For some time I’ve been doing an exercise I rarely talk about. I think it’s one of the most useful ideas I’ve learned in recent years, and also one of the most uncomfortable.
The exercise takes thirty seconds. You can do it anywhere – at your desk, on the train, after reading an article.
You ask yourself: could I explain this right now to someone who has never heard of the subject?
You explain it now, out loud, without jargon, without evasions, without “it’s complicated”, in sentences your mother or your neighbour could follow. Much of what I believe I understand fails this test.
The exercise isn’t mine, and it isn’t entirely the invention of the man whose name it bears today.
Richard Feynman prided himself on making even the deepest ideas accessible to first-year students. One day a colleague at Caltech asked him to explain why particles with half-integer spin obey Fermi-Dirac statistics. Feynman agreed: he would prepare a freshman lecture on it. A few days later he came back. He hadn’t managed it, he said; he couldn’t bring it down to that level – and that meant we didn’t really understand it.
This standard is documented: what cannot be explained at first-year level has not been understood. The short form that grew out of it sounds modest and is radical: if you can’t explain something simply, you haven’t understood it.
What circulates today as the “Feynman technique” was never called that by Feynman himself. The name spread through Scott Young’s self-learning project, and Young himself notes that the technique is only loosely based on Feynman’s actual practice.
Feynman distinguished between two kinds of knowledge: knowing the name of something and knowing the thing itself. You can know the name of a bird in every language in the world and know nothing about the bird. You can recite the formula for gravity flawlessly and still not understand what gravity is.
I first did the test before a presentation. I was supposed to explain to the team how our new scoring algorithm worked. I had read the documentation, studied the diagrams, followed the logic. I felt prepared.
Then I imagined a listener who knew nothing about the subject, someone who only asked: what does it actually do?
I tried, out loud, alone in my office, door closed. Two sentences. Then I stalled. The terms I was using I couldn’t define myself. I was talking around the essential point. I didn’t know what the essential point was.
Up to that moment I would have told anyone I understood the thing. The feeling was there; the substance was missing.
Why does the exercise work?
It’s about direction. When you read, your recognition runs: patterns appear, it’s cheap, it flows, it feels like understanding. When you explain, your brain has to generate – select, order, connect with prior knowledge. That’s more expensive, and it’s more honest.
Cognitive science calls this the generation effect: information you produce yourself is processed more deeply and remembered better than information you read passively. Slamecka and Graf showed this in 1978 with simple word tasks. The effect has since been widely replicated, moderate in size and robust, though its magnitude depends heavily on the experimental design. Chi and colleagues found the leap to understanding in 1994: eighth-graders who explained each line of a text to themselves understood it more deeply than the control group, most clearly those who explained a lot.
Exactly at the point where you stall while explaining, your feeling promised too much. It had reported familiarity, and you read understanding into it. I call the phenomenon behind this sham understanding: the feeling of having understood without having understood.
Your feeling doesn’t do this on purpose. Your brain is built to keep you able to act, and a person who constantly doubts their own knowledge gaps is less able to act than one who believes their explanations.
Writing it down, by the way, helps more than just thinking. On paper you see where you’ve been bluffing your way through with vague terms.
Three objections.
The exercise has limits, and I run into them regularly. If I did it all the time, I’d never get anything done. So I do it before decisions with consequences, before presentations and after explanations I got from a system – where doubt is worth it. These places are often different from the ones you’d first assume. ¶ It doesn’t create knowledge either. If you know nothing about a subject, explaining means talking into the void; the generation effect assumes there is something in memory that can be activated. The question only shows where your understanding ends. That’s why I read first and explain afterwards. ¶ The trickier part is that an explanation you give yourself can become just as smooth as one you receive. You explain something to yourself, it flows, you’re satisfied, and you’re wrong. An imagined listener nods as soon as you let them; a real one knows no more than you, but asks exactly where you kept talking. I have known the research on this for years, and yet after a talk on quantum computing I went home satisfied. Only when my wife asked what it had been about did I stall.
And the question doesn’t work everywhere. Asking yourself whether you can cook a recipe teaches you nothing new. It is a precision instrument for one kind of knowledge only: causal relationships in complex systems. And that is exactly what a machine gives you when you ask it how something works.
There, though, the basic form isn’t quite enough, because when you explain back you easily repeat the machine’s words instead of finding your own. Which variations hold up then is in my book (in German).
If you run out of words while explaining and fall back on jargon or platitudes, that’s a find, and a useful one, even if it feels embarrassing at first. You’ve just seen something that was invisible before.
The alternative is to carry on, nod, feel informed. That’s easier. It’s also what our information environment systematically rewards. And the question needs something in memory that can be activated. What happens to that store when every answer is only a tap away is the subject of the chapter on Google amnesia in my book.
Could you explain this now to someone who has never heard of it?
If yes – explain it to someone who asks questions. If no – interesting. Then you know where the work begins.
The full essay on this topic:
