In 1990, a Stanford graduate student named Elizabeth Newton ran an experiment so simple it barely sounds like science. She paired people up: one played the “tapper,” the other the “listener.” The tapper got a list of twenty-five songs everybody knows — “Happy Birthday to You,” “The Star-Spangled Banner” — picked one, and knocked out its rhythm on the table with a finger. The listener had to name the tune from the knocking alone. Before each round, Newton asked the tapper a single question: what are the odds your partner gets it?
Fifty percent, the tappers said. A coin flip. That was the average prediction across 120 taps.
The listeners got three right. Two and a half percent, one time in forty. And here is the detail that has stayed with me since I first read it: 2.5 percent sat below the entire range of tapper predictions. The most pessimistic person in the study guessed 10 percent. Everyone was at least four times too confident.
🧠 The Song Only You Can Hear
The tappers weren’t lying or showing off. While they knocked out the rhythm, they were hearing the whole song in their heads — melody, lyrics, the horn section — and their fingers were keeping time with that orchestra. What actually crossed the table was a run of disconnected knocks, like a strange Morse code.
The tappers could not hear the version they were sending. That is what makes this a curse rather than a mistake: it isn’t a gap in knowledge, it’s knowledge itself doing the damage. Once you know the song, you can’t un-hear it, and you can’t imagine hearing only knocks.
Economists named the phenomenon a year earlier. In 1989, Colin Camerer, George Loewenstein, and Martin Weber published the term “curse of knowledge” in a paper meant to puncture a standard economic assumption — the idea that whoever holds more information can reliably predict how the person holding less will judge things. Their test subjects were traders given inside information who were then asked to predict prices as if they didn’t have it. They couldn’t. Even when told exactly what the bias was, they could not switch it off.
🤔 Knowing It Doesn’t Help
Here is where the story turns. You would think learning about the curse would take the edge off it. It doesn’t.
Researchers have tried telling people about the bias. They have tried asking them to think harder about the other person’s perspective. They have tried paying them to be accurate. None of it moved the number. Awareness is not a cure, which makes this bias unusual: most of them yield at least a little when you slow down and focus on them.
There is one more layer of honesty worth adding. A 2014 replication effort on a related version of the effect — a study where people were told that one sister moved another sister’s violin — ran seven experiments with large samples and found the effect less than half the size originally reported. The tapping study itself holds up. The broader claim that we can precisely model this bias in everyday perspective-taking is still being argued about. Real phenomenon, contested measurements.
🔗 The Tapper in the Room
Everyone who makes something for other people taps out songs for a living.
The designer who thinks a button is obviously clickable. The engineer who assumes the error message explains itself. The writer who knows exactly who a character is and is baffled when readers ask. Any product that explains something most people have never learned — a reading, a chart, a rule, a system — is judged by people who only hear knocks.
The curse says the expert’s internal experience makes the worst possible guide. Not because experts are careless, but because the feeling of fluency wears the costume of evidence. “It’s obvious” is a sentence you can only say from inside the song.
There is a reason the design world keeps repeating one rule above all the others: put the thing in front of someone who has never seen it, then watch. It isn’t a courtesy. It’s the only instrument that registers what the tapper can’t.
🎲 Try It at Dinner
Pick a song everyone knows. Tap the rhythm on the table and have someone guess. Before they answer, say out loud how confident you are.
The predictions don’t shift much. Fifty percent, usually. The hit rate doesn’t either.
Then try it twice: once with a song you love, once with a song you barely know. That second round is a small taste of the listener’s side of the table — you hear the knocks, and nothing else.