There is a particular silence that falls in the middle of an explanation. You were sure you understood how a zipper works, or what a confidence interval means, or why the seasons change — and then, three sentences in, the words run out. The concept that felt solid in your head turns out to have been a label wrapped around fog.
That silence is one of the most useful signals in all of learning, and the Feynman Technique is a method for producing it on purpose. Instead of waiting for an exam or an awkward conversation to expose what you don't know, you sit down and try to explain the idea in plain language — and let every stall, hedge, and hand-wave show you exactly where the fog is.
The illusion of explanatory depth
In 2002, the psychologists Leonid Rozenblit and Frank Keil published a series of studies with a deceptively simple design. They asked people to rate how well they understood everyday objects — zippers, flush toilets, cylinder locks. Then they asked them to actually write out how those objects work, step by step. Afterward, participants rated their own understanding again.
The ratings collapsed. Having tried to explain, people realized they couldn't — and their confidence dropped to match. Rozenblit and Keil called this the illusion of explanatory depth: we routinely mistake familiarity with a thing for an understanding of it. We have seen zippers our whole lives, we can use the word fluently in a sentence, and so the mind quietly files the mechanism under "known."
The unsettling part is that the illusion is invisible from the inside. Confidence feels the same whether it is earned or not. The one reliable way Rozenblit and Keil found to puncture it was the act of explanation itself. You cannot introspect your way to the gap; you have to talk your way into it.
What the Feynman Technique actually is
The physicist Richard Feynman was famous among colleagues for something rarer than genius: he could not tolerate his own fog. His Caltech colleague David Goodstein recalled asking him once to explain why particles with half-integer spin obey the exclusion principle. Feynman said he would prepare a freshman lecture on it — then returned days later and admitted he couldn't. If it couldn't be brought down to that level, he said, it meant physicists didn't really understand it yet.
The four-step method that now circulates as "the Feynman Technique" was codified later by admirers rather than by Feynman himself, but it is faithful to how he worked:
Pick one concept and write its name at the top of a blank page. One concept — not a chapter.
Explain it in plain language, in writing, as if to a curious twelve-year-old. No jargon allowed unless you define it in the same breath. Jargon is where fog hides: a term like "statistical significance" can sit in a sentence doing no work at all, and only when you're forbidden from using it do you discover whether you know what it means.
Mark every stall. The places where your sentences slow down, hedge, or leap — "...and then it just, you know, equalizes" — are not writing problems. They are a map of what you don't know. Go back to your source material and repair exactly those spots.
Return and simplify. Rewrite the explanation, shorter. If you can carry the idea in an analogy that holds up under pressure, you've compressed it into something your memory can actually keep.
Why explaining works: the self-explanation effect
This isn't productivity folklore; it rests on one of the better-replicated findings in learning science. In the 1990s, the cognitive scientist Michelene Chi and her colleagues studied students reading a biology text about the circulatory system. Students prompted to explain each passage to themselves as they went — what does this mean, how does it connect to what came before — ended up understanding the material far more deeply than students who simply read it, and were better at answering questions the text never directly addressed. Researchers call this the self-explanation effect.
Three mechanisms do the lifting. Explaining is a form of generation: you produce the idea rather than recognize it, which strengthens memory more than passive review does. It forces integration: to say why the valve matters, you have to connect it to pressure, to flow, to things you already know — and connected knowledge has more retrieval routes leading back to it. And it surfaces contradictions: a vague mental model can coexist peacefully with a textbook for weeks, but a written sentence that contradicts itself is hard to miss.
There is even evidence that merely expecting to explain changes how you learn. In studies by John Nestojko and colleagues, students told they would later teach a passage to someone else remembered it better than students told they would be tested on it — same text, same time, different posture toward the material.
How to do it well
Write it down; don't just narrate it in your head. Thought is forgiving. It tolerates gaps, skips steps, and lets a vague gesture stand in for a mechanism. A written sentence has to end, and its verbs have to do something. The illusion of explanatory depth survives thinking about a topic; it rarely survives a paragraph.
Give yourself a real audience. "Explain it simply" is vague; "explain it to my sister, who is smart but never took statistics" is a constraint that does real work. The imagined listener determines which words you are allowed to lean on.
Collect your stalls as questions. Every hedge becomes a question: why does the current flow that way? What exactly does the enzyme bind to? Keep the list. It is the highest-value study agenda you will ever write, because it is made entirely of your own gaps rather than a textbook's guess at them.
Check the explanation against the source. Self-explanation has a known failure mode: fluent nonsense. You can produce a beautiful, simple, wrong explanation and feel wonderful about it. The loop only closes when you go back and verify that your plain-language version still matches the real thing.
Where the technique fails
Honesty requires the limits. A proper Feynman pass on one concept can take half an hour, so it belongs on the load-bearing ideas of a subject, not on every fact. It also has nothing to grip on pure associations: there is no "why" inside the fact that gato means cat, or that the ulnar nerve runs behind the medial epicondyle. Those simply have to be retained, and no amount of explaining will do it.
And there is a quieter limit: understanding something today does not mean you will have it in March. Explanatory depth, once earned, decays like everything else in memory. Feynman himself relearned things constantly; his notebooks were tools for returning to ideas, not trophies of having visited them.
Explanations fade too
Which suggests the natural pairing. Use explanation to find the gaps and build the structure; use spaced retrieval to keep the structure standing. The stall-questions you collected are already the right shape — turn them into prompts that demand explanation rather than recognition. Not "define osmosis" but "explain why the cell swells." A prompt like that, answered from memory at expanding intervals, makes you regenerate the understanding each time: a small Feynman pass, sixty seconds long, scheduled for exactly the moment you would otherwise begin to lose it.
This is where a tool earns its keep. Recall is a flashcard app built for that second half of the loop: you turn your stall-questions into cards, and its FSRS scheduler — a modern spaced-repetition algorithm — notices when each explanation is about to slip and resurfaces it just in time. It's fast enough that a one-minute re-explaining session actually happens, it works fully offline, and it imports your existing Anki or Quizlet decks if you've already started elsewhere. The Feynman Technique tells you what you don't yet know; Recall makes sure that once you finally know it, you keep it. Try it at recall.lumenlabs.works.