Everyone who has taught something knows the feeling: you did not really understand it until you had to explain it. The interesting finding is that you do not have to explain it. You only have to believe you will.
The experiment
Nestojko and colleagues gave two groups the same passage, the same time, the same conditions. One group was told they would be tested on it. The other was told they would teach it to another student.
Nobody taught anybody. The teaching never happened. Both groups were then tested identically.
The expecting-to-teach group recalled more of the passage, organised their recall better, and did particularly well on the main points rather than the incidental details. A single sentence of instruction, changing nothing about the material or the time spent, changed what people took away from it.
Why the expectation does the work
Because "I will be tested" and "I will explain this" are instructions to read in different ways.
Preparing for a test is a recognition task, and readers optimise for it accordingly: cover the material, keep everything, be able to spot the right answer when it appears. Preparing to teach is a reconstruction task. You cannot deliver a passage, you have to deliver its shape. So you start doing things the test-readers do not:
- Sorting by importance. A teacher has to lead with something. That forces a judgement about what the main point actually is, which many students never make.
- Building a structure. The study found the teaching group's recall was more hierarchically organised. They had built a tree, not a list.
- Hunting for the gaps. You read differently when someone might ask a question. The parts you were skimming past become the parts you slow down on, because you can feel that you could not defend them.
That last one is the real mechanism. Teaching expectation makes your own confusion visible to you, which is exactly what the illusion of competence normally hides.
It survives the audience being fake
The Stanford work on teachable agents pushed this further. Students taught a software character, a cartoon agent that would then take a quiz. Students worked harder and learned more on behalf of the agent than they did for themselves, and cared visibly about how their agent performed.
The audience was a piece of software. The effect held anyway.
This matters practically, because it means you do not need a study partner, a tutor, or anyone's time. An empty room works. A voice memo works. Writing to a hypothetical reader works, which is the entire premise behind evergreen notes being written in full sentences for a future stranger rather than as private shorthand.
How to use it deliberately
Set the intention before you read, not after. This is the whole trick and it is easy to get backwards. The effect comes from reading with the expectation in place. Deciding to teach it afterwards is just review.
Say it explicitly: I am going to have to explain this to someone on Thursday. Then read.
Give yourself a specific victim. "Explain it to someone" is vague enough to ignore. "Explain it to my sister, who is smart and knows nothing about this" produces different, better notes, because it fixes the level you have to hit.
Then actually deliver it, cheaply. The expectation gets you the encoding benefit. Delivering it gets you the retrieval benefit on top, which is the testing effect at full strength. Cheap deliveries that work:
- A three-minute voice memo, no script, recorded once.
- A written summary in your own words with the source closed. This is the Feynman technique.
- Explaining it to a study partner, alternating who teaches what.
- Answering questions about it on a forum, which is teaching with a real audience attached.
Notice where you go vague. When you explain out loud, listen for the moment your language turns abstract and hedged. That sentence is the one you do not understand. Vagueness is the tell, and it is far more reliable than your own sense of whether you know something.
Where it fits with everything else
The protégé effect is not a replacement for retrieval practice, it is an upgrade to the encoding stage that most techniques ignore.
A sensible stack looks like this: read expecting to teach, take notes as though someone will read them, ask why as you go using elaborative interrogation, then test yourself with active recall and space the reviews with spaced repetition.
The first two cost nothing and change the quality of everything downstream. A note taken in teaching mode is already halfway to being a good note, because the work of deciding what matters has already happened.
The uncomfortable corollary
If explaining something reveals whether you understand it, then avoiding explanation is a way of protecting the feeling of understanding.
Most people's study habits are quietly built around that protection: rereading, highlighting, nodding along to a lecture. All pleasant, all silent, all safe. The protégé effect works because it removes the option of staying silent, and the first few attempts are meant to feel bad. That is the technique reporting your actual position, which is the only useful thing a study technique can do.
More in why writing helps memory, how to study from notes, and the wider memory and focus section.