The Feynman technique: explain a concept and check the gaps

The name “Feynman technique” is popularly used for a routine of simple explanation and gap checking. Here it is a practical exercise, not a protocol validated by Feynman or a guarantee of learning anything.

A clear explanation is a useful signal, not a complete test of mastery. An explanation can be fluent and wrong. Compare it with a reliable source or ask someone to review it.

Choose a concept, explain it without looking, mark what you cannot justify and return to the source. Rewrite it and test it with a different question.

“Every rectangle is a square” is easy to understand but false. A correct explanation distinguishes the conditions: a rectangle has four right angles; a square also has four equal sides.

Try a rectangle with sides of length 2 and 3. It meets the rectangle condition but not the square condition. A counterexample reveals a gap that fluent wording did not.

Write one question per distinction: “What additional condition must a rectangle meet to be a square?” Use a separate card to ask for a counterexample to the original claim.

You can prepare those questions in Lumia AI. The Dunlosky review linked below provides context on self-explanation and practice testing; it does not establish that this app is superior or validate every popular version of the technique.

After explaining, look for a boundary condition, a counterexample and an application question. In science, change one variable and predict what follows. In history, distinguish evidence, interpretation and consequence. In law, separate the general rule from an exception. The purpose is to find where your model stops holding.

Compare every claim with the source and label what is your own inference. If you used an analogy, write where it breaks down. An analogy can make a relationship visible, but it can hide important differences when memorized as a literal description.

Simplifying does not mean removing necessary technical language. Define the term the first time, then use it precisely. Do not confuse speaking at length with explaining, either. Set one question, causal chain or procedure and check whether every part serves that objective.

For mathematics, programming or clinical tasks, add problems and supervised performance. For a quotation, legal definition or formula that must be reproduced exactly, retain the original wording. Explanation can locate uncertainty; other tasks must test performance and precision. Ask another reader to identify any step that your explanation assumes without stating.

Dunlosky and colleagues’ 2013 review, linked below, examined self-explanation as a distinct learning technique. It did not evaluate a routine named the “Feynman technique,” prescribe these four steps or establish that explaining to an imagined listener works equally well for every subject. This guide uses explanation as a diagnostic prompt that still requires verification.

End with a task you have not rehearsed word for word. In physics, predict what changes when one variable changes; in programming, run an edge case; in history, compare your account with a source offering another interpretation. If you cannot complete the task, record the exact premise or step that is missing and return only to that part of the material.