The Feynman technique, applied

The Feynman technique is the most-recommended study method on the internet, and one of the least-correctly-used ones. People treat it as a chant: "explain it like I'm five." Then they go back to highlighting. Then they wonder why they don't remember anything a week later.

The trick is that the Feynman technique isn't really an "explain it simply" exercise — that's just the surface action. What it actually does is force you to produce the kind of cognitive work you spent the entire study session avoiding: retrieval, comprehension monitoring, and the moment you realize you don't actually understand something.

If that sounds harsh, good. The technique works because it's harsh. Most studying is comfortable, which is why most studying doesn't work.

The four steps you've already half-read

The canonical version, attributed (loosely) to Richard Feynman:

1. Pick a concept you want to understand. Write its name at the top of a page.
2. Explain it in plain language, as if to someone who knows nothing about the field — usually framed as "explain it to a child," but a smart adult outside your field works just as well.
3. Find the gaps. The places where you stalled, got vague, used a jargon word as a shortcut, or wrote a sentence you couldn't justify if pushed. Those are your real learning targets.
4. Go back to the source, learn the gap, then re-explain. Repeat until the explanation flows without pauses or weasel words.

You've seen these steps before. You probably even tried them. Here's what you missed.

What the four steps are actually doing

The reason this works isn't because explaining-to-a-child is magic. It's because every step forces a specific cognitive act that ordinary reading lets you skip:

Step 2 forces retrieval. When you read a textbook, the information is in front of you — your brain doesn't need to fish it out of memory. When you have to explain from scratch, you're doing active recall. Active recall produces durable memory; passive re-reading produces an illusion of memory that lasts about 12 hours. This is the single largest finding in the cognitive psychology of learning, and it is the reason this technique is non-negotiable.

Step 3 forces comprehension monitoring. Most people can't tell when they don't understand something. They feel a vague hum of "I sort of get this," confuse it for comprehension, and move on. Writing a plain-language explanation makes the hum visible: the moment you write "because of quantum effects" without being able to say which quantum effects and why those particular ones, the gap is on the page. You can't unsee it. Comfortable studying never produces this experience; that's why comfortable studying doesn't work.

Step 4 forces depth. When you go back to fill a specific gap, you're studying one targeted thing with high attention, not skimming twenty pages with low attention. The information-per-minute is an order of magnitude higher.

This is why people who say "I tried the Feynman technique and it didn't work" almost always describe doing step 2, sort of, while looking at their notes. That's not the Feynman technique. That's reading aloud.

A real worked example: Bayes' theorem

Let me do this in front of you. The concept: Bayes' theorem.

Attempt 1. Bayes' theorem is a way to update your belief in something when you get new evidence. You take the probability of the evidence given the hypothesis, multiply by the prior probability of the hypothesis, and divide by the probability of the evidence overall.

OK. Stop. Where are the gaps?

• I said "probability of the evidence overall" — what does that mean? Why is that the denominator? I used the formula's structure as a shield. Step 3: gap.
• I didn't explain why this is useful. The whole point of Bayes is that human intuition gets it wrong; my explanation doesn't make that visible.
• "Updating your belief" sounds nice but isn't operationalized. If someone asks "update by how much?", I have no answer.

Attempt 2. Bayes' theorem solves a specific problem: most evidence is consistent with multiple hypotheses, and our intuition tends to update too strongly toward whichever one we thought of first. The theorem corrects this by anchoring every update to two numbers we usually ignore: (a) how often the hypothesis is true before we see any evidence (the base rate), and (b) how often the evidence shows up under each alternative hypothesis, not just the one we're considering.

This is a much better attempt because it survives the "so what?" test. But there's still a gap: I haven't said how the formula's denominator implements "alternative hypotheses." That's the next attempt.

Attempt 3. The denominator — P(evidence) — is the total probability that the evidence happens at all, summed across every way it could happen. So if the hypothesis is true, the evidence shows up at one rate; if the hypothesis is false, it shows up at another rate. The denominator sums those weighted by how likely each world is. Dividing by it normalizes — it forces the answer to account for the alternative explanations, which is the part untrained intuition skips.

Now we're somewhere. After a third pass, I have a story that doesn't lean on the formula's notation as a crutch. And — crucially — I noticed exactly what I didn't understand: the role of the denominator. I would never have noticed that by re-reading a textbook. I noticed it by trying to explain it.

That whole loop, including writing the gap audit, took ~6 minutes. That is more productive than 6 hours of highlighting.

Common misuses

1. Explaining it to yourself in your head. Working memory has a 7-item-ish cap. Internal explanations let you cheat: you can fudge a step and forget you fudged. Writing makes the fudges permanent. Write it down or it didn't happen.

2. Explaining it from your notes. You're not retrieving; you're transcribing. The whole point is that the page should be blank when you start. If you peek every 30 seconds, your brain marks the information as "stored externally" and stops trying to hold it.

3. Using jargon as a shortcut. "Bayes' theorem is just probabilistic inference under uncertainty using the conditional structure of P(A|B)." This is technically true and totally useless — the jargon is doing all the work. A useful rule: any time you use a phrase your imagined audience wouldn't know, you've found a gap. Replace it.

4. Stopping at attempt 1. The first explanation is always bad. Step 4 — re-explain — is the entire technique. People skip it because the first attempt feels like the deliverable. It isn't. It's the diagnostic.

When not to use it

Feynman is for conceptual material — ideas, theorems, frameworks, mechanisms. It is wrong for:

• Procedural skill. You don't learn the violin or a programming language by explaining them. You learn them by doing them. The explanation-style retrieval doesn't map onto the muscle-memory retrieval those skills require.
• Pure memorization. Memorizing the periodic table benefits from spaced repetition, not from explaining. There's no "understanding" to find gaps in.
• Wide reading. Some material is exposure-based — you read a hundred history papers to absorb a sense of the field. You wouldn't Feynman every one; you'd never finish.

A good heuristic: if someone could ask you "why is that?" and you'd need to answer in a sentence longer than the original claim, Feynman it. Otherwise, use the right tool instead.

Pairing with other techniques

The Feynman technique gets oversold as a one-size-fits-all method. It works best as one node in a study workflow:

• Pair it with spaced repetition for retention. Feynman gets the concept in; spaced review keeps it there. Each gap you fill becomes a flashcard you actually understand.
• Pair it with active recall practice problems. Feynman tests whether you can explain; problems test whether you can use. You need both.
• Pair it with a deadline. The most reliable failure mode is open-endedly Feynman-ing for 90 minutes without finishing one concept. Cap each concept at 25 minutes. If you blow through, the concept needs to be split — it's two concepts in a trench coat.

How to actually run it this week

Pick one thing you've been "studying" for more than a week and can't fully explain. Sit down with a blank page. Set a 25-minute timer. Don't open notes. Write the explanation as if you were emailing it to someone in a different field.

Stop at the first stall. Don't push through it — note it. ("I don't know why we use the chain rule there.") Then go look up just that gap. Come back. Continue.

When the timer runs out, you should have two artifacts: a draft explanation you can show a friend, and a list of 3-6 specific gaps you didn't know you had until you started writing.

The next day, do it again, blank page. The gaps should be smaller. The third day, they should be different gaps — deeper ones. By day five, you'll have an explanation you can give without notes, which is the only criterion that matters.

If you want to keep going, the natural complements are active recall (the broader principle Feynman is one form of), spaced repetition (for retention), and getting unstuck on a hard topic (when a gap turns out to be a wall).