Explain it to a 12-year-old, the advice goes, and you will find out whether you understand it. In one study of college students, though, those who explained a text to another student did worse on deeper questions than those who simply explained it aloud to themselves (Roscoe & Chi, 2008, as cited in Fonseca & Chi, 2011).
The technique is also hard to pin down. One study-skills blog gives three steps (Highley, 2022). Another gives five, with a sixth-grader as the audience (Tamm, 2023). An academic paper gives four: write down what you know, explain it as if to a child, look for gaps, then simplify or build analogies (Reyes et al., 2021).
No version is official. The last post covered self-explanation, explaining material to yourself while you study it. The Feynman Technique is its popular cousin, with two changes: you explain after studying, and you aim the explanation at an imagined listener.
Four Activities Under One Name
Taken apart, the technique bundles at least four activities. Generating an explanation is close to self-explanation and to elaboration, linking new material to what you already know. Simplifying can mean either summarizing or explaining, depending on what you actually write. Looking for gaps is what researchers call comprehension monitoring: noticing when you do not understand. And going back to the source is feedback followed by restudy.
The ingredients do not have equal support. Reviewing ten learning techniques, Dunlosky et al. (2013) rated self-explanation and elaborative interrogation as moderately useful and summarization as low. In the comparisons they describe, summarizing was “less powerful than generating explanations” (Bednall & Kehoe, 2011, as cited in Dunlosky et al., 2013, p. 18). The review does not evaluate the Feynman Technique, and support for each ingredient does not show that the combination adds anything.
Timing is the first departure from self-explanation as researchers have studied it. In those studies, learners explain while they work. In a logical-reasoning study the review describes, explaining after solving problems helped on later transfer problems, but less than explaining while solving them (Berry, 1983, as cited in Dunlosky et al., 2013).
Explaining to Someone Can Turn Into Recitation
The audience is the second departure. In the study from the opening, Roscoe and Chi had college students read a text on the human eye and then spend half an hour explaining it. Some explained to a peer who had not read it, some to a video camera for a future student, and some aloud to themselves. On deeper questions about the text, the self-explainers outscored both groups who had explained to an audience.
When the researchers coded what students said, most of the peer explainers’ talk was knowledge-telling, meaning summaries and paraphrases of the text with nothing added. The video explainers did even more of it (Roscoe & Chi, 2008, as cited in Fonseca & Chi, 2011). Knowledge-building, talk that connected ideas and drew inferences, went along with higher scores on the deeper questions (Roscoe & Chi, 2007a, as cited in Chi, 2009). Fonseca and Chi (2011) offer a possible reason, one the study was not designed to test: “the explainers tend to teach only what they already know.”
A live listener did make some difference. In Chi’s (2009) account, explainers with a partner did slightly but consistently better than those who recorded a lesson. The listening partners asked many questions, a good share of them deep, and started more knowledge-building episodes than knowledge-telling ones. Even so, the students who explained to themselves did best. The imagined child of the Feynman Technique never asks anything, and whether it works like the peer or the silent camera is untested. The deeper questions also asked only about the text just read; nobody tested whether students could apply the ideas to new problems.
The One Trial of the Whole Technique, and a Medical Cousin
Among the sources behind this series, only one randomized trial tests the full technique (Reyes et al., 2021). Students in grades 4, 7 and 11 in the Philippines were randomly assigned to a Feynman lesson or the standard lesson on English topics. The Feynman lesson was a 35-minute sequence of reading and note-taking, explaining to oneself, a review, and a final simplification. In all three grades, the Feynman groups scored higher on the test afterward and gained more.
The authors call this “conclusive evidence” (p. 9), a claim the design cannot carry. The samples were small, students were tested once right after the lessons and never again, and the researchers, who knew the students, ran the experiment themselves. The Feynman lesson was also structured differently from the standard one, so any benefit of the technique is tangled up with simply getting a new kind of lesson. The result is encouraging, and it is short-term.
The closest medical evidence concerns teaching rather than the technique. Gregory et al. (2011) worked with 17 third-year medical students who volunteered to teach Advanced Cardiac Life Support to second-years. All prepared two topics, resuscitation algorithms and ECG interpretation, and on the day each was randomly assigned to teach one topic, the other, or neither. Preparing alone raised scores; preparing and then teaching raised them most. Sixty days later both gains were still above where students started, and the extra benefit of teaching had shrunk to the point where it could have been chance. But only 13 of the 17 students came back for that test, and only 2 of the 5 who had not taught, so the study cannot say whether the gap really closed.
Those who taught also spent about four extra hours with the content, and no group simply studied. The content was rule-based, and the authors do not assume the same benefit for complex skills such as diagnostic reasoning. Like every study in this post, it tested the content students had just been taught, with no new problems.
