, I think, a genuine talent in itself. That is exactly why I would ask you to point it not at fair copies, but at selection and spacing. As the struck-through lines multiply and the notebook grows thinner and messier, you are getting closer to the "passed" that passed.jp hangs on its signboard.
FAQ
Q. Do I really leave the model answer and the explanation out of the notebook? A. Yes. If the answer is in the notebook, your eyes catch it when you re-solve, and seeing happens before recalling. The answer and the explanation already exist in the original problem set or past-exam solutions, so write only the location, such as "workbook p. 34, problem 3," and go back there to check.
Q. Should careless errors go into the mistake notebook? A. No. A slip of the pen or a copying mistake is not a gap in knowledge, so re-solving is unlikely to fix it. Prevent those with mechanisms instead, such as physically checking your answer-sheet rows and a fixed priority order for your final review, as described in the column on exam-day time allocation.
Q. Do I have to follow the "next day, 3 days, 1 week" schedule exactly? A. No. Drifting a little is fine. The exact numbers have no rigorous basis; what the research shows is that repetitions spaced further apart make memory last longer. The optimal number of days shifts with the material and with how far away your exam is. What you protect is the rhythm of trying again right when you start to forget.
Q. When can I strike a problem out? A. When it has been answered correctly twice in a row on retests separated by days. Two correct answers on the same day do not count, because they may only reflect fresh memory. A graduated problem is not re-solved again until the final check right before the exam.
Q. Should I build it on paper or in an app? A. Whichever has less friction, for you, before you start re-solving. Photographing problems into a flashcard-style app is strong at collecting across multiple materials and automating the spacing. For subjects like math, where you write answers by hand, a paper notebook you can solve in right away is often faster. Splitting by subject is fine.
