Why you learn better with someone else (even a non-expert)

You and a friend are staring at the same confusing thing: a tenancy agreement, a chart in the news, the rules of a board game nobody has played since 2019. Neither of you is an expert. You talk it through for five minutes, argue a bit, and suddenly it makes sense to both of you.

That feels like luck. It isn’t. It’s one of the better-tested findings in the science of learning: two people who don’t yet know something can often talk their way to understanding it, better than either would manage alone. Here’s what the research shows, why it works, and how to use it on purpose.

Can two people who don’t know the answer work it out together?

More often than you’d think. The classic evidence comes from a Harvard lecture theatre. Around 1990, physicist Eric Mazur came across work by Ibrahim Halloun and David Hestenes showing that standard teaching barely shifted students’ misconceptions about basic physics. He gave a test of the core ideas to his own students, expecting them to do fine. Many who could crunch complicated equations didn’t really understand the concepts underneath.

So he changed how he taught. In what became known as peer instruction, he posed a tricky conceptual question, had everyone commit to an answer on their own, then asked them to turn to a neighbour and try to convince each other. Then everyone answered again.

In 2001, Catherine Crouch and Mazur published ten years of results. Students taught this way improved in both conceptual reasoning and quantitative problem solving. When people changed their answer after talking, it was overwhelmingly from wrong to right. They also spotted a sweet spot: discussion helped most when roughly half the room got the question right at first. Too easy and there’s nothing to argue about. Too hard and there’s nothing to reason with.

Is it real learning, or just copying the smart one?

The obvious objection: maybe the person who knows just tells the person who doesn’t. Scores would go up without anyone learning a thing.

In 2009, Michelle Smith and colleagues at the University of Colorado Boulder tested exactly this in a genetics course of around 350 students, and published the results in Science. Students answered a question alone, discussed it with neighbours, and answered again. Then came the twist: a second, different question testing the same idea, answered alone with no discussion.

Roughly half got the first question right on their own, 68% after talking, and more than 70% on the new question solo. If they’d simply been copying, that last score should have dropped. It went up. And the improvement showed up even in groups where nobody had the right answer to begin with. The conversation wasn’t passing knowledge along. It was building it.

Why does talking it through help you learn?

A more cynical explanation: maybe whoever sounds surest just wins, and confident people happen to be right more often. Psychologists Jonathan Tullis and Robert Goldstone tested this in 2020, following 208 university students across six classes and 86 questions. For each one, students answered alone, rated their confidence, discussed it with a partner, then answered and rated their confidence again.

After discussion, students were both more accurate and more confident. Only 5% of right answers were switched to wrong, while 28% of wrong answers were switched to right. When partners disagreed, 42% of the initially wrong answers flipped to correct.

Here’s the key part. The pairs did better than a simple “go with whoever is more confident” rule would predict. People who started out right were more likely to stick to their answer, and confidence only partly explained that. Being right had a pull of its own.

The researchers describe this as a process gain: the pair produces more than you’d predict from each person alone. One explanation they offer is coherence. When you have to explain an answer out loud and someone pokes at it, a correct explanation tends to hold together and a wrong one tends to wobble. You can often hear the difference, even if you couldn’t have produced the right answer yourself.

Why a fellow learner can sometimes explain it better than an expert

There’s another reason the person next to you can be useful: they still remember what it’s like not to get it. Experts often don’t. Psychologists call this the curse of knowledge: once you know something, it’s surprisingly hard to imagine not knowing it.

A famous demonstration comes from Elizabeth Newton’s 1990 doctoral research at Stanford. She asked people to tap out the rhythm of well-known songs and predict how often listeners would name the tune. Tappers expected about half. Listeners got 3 out of 120, roughly 2.5%. The tappers could hear the melody in their heads. The listeners just heard knocking.

A friend who was confused by the same thing last week is far less likely to be tapping a song only they can hear. They know where the tricky bit is, because they just tripped over it.

When doesn’t learning with someone else work?

Peer discussion isn’t magic. If you both hold the same wrong idea, talking can simply make you both surer of it. It works best on questions that need reasoning, not on facts you either know or don’t: no amount of chatting will produce the capital of Burkina Faso if neither of you knows it. And most of this research comes from university classrooms, so applying it to everyday life is a sensible extrapolation rather than a tested guarantee.

How to learn with a partner: try this

1. Think alone first. Before you talk, each of you jot down your answer and how sure you are. This stops you both drifting towards whoever speaks first.
2. Pick questions with room to disagree. “What is this graph actually showing?” beats “What year did this happen?”. Aim for things you each half-understand.
3. Explain, don’t announce. Say why you think what you think. “I’m pretty sure” doesn’t count as a reason.
4. Poke holes kindly. Ask “How would that work if…?”. Wrong explanations tend to wobble under gentle pressure.
5. Swap who goes first. Confident talkers dominate. Take turns leading.
6. Test yourself on a fresh example. Like the Colorado students, try a new version of the problem on your own afterwards. If you can do it solo, you’ve learned it.

Where Everconnected fits

If you’d like to try this with someone new, Everconnected matches you with people who share your values and interests, for one-to-one conversations by voice. It’s free to join: create an account at app.everconnected.com, or get the app from the App Store or Google Play.

Questions people ask

Sources

1. Crouch, C. H., & Mazur, E. (2001). Peer Instruction: Ten years of experience and results. American Journal of Physics, 69(9), 970-977. doi.org/10.1119/1.1374249
2. Smith, M. K., Wood, W. B., Adams, W. K., Wieman, C., Knight, J. K., Guild, N., & Su, T. T. (2009). Why peer discussion improves student performance on in-class concept questions. Science, 323(5910), 122-124. doi.org/10.1126/science.1165919
3. Tullis, J. G., & Goldstone, R. L. (2020). Why does peer instruction benefit student learning? Cognitive Research: Principles and Implications, 5, 15. doi.org/10.1186/s41235-020-00218-5
4. Kober, N. (2015). Reaching Students: What Research Says About Effective Instruction in Undergraduate Science and Engineering. National Academies Press (account of Mazur’s discovery via Halloun and Hestenes’ work). www.nationalacademies.org/read/18687/chapter/2
5. University of Colorado Boulder (2009). Peer discussion improves student performance with ‘clickers’. ScienceDaily, 2 January 2009 (summary figures for Smith et al.). sciencedaily.com/releases/2009/01/090102100234.htm
6. Newton, E. L. (1990). Overconfidence in the communication of intent: Heard and unheard melodies. Doctoral dissertation, Stanford University. Described in: The Psychologist (British Psychological Society), “Cursed knowledge”. bps.org.uk/psychologist/cursed-knowledge