Does teaching someone else actually help you learn?
In short
- Expecting to teach does almost nothing on its own. Actually explaining does the work.
- The explaining has to be from memory. Reading your notes aloud to someone is not teaching.
- Questions from the listener are where most of the gap gets found.
- None of the research requires the listener to know the subject.
Every study guide says it: if you want to know whether you understand something, teach it to someone else. It is one of the few pieces of study advice that is both widely repeated and actually supported. It is also routinely misquoted in a way that removes the part that works.
The research question is narrower than the slogan, and the narrow version is more useful.
What was actually tested
The cleanest work on this splits the folk advice into two separable things: expecting to teach, and actually explaining.
Fiorella and Mayer ran the comparison directly. Students studied a passage. Some were told they would be tested; some were told they would have to teach it. Of the ones told they would teach, some then actually taught — explaining the material to a camera without their notes — and some only prepared and were tested instead.
The result that matters: preparing-to-teach and then stopping produced little advantage. Preparing to teach and then actually doing it produced a real one, and the gap grew on the delayed test rather than shrinking.
That is the sentence to keep. The expectation is not the mechanism. The explaining is.
Why the explaining works
Nothing mystical is happening. Explaining from memory forces three things that reading does not:
You have to retrieve it. You cannot explain a mechanism you are simultaneously reading. Retrieval is the single most reliable memory intervention in the literature — being made to produce an answer beats re-reading the same material, and the advantage widens as the delay grows (Roediger & Karpicke, 2006).
You have to produce it out loud. Material spoken aloud is remembered better than the same material read silently — the production effect (MacLeod et al., 2010). The effect is modest per item and it is essentially free, because you were going to review the material anyway.
You have to organise it. A list of facts can be recognised in any order. An explanation has to have a beginning, a because, and an end. Building that structure is where people discover they have four facts and no thread connecting them.
What the research does not say. It does not say teaching beats every other method, and it does not say you need an audience who can evaluate you. It also does not license the version of the Feynman technique where you write a simplified explanation with the textbook open — that removes the retrieval, which is the part carrying most of the effect.
The part nobody mentions: the questions
In the controlled studies, the “teaching” is often a monologue to a camera. That is enough to produce the effect. But anyone who has actually explained something to a classmate knows the monologue is not where the holes show up.
The holes show up when they say wait, why?
An explanation you give unchallenged is shaped by what you happen to remember. A follow-up question is shaped by what you failed to say. Those are different sets, and the second one is the one you needed.
This is also why explaining to a patient audience who nods along is much less useful than explaining to someone slightly sceptical. You are not trying to be understood. You are trying to be caught.
The practical problem
So the method is: explain the topic from memory, out loud, to someone who will ask you why.
Try to schedule that daily.
The obstacle is not motivation. It is that a study partner has to be free at the same time as you, working on the same material as you, and willing to interrupt you. Most students can manage that a few times a term. The students who manage it weekly are usually in a cohort small enough to make it happen by accident.
Which leaves most people doing the thing the research supports least: re-reading, with the notes open, silently.
Getting the mechanism without the person
Look at what the listener actually has to do. They have to be there, listen, and ask a question that targets what you skipped. They do not have to know the subject better than you — in the studies, the “student” was often a camera.
That is a low bar for a person and a normal one for software. It is why Live Talk is the agent KoiSwarm is built around: you explain the topic out loud from memory, it asks the follow-up, and it marks what you knew separately from how clearly you said it.
The thing to keep from the research, whether or not you use any software for it: close the notes, say it out loud, and make something ask you why. That is the whole of it.
Sources
- Fiorella, L., & Mayer, R. E. (2013). The relative benefits of learning by teaching and teaching expectancy. Contemporary Educational Psychology, 38(4), 281–288.
- Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
- MacLeod, C. M., Gopie, N., Hourihan, K. L., Neary, K. R., & Ozubko, J. D. (2010). The production effect: Delineation of a phenomenon. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(3), 671–685.