한국어

Does the Feynman Technique Work? What the Research Says

At a glance
  • No large study we could find has tested the Feynman Technique by name against other ways of studying.
  • Its core part, explaining to someone else, has helped memory and understanding in many experiments.
  • Students who taught without notes did better on a test a week later than students who taught from a script.
  • Their gain was about the same as students who only practiced recalling. The work is done by retrieval.
  • So don't ask AI to explain it simply. Explain it yourself with everything closed, then ask AI for the gaps.

Yes, the Feynman Technique works, but the benefit does not come from using simple words. It comes from explaining with your notes closed. Students who taught without notes understood more a week later than students who taught from a script, and about as much as students who only practiced recalling. That is why asking AI to "explain it simply" skips the step that does the work.

What is the Feynman Technique?

The Feynman Technique is a way to study by explaining a topic in plain words, as if to someone who knows nothing about it, and then going back to fill the gaps you hit. It carries the name of the physicist Richard Feynman.

It is usually taught in four steps.

StepWhat you doClosest research idea
1Pick one concept
2Explain it in plain words, as if teachingLearning by teaching, retrieval
3Find where you got stuck and go backCorrecting the feeling of knowing
4Make it shorter and simplerElaboration

This note takes those steps apart and looks at what research found for each one.

Is there research on the Feynman Technique?

We found no large experiment that tests the Feynman Technique by name against other ways of studying. The top search results describe the steps or cite small classroom studies.

Its central step, explaining to someone else, has a long research record. Kobayashi (2019) pooled earlier studies in a meta-analysis. The comparison group is students who studied with no plan to teach.

Students learned most when they prepared to teach and then taught face to face

  • Indirect (video, writing)
  • Direct (face to face)
00.250.50.751Indirect (video, writing) · Prepared to teach only · 0.270.27Direct (face to face) · Prepared to teach only · 0.50.5Prepared toteach onlyIndirect (video, writing) · Prepared and then taught · 0.480.48Direct (face to face) · Prepared and then taught · 0.840.84Prepared andthen taughtPreparing alonealready helped
Effect size (Hedges' g) against students who studied with no plan to teach. 0.2 is usually read as small, 0.5 as medium, 0.8 as large. Indirect means teaching by video or in writing; direct means teaching a person face to face.Source: Kobayashi (2019), Table 2
See the numbers
Indirect (video, writing)Direct (face to face)
Prepared to teach only0.270.5
Prepared and then taught0.480.84

Two things stand out. First, just studying as if you will teach helps. In Nestojko and colleagues' experiments (2014), students told they would teach the passage later recalled it more completely and in a better organized way, and generally answered more questions about it, especially about the main points, than students told they would take a test. In reality nobody taught.

Second, the effect grew when there was a real listener. Preparing to teach someone face to face (g = 0.50) beat preparing to teach by video or in writing (g = 0.27). Only four comparisons involved actual face-to-face teaching, though, so read the 0.84 as a direction, not a precise number.

Does it work if you explain with your notes open?

No. The benefit appeared when people explained from memory. This is the part of the Feynman Technique that is easiest to miss.

Koh and colleagues (2018) had students study the same text and then split them into four groups:

  • made a teaching video without notes (explaining while recalling)
  • made a teaching video by reading a script they were given (explaining without recalling)
  • did not teach, only practiced recalling (a recall test)
  • solved arithmetic problems (neither)

On a comprehension test a week later, the students who taught without notes and the students who only recalled both beat the script readers and the arithmetic group. The authors concluded that the benefit of teaching likely comes from retrieval practice.

Explaining without notes and recalling alone did about equally well

  • Ahead of script readers
  • Ahead of arithmetic group
00.250.50.751Ahead of script readers · Explained without notes · 0.570.57Ahead of arithmetic group · Explained without notes · 0.820.82Explained without notesAhead of script readers · Recalled only · 0.650.65Ahead of arithmetic group · Recalled only · 0.910.91Recalled onlyReading a scriptloses this gain
Effect size (Cohen's d) on a comprehension test one week later. Each bar shows how far that group was ahead of the comparison group. The two leading groups differed by d = 0.10.Source: Koh et al. (2018), effect sizes as reported in Kobayashi (2022)
See the numbers
Ahead of script readersAhead of arithmetic group
Explained without notes0.570.82
Recalled only0.650.91

Saying the same words from a script stuck less. The effort of pulling it out of your own head is what builds the memory.

