The Curse of Knowledge: The Neuroscience of Why You Can't Explain Your Own Product

In 1990, a Stanford doctoral student named Elizabeth Newton designed an experiment so simple it almost didn't look like science. She divided participants into two groups: tappers and listeners. Each tapper received one of twenty-five well-known songs ("Happy Birthday," "The Star-Spangled Banner," "Mary Had a Little Lamb") and tapped out the rhythm on a table. The listener's job was to identify the song.

Before the guessing began, Newton asked the tappers to predict how often the listeners would get it right. The tappers predicted fifty percent. They figured the songs were so obvious, so universally known, that at least half the time the rhythm alone would be enough.

The actual success rate was 2.5 percent. Three correct guesses out of 120 attempts.

The tappers were stunned. Some were visibly frustrated with the listeners, as if the listeners were being obtuse on purpose. Newton later documented the tappers' reactions: they would tap out "The Star-Spangled Banner" and look at the listener with an expression that said, How can you not hear this?

But that reaction was itself the finding. The tappers couldn't help hearing the full melody in their heads while they tapped. They experienced a rich, layered song: horns, lyrics, emotional crescendos. The listeners heard a series of arrhythmic knocks on a table. The tappers were not bad at tapping. They were incapable of imagining what the tapping sounded like without the song. Once they knew the melody, they could not unknow it.

Newton had demonstrated, in the space of a university classroom, the single most damaging cognitive bias in business communication. And if you've ever watched a potential customer stare blankly at your landing page, or a promising investor's eyes glaze over halfway through your pitch, or a friend politely nod when you explain what your company does while clearly understanding none of it, you've lived the experiment from the wrong side of the table.

You are the tapper. And you cannot hear how bad the tapping sounds.

What Is the Curse of Knowledge?

A year before Newton's tappers sat down at Stanford, three economists published a paper that gave this phenomenon its name. Colin Camerer, George Loewenstein, and Martin Weber authored "The Curse of Knowledge in Economic Settings" in the Journal of Political Economy in 1989. Their work built on Baruch Fischhoff's 1975 research on hindsight bias, the finding that once people know the outcome of an event, they can't accurately reconstruct their prior state of uncertainty. Fischhoff showed that knowing the ending makes the ending feel inevitable, even to the person who didn't see it coming.

Camerer, Loewenstein, and Weber extended this into business decision-making. They designed experiments in which informed participants were asked to predict the judgments of less-informed participants. The question was simple: can people who know more accurately predict how people who know less will respond?

The answer was no. And it wasn't close.

In one experiment, participants who knew the value of a company's assets were asked to predict what price an uninformed buyer would offer. The informed participants consistently overestimated what the uninformed buyer would pay. They couldn't set aside their own knowledge. When the assets were worth more, they expected higher offers, even though the buyer had no way to assess the true value. Their brains were defaulting to a projection: I know this, therefore you probably know something close to this.

The researchers called this the curse of knowledge. Their most striking finding was about what doesn't fix it. Incentives (real money for accurate predictions) shrank the curse slightly but didn't eliminate it. Market settings with multiple rounds of feedback reduced it by about fifty percent. It never vanished entirely. Even when money was on the line, the brain's default was to project its own knowledge onto others.

Chip and Dan Heath later elevated this research to a mainstream audience in their 2007 book Made to Stick. They called the curse of knowledge "the single biggest problem in communication" and used Newton's tapper experiment as the book's central metaphor. Every CEO communicating with frontline employees, every teacher explaining to students, every marketer writing for customers, all of them were tappers, hearing the melody while their audience heard only knocks.

The metaphor is memorable. But it understates the problem. The curse of knowledge has a neurological basis, it gets worse with expertise, and founders are the most cursed people in any room.

Why Founders Are the Most Cursed People in the Room

Here is a pattern that repeats across nearly every early-stage company. The founder spends months, sometimes years, immersed in a problem. They talk to hundreds of potential customers. They read every competitor's documentation. They build prototypes, tear them down, rebuild them. They learn the market's vocabulary, its economics, its hidden pain points. By the time they sit down to write a landing page or rehearse a pitch, they carry a mental model so detailed and interconnected that it feels like common sense.

Then they try to explain the product to a stranger, and the stranger has no idea what they're talking about.

Not because the stranger is unintelligent. Because the founder has forgotten what it was like to not know. The jargon that feels precise sounds like noise to an outsider. The problem that feels urgent after six months of customer interviews feels abstract to someone encountering it for the first time. The founder taps "The Star-Spangled Banner." The customer hears knocks.

