Using the Feynman technique when you’re studying alone takes five moves:
- Close the source and explain one mechanism, not a whole chapter.
- Mark every stall, borrowed phrase and missing “because.” Those are the gaps the technique catches well.
- Put the explanation against something outside your own judgment — the source, a worked solution, a knowledgeable person or a source-grounded tool.
- Use the idea on a fresh problem. A good explanation and a usable idea are not the same verdict.
- Explain or apply it again later. A polished answer five minutes after reading says little about what survives next week.
You do not need a human audience for step 2. You do need an answer key for step 3. The empty chair is a useful prompt and a nonexistent judge.
How do you do the Feynman technique when studying alone?
The popular protocol is simple: study a concept, teach it in plain language, return to the parts you could not explain, then simplify again. Kansas State’s one-page version says you can teach a friend, roommate or family member — or talk aloud to yourself.
That substitution mostly works. Explaining to yourself still forces you to produce the idea instead of nodding along with the author’s version. You can hear the sentence that never reaches a verb. You notice when “it just balances out” is covering for a mechanism you cannot name. You discover that the analogy you loved does not map back to the thing it was meant to explain.
Use a voice note, a blank page or an empty chair. The format barely matters. The constraints do:
- close the book before you begin;
- choose one causal claim, argument or worked step;
- ban any technical term you cannot define in the same breath;
- make at least one prediction or example; and
- stop after five minutes and mark the exact breakpoints.
Do not turn it into a lecture. A forty-minute monologue gives you forty minutes to talk around the hole. Three precise minutes on one idea makes evasion much harder.
What does a solo explanation fail to catch?
It fails to catch the explanation that is clear, complete and wrong.
If you go blank, you know you have a gap. If you confidently reverse cause and effect, omit a condition you never noticed, or use a beautiful analogy that breaks at the important edge case, the performance can feel excellent. The learner who produced the error is also the learner deciding whether it passed.
This is not an argument against self-explanation. It is an argument against asking one exercise for four verdicts.
| What you need to know | What reveals it |
|---|---|
| Can I produce the idea without the book? | Explain it from memory |
| Is my explanation accurate? | Compare it with the source or get outside feedback |
| Can I use it rather than recite it? | Solve a fresh problem or apply it to a new case |
| Will it still be there later? | Retrieve it again after a delay |
The ordinary Feynman loop is good at the first row and gives useful clues about the second. It does not settle the other three.
That is the same structural problem behind testing yourself on a textbook you are studying alone: you can write the question, sit the test and mark the answer, but every stage can inherit the same original misreading. A smooth result is still circular.
Does the Feynman technique actually work?
The evidence supports a more modest answer than most study blogs give.
The named four-step protocol is built from real learning activities, especially self-explanation and retrieval. But the large 2013 review by Dunlosky and colleagues rated self-explanation moderate utility, not high utility: promising across some settings, but not adequately tested in educational contexts. Practice testing and distributed practice received the high-utility ratings.
A preregistered 2026 experiment makes the distinction sharper. Harders and Ebersbach assigned 208 participants either to self-explain or reread fictional factual material, held study time constant, and tested them immediately and two weeks later. Self-explanation did not beat rereading on factual memory. Within the explanation group, correct inferences predicted performance; paraphrases did not.
That study does not prove that explaining is useless. It tested factual material, not every kind of conceptual learning. It does show why “say it in your own words” is too weak a recipe. A paraphrase can be active, effortful and educationally inert. The quality and correctness of what you generate matter.
Delayed retrieval has sturdier support. In Roediger and Karpicke’s 2006 experiments, restudying won on a test five minutes later, while prior testing produced greater retention after two days and one week. Restudying also raised confidence. That is exactly why your solo loop needs a later attempt: immediate fluency is an unreliable finish line.
The honest conclusion is not “Feynman works” or “Feynman does not work.” It is: explaining is a strong diagnostic task when you make the output checkable, and it is one part of learning rather than the whole system.
How do you check your explanation without a teacher?
Run the cheapest independent check your subject allows.
1. Check each claim against the source
Underline the claims in your explanation that could be false. For each one, find the passage, definition, proof step or worked example that supports it. If you cannot locate the support, you have either made an inference or invented one. Label it before deciding which.
Do not compare prose for identical wording. Write the explanation as X because Y, under condition Z. Then check X, Y and Z separately. This catches more than asking whether your paragraph “basically matches.”
2. Make the idea move
Change one condition and predict what follows. Apply the argument to a case the author did not use. Solve a problem whose surface details differ from the worked example.