How to Use It Without Fooling Yourself
Learn the material first. The benefits of elaborative interrogation, covered two posts ago, are often larger for learners who already know more (Dunlosky et al., 2013); whether the same holds for self-explanation has not been tested systematically.
When you simplify, explain the mechanism instead of shrinking it. Suppose you “simplify” oxidative phosphorylation by writing cells make energy from food. That is a summary. Suppose you work out why the electron transport chain needs a membrane, what would happen without the proton gradient, and how this connects to ATP synthase. That is explaining. Both feel like progress, and the branded steps do not tell you which one you did.
Give your explanation a listener who asks questions. A study partner who interrupts with “why?” is closer to the listeners in Chi’s (2009) account than an imagined child is. Studying alone, play that partner yourself: after each step of the oxidative phosphorylation explanation, ask why it has to happen that way and what would break if it didn’t.
Check your explanation against the source. In the studies Chi (2000) reviews, wrong self-explanations did no apparent harm, and the likely reason is that the text the students kept reading contradicted most of them. A Feynman explanation comes after you have closed the book, so nothing later in the text will catch your mistakes. Going back to the source has to do that job; skip it, and a confident error stays in place.
Then turn the explanation into retrieval practice. Dunlosky et al. (2013) gave practice testing, their term for retrieval practice, their highest rating. Once the explanation holds up, close the notes and rebuild it from memory a few days later.
For anyone who runs peer teaching, two design choices follow from these studies. Assign the listeners the job of asking questions, since in Chi’s (2009) account the listening partners, who asked many questions, started more knowledge-building episodes than knowledge-telling ones. And treat preparation as part of the learning: in Gregory et al. (2011), students gained on content they only prepared, and those gains were still above baseline at 60 days.
What This Looked Like for Me
I came to medicine as an average standardized-test taker. One habit that helped change that looks a lot like the Feynman Technique.
When a mechanism would not stick, I tried to explain it out loud, in plain language, as if a classmate two years behind me were sitting across the table. Why does this drug raise potassium? What happens to the gradient if this pump fails? A few sentences in, I would hit a spot where my explanation went fuzzy and realize I did not understand the step I had been reciting. That sent me back to the source to fix it. In the first years of medical school, we also presented topics to small groups of classmates with a faculty member watching.
I changed several strategies at once, spacing and self-testing around the same time, so I cannot isolate what the explaining contributed. What explaining aloud did, every time, was show me where I was recognizing a fact rather than understanding it. The subjects I prepared to teach are the ones I still remember best.
The Core Lesson
The Feynman Technique has no fixed definition, and most of its evidence is borrowed from its ingredients. The one trial of the whole technique compared it against a differently structured lesson and tested students only once. Explaining seems to help most when it adds inferences rather than restating. A listener who asks real questions appears to push it that way, and checking your explanation against the source is the step meant to catch errors, though neither has been tested as part of the technique. Then test yourself on it, because a good explanation still fades.
The next post moves from what happens inside a study session to how you structure the time around it: the Pomodoro Technique.
This post is part of a series on the science of learning. Each post covers one evidence-based principle and how to apply it to your own studying. Follow us on Instagram @acemeded to keep up with future blog posts and related content
References
Chi, M. T. H. (2000). Self-explaining expository texts: The dual processes of generating inferences and repairing mental models. In R. Glaser (Ed.), Advances in instructional psychology (Vol. 5, pp. 161–238). Lawrence Erlbaum Associates.
Chi, M. T. H. (2009). Active-constructive-interactive: A conceptual framework for differentiating learning activities. Topics in Cognitive Science, 1(1), 73–105. https://doi.org/10.1111/j.1756-8765.2008.01005.x
Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques: Promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4–58. https://doi.org/10.1177/1529100612453266
Fonseca, B. A., & Chi, M. T. H. (2011). Instruction based on self-explanation. In R. E. Mayer & P. A. Alexander (Eds.), Handbook of research on learning and instruction (pp. 310–335). Routledge. https://doi.org/10.4324/9780203839089.ch15
Gregory, A., Walker, I., McLaughlin, K., & Peets, A. D. (2011). Both preparing to teach and teaching positively impact learning outcomes for peer teachers. Medical Teacher, 33(8), e417–e422. https://doi.org/10.3109/0142159X.2011.586747
Highley, Z. (2022, June 19). The danger of the Feynman Technique (why most people are using it wrong). https://zhighley.com/article/feynman/
Reyes, E. P., Blanco, R. M. F. L., Doroon, D. R. L., Limana, J. L. B., & Torcende, A. M. A. (2021). Feynman technique as a heutagogical learning strategy for independent and remote learning. Recoletos Multidisciplinary Research Journal, 9(2), 1–13. https://doi.org/10.32871/rmrj2109.02.06
Tamm, S. (2023, January 18). Feynman Technique: A complete beginner’s guide. E-student.org. https://e-student.org/feynman-technique/
