Testing yourself on a textbook you’re studying alone comes down to five steps:
- Read one chunk, not one chapter. UNC’s Learning Center suggests no more than 5–10 pages at a time; the Learning Scientists work in chunks of “a few paragraphs at most.”
- Close the book and write down what it said. From memory, before you look.
- Write questions on the chunk — the ones you’d set if you were teaching it.
- Answer them cold, days later, in prose or worked steps, not in your head.
- Get the answers checked by something that is not you.
Steps 1 through 4 are on every page that ranks for this question, and they work. Step 5 is where every one of those pages quietly hands the job back to you — check your answers against the answer sheet you wrote, or go see your professor. If you’re teaching yourself, you have neither.
This article is about step 5, because it’s the only one that can tell you you’re wrong.
Does testing yourself on a textbook actually work?
Yes — including the version where you write the questions yourself.
In a 2020 study published in Applied Cognitive Psychology, Ebersbach, Feierabend and Barzagar Nazari put 82 university students into three groups: one restudied the material, one took a test on it, and one wrote their own questions about it. On a final test a week later, “students in both experimental conditions (i.e., generating questions and testing) scored on average 11 percentage points higher on the final test compared with students in the restudy condition.”
Writing your own questions was as good as being tested. Rereading lost to both. If you take one thing from the standard advice, take that.
Now the part that rarely gets quoted from the same paper. The authors noted that the questions students wrote “mostly addressed superficial facts rather than requiring conceptual analyses or inferences (i.e., low question depth in general)” — and that question depth showed no correlation with the final score. They also list, among the study’s limitations, that the questions were “not evaluated by the instructor.”
So: generating your own questions is a memory intervention. It reliably makes material stick. It is not a verification method, and nothing in that result says the material you fixed in memory was correct. Those are two different jobs, and almost everything written about self-testing treats them as one.
Why checking your own answer sheet doesn’t work
The best method on this search is the Learning Scientists’ “Be Your Own Teacher: How to Study a Textbook” (Rachel Adragna, 12 February 2016). It’s a genuinely good nine-step procedure: take a small chunk, note the key points, form questions that elicit them, write the answers on a separate sheet, repeat, then shuffle the pages and answer at random days later.
Then step eight:
“After each set of questions, go back and check your answers with the answer sheet. If you got something wrong, review the notes or textbook if you need clarification.”
The answer sheet is the one you wrote. You read the chapter, you decided which points mattered, you wrote the question, and you wrote the answer — all from the same reading. If you misread the chapter, the misreading is in the question and in the key. You sit the exam, you mark it, and it comes back clean.
This isn’t a flaw in that method specifically. It’s the shape of every method written for someone who has a class. UNC’s Learning Center also has you author the test — “create questions that you would ask on a test if you were a professor giving a test about this chapter” — and where it does supply a real check, the check is a person: “Office Hours: Schedule an appointment with your professor.” It also assumes you’ll “come to class prepared.”
The advice isn’t wrong. It’s just missing a component you’re expected to already have.
At the other end of the results, practitioner accounts of self-studying technical books often skip the step entirely. Statistically Relevant’s four strategies (Joleen, 8 November 2021) — annotate, don’t over-note, implement fast, iterate — is a sincere account of the process and contains no discussion of exercises, answer keys, or how you’d know you were right. When the book stops making sense, the move is to go read a different explanation of it. Which helps with confusion, and does nothing about confidence.
Search for the honest version of the question — how do you do textbook exercises without an answer key or a teacher — and the results are worse than unhelpful. As of August 2026, most of the first page is a hijacked domain serving fake “solution manual” PDFs. The rest tells you to ask your instructor, your tutor, or your classmates. That is the question, returned as the answer.
You can be your own examiner. You cannot be your own answer key.
How do you check your answers with no answer key and no teacher?
You need at least one source of correction that didn’t come out of your own reading. There are four, and it’s worth knowing exactly what each one can and can’t catch.
| Check | What it catches | What it misses |
|---|---|---|
| The book’s own worked solutions | Whether your final answer matches | Why yours went wrong, and everything the exercises don’t cover |
| A published solutions manual | The same, for more problems | Same — plus most books don’t have one, and the search for one is a spam minefield |
| Checking your prose against the text | Claims you can’t find support for | Anything you misread the same way twice |
| Another person | All of it, including reasoning | Availability. This is the one you don’t have |
Use them in that order, and use them properly:
- Work the exercises before you read the solution, and write the whole solution down. A worked answer you followed along with is not a test. If the book gives final answers only, the number matching is weak evidence — you can reach a right answer by a wrong route, and in most subjects that’s the more common failure.
- Make your prose checkable. Instead of “explain the concept,” write a claim about the chapter that could be false, then go find the sentences that support it. If you can’t find them, you invented it. This works on books with no exercises at all, which is most non-technical ones.
- Date your errors. Keep one running list of every place you were wrong and what the actual answer was. It’s the only artifact of self-study that gets more valuable over time, and it turns your confirmed mistakes into the question bank you re-test yourself on — which is retrieval practice aimed at exactly the places you already know you were wrong, rather than at the chapter as a whole.
- Spend a real person carefully. A study partner, a subject forum, a friend who did the degree — all of them beat every automated option in this article. They also run out. Bring them your error list and your specific wrong reasoning, not “can someone explain chapter 4.”
What about AI quiz generators?
There’s a whole category now — Quizlet, Quizgecko, Revisely, StudyPDF and others. Upload a PDF, get questions back in seconds. For step 3 — generating the questions — they genuinely save time, and per the research above that step has real value.
But look at what they do with step 5. Their question types are multiple choice, true/false, matching and fill-in-the-blank: formats that can be graded precisely because they don’t require reading a sentence. And where they do offer free-text, StudyPDF states the arrangement plainly on its own page: “Short-answer questions come with a model answer that you self-grade.” “Write your answer. You will self-grade it against the model answer at the end.”
That’s the Learning Scientists’ circularity with a faster question-writing step. The questions now come from the book instead of from your reading, which is a real improvement. The verdict still comes from you.
What do you do when the textbook has no exercises?
Most books outside math, physics and engineering have none, and this is where self-testing usually collapses into re-reading.
The substitute is to stop asking yourself to recall and start asking yourself to commit. Write the argument, not the summary:
- Take a position the chapter would judge. “The author claims X, and the argument for it fails at Y, because Z.” A recall question can be answered by a good memory. An argument can be answered wrongly.
- Apply it to a case the book never mentions. Transfer is where a shallow reading breaks, and it breaks visibly, in your own writing.
- Reconstruct the chapter’s structure from memory before re-opening it — what it was trying to establish, in what order, and what it needed to prove first.
Each of these produces a page of prose that is either defensible against the text or isn’t. That’s a testable object. “I understood the chapter” isn’t.
How Study Junkie closes the checking step
Study Junkie exists because of step 5. You upload the textbook — PDF, DOCX or a text file — and it becomes a structured course with generated chapters, usually in two or three minutes even for a five-hundred-page book. From there the loop is the one above, with the last step done by something outside your head:
- Consume — read the chapter, or listen to it narrated at 0.75× to 2× while you’re walking.
- Discuss — ask a tutor that’s grounded in your uploaded source rather than in general internet knowledge, so the answers come back from the book you’re actually reading.
- Practice — assignments generated from your material, which you answer in writing, and which come back with feedback on what you wrote.
- Recall — scheduled spaced-recall quizzes that bring the material back days and weeks later, without you building a card deck.
The part that matters here is the third one, and it’s worth being concrete about what “feedback” means, because the word is doing a lot of work in this category. We recorded three short demos on three real books, and each shows a different shape of being wrong:
- On Jim Hefferon’s open linear algebra text, it finds the sign error in a written solution and shows where it happened — a computation.
- On Frank A. Fetter’s The Principles of Economics, it reads the argument and challenges the weak claim in it — prose, not a number.
- On Kant’s Critique of Pure Reason, it reads the answer and names the specific misconception behind it — a concept.
That range is the point. A right-or-wrong verdict you could have got from an answer key. Naming which idea you have crossed with which other idea is the thing you cannot do for yourself, because doing it requires already knowing.
Honest limits. An AI reading your textbook is not a subject expert, and its feedback is a second reading of your work rather than an authority — treat a disagreement between it and the book as something to go check, not something settled. If your book ships fully worked solutions, use them; they’re better than anything here for the problems they cover. And a study partner who knows the subject beats all of it.
You get 1,500 Study Credits when you create an account, no card required — enough to put a real chapter through the loop and see whether the feedback tells you something you didn’t already know. After that it’s one-time credit packs from $10, and they don’t expire. There’s no subscription, because there’s no reason a book you read once a year should bill you monthly. It works across eight languages, and the language you upload in doesn’t have to be the language you study in.
FAQ
Does writing your own practice questions actually work?
Yes, for memory. In the 2020 Ebersbach, Feierabend and Barzagar Nazari study, students who generated their own questions scored 11 percentage points higher on a test a week later than students who restudied the material. But the same authors observed that the questions those students wrote “mostly addressed superficial facts,” and that nobody evaluated them. Write your own questions to remember the material; don’t rely on them to tell you whether you understood it.
How do you check your answers when the textbook has no answer key?
In order of preference: find whether the book publishes worked solutions for a subset of problems and use those to calibrate; write claims about the text that can be checked against specific passages; keep a running error list and re-test yourself on it; and spend a knowledgeable person on the things that stay unresolved. If you want an automated check on written work, it has to be something that reads your sentences rather than matching them, which is what Study Junkie’s assignment feedback does. Note that searching for “solutions manual” PDFs is currently a spam minefield.
How much of a textbook should you read before testing yourself?
Less than you think. UNC’s Learning Center recommends tackling “no more than 5-10 pages at a time”; the Learning Scientists work in chunks of a few paragraphs with breaks every 25 to 30 minutes. The practical test is whether you can still close the book and reconstruct what it said. If you can’t, the chunk was too big.
Is rereading the chapter enough?
No. Rereading was the control group in the 2020 study above, and it lost to both self-testing and question generation by 11 percentage points. It feels productive because the text gets easier each pass, which is a measure of familiarity rather than of understanding. If you want the mechanism behind that, we cover it in why you forget what you read.
What if I can’t tell whether my answer is wrong or just worded differently?
That ambiguity is the signal that you need an external reader. A keyword match can’t distinguish “said the same thing differently” from “said a different thing,” which is why self-marked short-answer questions are so weak. Rewrite the answer as a claim with a reason attached — “X, because Y” — and check whether the book supports Y specifically. A wrong reason survives a paraphrase; it doesn’t survive being located in the text.
If you want the subject-specific versions of this problem, the checking gap has a different shape in each one: in math there’s often no answer key at all, in economics the answer keys exist and are excellent but can’t read prose, and in philosophy the misreading happens silently while you read. If you’re working through a whole shelf rather than one book, start with turning a reading list into a study plan. And for the memory half of the loop, spaced repetition without making flashcards covers what to do after you’ve been marked wrong.