The Feynman Technique, Used on Real Coursework

The Feynman technique is usually explained with a line about a five-year-old and then never demonstrated on anything harder than photosynthesis. So here it is applied once, in full, to a topic from a second-year course, including the parts of the first explanation that fell over.

What the technique is asking of you

Richard Feynman was a physicist known for explaining difficult physics in plain language, and for the point that the person you are most likely to fool about your own understanding is yourself. The tidy numbered version of the technique that circulates under his name was assembled by other people later. That is not a problem, but treat it as a method someone wrote down rather than something handed over by a Nobel laureate.

The method: explain the topic in writing, without notes, until you hit the places where you cannot. Then go back for those specific places only, and explain it again.

The audience is where most attempts go wrong. A five-year-old accepts anything, so imagining one gives you no feedback. Pick a real reader instead: a classmate on the same course, who has the prerequisites and missed this week's lecture. That person will not let you hide behind jargon, and will not need everything reduced to analogy either.

The topic: the central limit theorem

Statistics, second year. Everyone can state it. Far fewer can say what it does not claim, which is where the marks are.

First attempt, written cold

"The central limit theorem says that if your sample is large enough, the data becomes normally distributed. That is why we can use normal-based tests on almost any data, as long as n is above about 30."

Two sentences, produced fluently, and mostly wrong. Reading the chapter again would not have revealed that, because the chapter says all of this correctly and my eyes would have nodded along with it.

Where it broke

Three separate failures, and they are three different kinds.

A claim that is simply false. The data does not become normal. Sample a thousand household incomes and you have a thousand skewed incomes. The theorem is about the distribution of the sample mean across repeated samples, not the shape of the observations inside one sample. That single confusion is enough to lose an entire question.

A borrowed phrase doing no work. "n above about 30" arrived in the sentence fully formed because a lecturer said it. Asked where 30 comes from, I had nothing. It is a rule of thumb, not part of the theorem, and how large a sample needs to be depends on how skewed the population is.

A sentence I could not finish. I started "as long as" and stopped, because I could not name the conditions. The theorem needs independent draws from a distribution with finite variance. I did not know those were missing from my understanding until I tried to say them.

Those three failures are the entire product of the exercise. None of them would have surfaced from re-reading, and only the first would have surfaced from a multiple-choice quiz.

Going back for three things, not for the chapter

The instinct at this point is to restudy the topic. Do not. You now have a list of three specific gaps, and restudying the chapter buries them in forty minutes of things you already knew.

Go back for exactly these: what the theorem is a statement about, what conditions it requires, and where the 30 came from. That is a few paragraphs of a textbook and perhaps two minutes of a recorded lecture. The difficulty is usually finding them, which is why it helps when the topic keeps its own locations. In SymbioLearn a Concept holds every Source underneath it with the page, section or timestamp attached, so going back for one paragraph does not mean scrubbing through a fifty-minute recording to find it.

Second attempt

"Draw repeated independent samples of the same size from a population with finite variance, and take the mean of each sample. Those means have a distribution of their own. As the sample size grows, that distribution approaches a normal one, centred on the population mean, and narrower for larger samples. This is what lets us use normal-based tests on the mean of skewed data. It says nothing at all about the shape of the data itself. How large the sample must be depends on the population: with mild skew a small sample is enough, and with heavy tails it can take far more than the usual rule of thumb suggests. If the variance is infinite, the theorem does not apply."

That version is longer than the first one, which is fine: you are aiming for an explanation with no borrowed phrases in it and no sentences you cannot finish, not for a short one.

Doing this on the rest of your course

A few things make the difference between the technique working and it becoming another way to re-read.

Write it, or say it out loud. Running it silently in your head lets you skip the gaps without noticing, because thought does not have to be grammatical and a sentence does.

Keep the first attempt. The wrong version is more useful for revision than the corrected one, because it names how your understanding actually bends, and it will bend that way again in a month.

Underline anything you wrote in the lecturer's exact words. Those are the hollow spots, because a phrase you can only reproduce is one you have not had to unpack yet.

Use it once per topic, not once per fact. Fifteen to thirty minutes on an idea is reasonable. The same on a definition is not.

When it is the wrong tool

Ten minutes explaining why heat denatures an enzyme is time well spent. Ten minutes explaining what "denature" means is not. Terminology, notation, formulas and mappings need repetition, not explanation, and Flashcards or a short quiz will do more for them in a quarter of the time. Applied to a definition, this technique produces a longer definition.

What to do next

Pick the topic on your current course you would least like to be asked about out loud, and write your explanation of it before you reopen anything. Mark the three failures above: the false claim, the borrowed phrase, the sentence you could not finish. What you find is your revision list for the week, and it will be shorter and more specific than one you would have written by reading.