The Protégé Effect: Why Children Learn More From Teaching Than From Being Taught
Children will work harder to learn something for someone else than they will for themselves. That’s the protégé effect, and it isn’t a metaphor - it’s been measured. In a Stanford study, eighth-graders who believed they were teaching a computer character spent more time reading and learned more biology than classmates doing identical work for their own benefit. The gap was widest among the lowest-achieving children.
Same software. Same material. The only difference was who the child thought the effort was for.
What the protégé effect is
The claim is narrow and specific: taking on the role of teacher changes how a learner engages with material, and improves what they retain - beyond the benefit of simply studying it.
It’s distinct from two things it gets confused with. It isn’t the general observation that teaching is a good job for learning things, which is true but unhelpful. And it isn’t peer tutoring, where an older child helps a younger one - though the two overlap, and peer tutoring is where most of the hard numbers come from.
The protégé effect is about motivation and effort. The teaching role recruits something that studying for yourself doesn’t.
The Betty’s Brain studies
The canonical work is Chase, Chin, Oppezzo and Schwartz, published in the Journal of Science Education and Technology in 2009 under a title that says it plainly: “Teachable Agents and the Protégé Effect: Increasing the Effort Towards Learning.”
The setup is elegant. Betty’s Brain is software where a student teaches a character - building out what she knows about, say, a river ecosystem, then watching her reason from it. When Betty gets a question wrong, she’s wrong because of how you taught her. Her errors are your gaps, externalised.
The experiment’s cleverness is in the control. Eighth-graders all used the same Betty’s Brain software. One group was told they were teaching Betty. The other was told they were learning for themselves. Nothing else differed.
The teaching group spent more time on learning activities - more time reading, specifically - and learned more. And the effect was most pronounced for lower-achieving children, which is the opposite of how most interventions behave. A second study, using verbal protocols with fifth-graders, went looking for why.
The finding that makes this interesting isn’t “teaching helps.” It’s that a child’s willingness to do the boring part - read the passage again, carefully - moved because of who they believed they were doing it for.
How big is the effect?
Here’s where you need to be careful, because the protégé-effect literature is small and the peer-tutoring literature is large, and people blur them.
| Intervention | Effect size | Source |
|---|---|---|
| Peer tutoring, tutors’ own achievement | 0.43 | 2019 meta-analysis |
| Cross-age tutoring, overall | 0.34 | 2025 meta-analysis |
| - tutees | 0.33 | same |
| - tutors | 0.39 | same |
| Peer tutoring, single-case research | TauU 0.75 | 2013 meta-analytic review |
| Tutoring programmes generally (RCTs) | 0.30–0.40 SD | multiple meta-analyses |
Read the cross-age row twice. Tutors gain more than tutees - 0.39 against 0.33. The person doing the explaining learns more than the person receiving it.
For context, an effect size around 0.3–0.4 is a serious educational intervention. It’s roughly what well-run tutoring produces overall, and tutoring is about the most reliable thing we know how to do in education.
One more number worth knowing, because it’s the one that punctures a common shortcut: teacher and paraprofessional tutoring programmes outperform non-professional and parent tutoring on average. Enthusiasm doesn’t substitute for training. Which is awkward, and we’ll come back to it.
Why it works: three mechanisms
Retrieval. To explain something you have to pull it out of memory and assemble it, not recognise it on a page. Retrieval practice reliably outperforms passive review - and in one well-known comparison it beat elaborative concept mapping even when the final test asked students to produce a concept map. Explaining is retrieval with the volume turned up.
Gap detection. This is the mechanism people underrate. When you explain, you find out where your understanding stops - not vaguely, but at a precise word. Every place you reach for jargon, or say “and then it just works”, is a located gap. Reading the chapter again doesn’t produce that map. It produces a warm feeling of familiarity, which is not the same thing and is frequently mistaken for it.
Accountability. The Betty’s Brain finding is squarely about this. A child who has to answer for someone else’s understanding behaves differently from a child answering only for their own. It’s not that they care more about Betty than themselves. It’s that responsibility to another party converts a vague intention into a specific obligation.