Not everyone reads the result the same way. Kobayashi (2022), citing Lachner and colleagues (2020), points out that the script readers may have lost more than retrieval: they also did not build an explanation in their own words. In some other studies, explaining to others did better than recalling alone. The practical point for you is the same either way. Explaining with your notes open throws away much of the benefit. That recalling beats rereading over days is also covered in our note on spaced repetition schedules.

Why does explaining show you what you don't know?

Because until you try to explain something, you feel you understand it more deeply than you do. Rozenblit and Keil (2002) called this the illusion of explanatory depth.

In a series of studies, people felt they understood complex things with far more precision and depth than they really did. The illusion was much stronger for knowledge of how things work than for facts, procedures or stories.

In study, it looks like this. While you listen to a lecture or read an AI's explanation, everything seems clear. Then you close it, try to say the first sentence, and stall. Step 3 of the Feynman Technique, finding where you got stuck, is the tool that breaks the illusion. Roediger and Karpicke (2006) saw the same pattern: students who reread felt more confident they would remember, but on tests days later the students who had practiced recalling remembered more.

What happens when you do the Feynman Technique with AI?

When you ask AI to "explain it simply", the AI becomes the one explaining. It then does the part that works, pulling the idea out of memory into your own words, for you.

Searches for Feynman Technique prompts mostly turn up prompts where AI does the simplifying. A clear explanation can help you understand, but reading it is closer to the script group in Koh et al. (2018) than to the teachers. Flip the order:

  1. Close your material and the AI chat, then explain one concept out loud or in writing for 3 to 5 minutes.
  2. Mark the places where you got stuck or skipped ahead.
  3. Then open the AI, paste in your explanation, and ask it to point out only what is wrong or missing.
  4. Reread only those parts, and explain the same concept again tomorrow with everything closed.

It is also worth asking the AI to play a student who asks follow-up questions. Asking and answering questions is one of the proposed reasons face-to-face teaching helped more. We have not seen a study that tests explaining to an AI student, though. Learning through questions is covered in our note on whether asking questions helps you learn.

A version you can try today

All you need is a blank page and five minutes.

  • Pick one thing you learned today and write it at the top as a question (for example, "Why does spacing out review make memories last?").
  • Close the book and the screen, and answer it as if a friend who has never heard of it asked.
  • Put a question mark on each line where you stalled. Look up only those.
  • Tomorrow, answer the same question again with everything closed. Check whether yesterday's question marks come back.

A real listener makes it better. If you don't have one, you can start by studying as if you will explain it to someone tomorrow.

In Aloom

Aloom turns this order into a daily habit. You ask AI a question, keep only the answers worth keeping, and each one joins what you already know in a knowledge graph. Then each day you recall the number you chose, 1, 3 or 5, as questions before you see the answers. Pulling it out in your own words first and opening the answer second is the core of this note. The more you recall a node, the darker its ink, so you can tell what you have read from what you can bring back. See how Aloom works

FAQ

Is the Feynman Technique effective?

Its core step is. Explaining to others has helped learning across many experiments, and the benefit was clearest when people explained from memory rather than from notes (Koh et al. 2018; Kobayashi 2019). The technique as a whole has not been tested in a large trial.

Does the Feynman Technique work for math?

The research here mostly used texts and concepts, not problem solving, so we can't say from it how well it works for math. Explaining why a method works, with your notes closed, uses the same retrieval step. Practicing problems is still needed.

What is a good Feynman Technique prompt for AI?

Write your own explanation first, then ask the AI to point out only what is wrong or missing. Or ask it to play a student who knows nothing and asks follow-up questions. Either way, recalling with everything closed comes first.

References