This is especially important in technical subjects. A correct explanation of a method does not prove you can select it when the problem stops naming the method for you. Our guide to self-studying physics from a textbook uses a simple version: write why the principle applies, predict what changes when a quantity doubles, and only then check the number.
3. Bring it back after the glow has worn off
Put the explanation away. Return after a day, then after a longer gap. Rebuild it before reading the old version. If the concept matters for months, test it on that horizon rather than congratulating yourself after five minutes.
There is no universal calendar. The useful gap depends on how long you need the idea and how the last retrieval went; our forgetting-curve guide covers that decision without pretending one interval fits every book.
4. Spend a person on disagreement, not performance
If you have access to someone who knows the subject, do not make them sit through your whole lecture. Give them the shortest claim you could not reconcile with the source, your reason, and the passage you used. A focused disagreement is easier to answer and more useful than “Can I explain chapter six to you?”
Which AI tools can be your Feynman partner?
As of August 2026, this is a real product category. The old advice to use a generic chatbot is no longer the whole answer.
- InkRecall accepts a PDF, image or text, compares your explanation with the source, and challenges you with scenarios.
- Ankra accepts notes, PDFs, slides, photos or a recording, marks what was right, wrong and missed, then turns gaps into spaced-repetition cards.
- Revaldo evaluates explanations against uploaded material and also generates and marks short-answer and essay questions.
- Axiom Flow takes a different route: you teach an AI student, then the student’s follow-up test becomes the check on what you managed to teach.
If all you want is a dedicated Feynman session, use one of those. They name the job directly and are built around it. None should be treated as an oracle: an AI verdict is another reading of your work, and a disagreement with the book is a reason to inspect the book, not overrule it.
A small 2025 preprint tested the broader idea directly. Rajesh and Khan built a retrieval-augmented Feynman Bot for learners without peer or instructor support and reported higher learning gains than passive study. The controlled experiment lasted three days and involved 14 participants. Promising, yes. Settled evidence for AI as a teacher, no.
How Study Junkie fits the solo Feynman loop
Study Junkie does not have a button called “Feynman mode.” It does the wider job across the book you chose.
You upload a PDF, DOCX or text file and it becomes a chaptered course built from that source. The useful Feynman sequence then sits inside a larger loop:
- Explain and discuss: the tutor is grounded in the uploaded material, so you can press on the exact definition or passage your explanation depends on.
- Get a verdict: assignments generated from the chapter make you submit written work, and the feedback responds to what you wrote.
- Make the idea move: the assignment is a new question rather than a request to repeat the page.
- Return later: scheduled recall brings completed material back after the immediate fluency has faded.
The concrete proof is the written verdict, not the feature list. In a 39-second recorded run on Kant’s Critique of Pure Reason, you can watch the course read an answer and identify the specific misconception, then schedule the material for one-, three- and seven-day recall.
Honest limits: a knowledgeable human is better than an AI reader. A fully worked solution is authoritative for the problem it covers. And if you only want to practise one explanation, the dedicated tools above are the cleaner choice. Study Junkie makes sense when the explanation belongs to a whole book you want to read, practise and remember rather than to a one-off concept.
You can see how the full book-to-course loop works or start with 1,500 free Study Credits, with no card required.
FAQ
Do I need another person for the Feynman technique?
No, not for the explain-it-back step. Speaking to yourself, writing a short lesson or recording a voice note still exposes stalls and borrowed language. You need an outside check only for the claims that sounded complete: a source, worked solution, fresh problem, knowledgeable person or source-grounded tool.
Who do I explain what I learned to when studying alone?
Explain it to an imaginary beginner, but write or record the explanation so it becomes inspectable. The audience creates the plain-language constraint; it does not supply the verdict. Afterward, check the claims against the source and apply one to a new case.
How do I use the Feynman technique for reading a book?
Do it after a small section, not after the whole book. Close the book, explain one mechanism or argument in three minutes, mark the gaps, verify the explanation against the relevant passage, then answer a question the passage did not answer verbatim. Repeat later. For a broader route from approachable reading into technical work, use our guide from popular science to real science.
Is the Feynman technique the same as active recall?
It includes recall because you produce the explanation without looking. It adds organization, causal links, examples and simplification. Ordinary recall can ask for one fact; a Feynman explanation asks you to rebuild a connected account. Neither guarantees the account is correct, which is why feedback remains a separate step.
How do I know if I actually understand something?
Require four different pieces of evidence: you can explain it without the source, the explanation survives an external check, you can use it on a fresh case, and you can retrieve it after a meaningful delay. Passing one is encouraging. Passing all four is evidence.