The catch: expecting to teach isn’t enough
There’s a finding that keeps the whole thing honest. A 2013 study found that people who expect to teach but then don’t are outperformed by people who expect to teach and actually do.
So the effect isn’t a mindset trick. Telling your child “imagine you’ll have to explain this to your sister” will produce some benefit at the encoding stage, and then most of it evaporates if the explaining never happens. The act is doing the work, not the anticipation.
This matters practically, because the easy version of this advice - “get them to think of themselves as a teacher” - is the version that doesn’t work.
What to do when your child has nobody to teach
Which brings us to the real problem. The protégé effect requires an audience, and most children don’t have one. A younger sibling is the classic answer and is available to roughly half of families, tolerant for roughly ten minutes, and uninterested in Class 8 chemistry.
The obvious substitute is you. And here the research gets uncomfortable: parent tutoring underperforms professional tutoring, and separately, a large study across two nationally representative cohorts of about 20,000 children each found parental homework help produced no measurable achievement benefit in primary school. Worse, where a parent is anxious about maths and helps frequently, children learn less maths and are more anxious about it than children of equally anxious parents who help rarely.
So “sit down and let them explain it to you” is better than nothing but carries a specific hazard: the moment you correct, you’ve become the teacher again and the child has become the student again, and the whole mechanism inverts.
The practical version, then, is to be a bad audience on purpose:
Don’t know the answer, visibly. “I genuinely don’t remember how this works - walk me through it” is a better opening than “show me what you learned.” The first invites explanation. The second invites performance.
Don’t correct. Ask. When they say something wrong, the useful move is “hang on, I didn’t follow that bit” rather than “no, it’s actually…”. You’re pointing at the gap, not filling it. This is hard and you will fail at it, which is fine.
Ask for the wrong version. “What would someone who got this wrong have done?” is close to unguessable. A child can produce a correct answer they don’t understand. They cannot construct a plausible misconception without understanding where the reasoning forks.
Let the silence run. The gap-detection mechanism operates in the pause where they realise they can’t finish the sentence. Rescuing them from that pause is the single most common way parents dissolve the benefit.
Ten minutes, not an hour. This is effortful in a way re-reading isn’t. That’s the point - and it’s also why it doesn’t scale to a long session.
The short version
Teaching is not a reward you earn after learning. It is one of the more reliable ways to learn, it works best for the children who are struggling most, and the person explaining gains more than the person listening.
The hard part has never been the principle. It’s that it needs an audience who doesn’t know the answer, won’t get bored, and won’t take over - and that is genuinely difficult to supply at 8pm on a Tuesday.
Frequently asked questions
What is the protégé effect?
The finding that learners put in more effort and learn more when they believe they’re learning in order to teach someone else. It was named and demonstrated in Chase, Chin, Oppezzo and Schwartz’s 2009 teachable-agent studies at Stanford.
Does teaching someone else really help you learn?
Yes, and it’s measurable. In cross-age tutoring meta-analyses the tutors gain slightly more than the tutees (0.39 vs 0.33), and peer-tutoring meta-analyses put tutors’ own achievement gains around 0.43.
Is the protégé effect the same as the Feynman Technique?
They’re closely related. The Feynman Technique is a study method - explain it simply, find where you stall, go back and fill the gap. The protégé effect is the underlying research finding about why that works and why a real audience matters.
Does it work if you only imagine teaching?
Partly, and much less. A 2013 study found people who expected to teach but didn’t were outperformed by those who expected to teach and did.
Who benefits most?
Lower-achieving children. In the Betty’s Brain study the gains were most pronounced for them, which is unusual - most educational interventions widen gaps rather than narrowing them.
Can a parent be the audience?
Yes, with care. Parental homework help shows no measurable benefit on average in primary school, and frequent help from a maths-anxious parent is associated with worse outcomes. The mechanism survives only if you stay the audience and don’t become the teacher.
Related reading
- 86% of Kids Aged 9–17 Now Use AI - the data behind why “explain it back” is the real test for AI-assisted homework.
- CBSE’s AI Curriculum for Classes 3–8 - what the new curriculum actually asks a Class 6–8 child to do, hour by hour.