This is why so many startup landing pages read like they were written for the company's own engineers. "We leverage AI-driven predictive analytics to optimize cross-functional workflow integration." The founder reads that and hears a symphony. The visitor reads it and bounces. The visitor doesn't leave because the product is bad. They leave because the words on the screen activated no mental model they could connect to.

The curse intensifies with time. A first-time founder explaining their product at week two is less cursed than the same founder at month eight. By year two, the curse is so deep that the founder often can't identify which parts of their explanation assume prior knowledge, because all of it feels like basic vocabulary.

This is also why investor pitches fail at the rates they do. DocSend's analysis of fundraising decks shows that investors spend an average of two minutes and thirty seconds reviewing a pitch deck. In that window, the founder needs to transmit not just information, but context. The curse of knowledge systematically strips context out. The founder assumes the investor understands the problem, the market, the competitive terrain. The investor sees assertions with no scaffolding. The pitch doesn't fail because the idea is bad. It fails because the translation is bad.

What Happens in the Brain When You Can't Take Another Perspective

The curse of knowledge is not a character flaw. It's a design feature of neural architecture. And understanding the brain mechanism explains why the curse is so hard to break, and what it actually takes to override it.

When the brain processes other people's mental states (what cognitive scientists call "theory of mind"), it relies heavily on a region called the temporoparietal junction, or TPJ. Rebecca Saxe at MIT has spent over two decades mapping this region using fMRI, and her work has established the right TPJ as the brain's dedicated hardware for thinking about other people's thoughts. In a classic experiment, participants read stories requiring them to reason about another character's false belief, predicting where Sally will look for a marble she doesn't know has been moved. The right TPJ activates specifically when the task requires holding two knowledge states at once: what you know and what someone else knows.

But here is the critical finding. Perspective-taking is not the brain's default mode. Nicholas Epley, Boaz Keysar, and their colleagues demonstrated in 2004 that people take others' perspectives through a two-step process: they start from their own perspective (an egocentric anchor) and then adjust outward. The first step is automatic. The second step, the adjustment, requires cognitive effort, and it is almost always insufficient.

The brain's default when estimating what someone else knows is to start with what you know and then try to subtract. The subtraction is consistently too small.

Under time pressure, the adjustment shrinks further. Under cognitive load (when you're stressed, multitasking, or tired), it nearly disappears. A founder rehearsing a pitch the night before a meeting, running on four hours of sleep, is operating with almost no perspective-taking correction. They are fully anchored in their own knowledge. The tapping has never sounded more like a song.

Susan Birch and Paul Bloom confirmed in 2007 that even adults show systematic failures in false-belief tasks when their own knowledge is salient. The curse of knowledge isn't something we outgrow. It's a permanent feature of how the brain models other minds, counteracted only by deliberate effort and external feedback. Left to its own devices, the brain assumes shared knowledge. Every time.

This explains a pattern every startup mentor has observed: the founder who gets worse at explaining their product over time. Their egocentric anchor grows heavier with each week of immersion, and the adjustment required to reach the customer's perspective grows proportionally larger.

The Five Techniques That Actually Break the Curse

If the curse of knowledge were simply a matter of awareness, if knowing about it were enough to fix it, Newton's tappers would have adjusted their predictions. They didn't. Camerer's participants would have corrected when offered cash incentives. They only partially did. The curse resists awareness. It requires structure.

These five techniques work because each one forces an external input that the brain's internal perspective-taking system can't generate on its own.

1. The Mom Test

The name comes from Rob Fitzpatrick's book, but the principle is older than the book. Show your landing page, your pitch deck, or your product description to someone who has no context for what you do. Not a friend in your industry. Not a fellow founder. Your mother. Your neighbor. The barista who makes your coffee. Ask them one question: "What does this company do?"

Then be quiet. Don't correct them. Don't clarify. Don't say "well, what I meant was." Just listen. If they can't explain it back to you in their own words, words you didn't feed them, your messaging has the curse. The gap between what you wrote and what they understood is a direct measurement of how much knowledge you're projecting onto your audience.

This works because it replaces your internal simulation of the audience with an actual audience. Your brain is incapable of accurately simulating what a naive reader experiences. An actual naive reader makes the simulation unnecessary.

2. The Five-Year-Old Rewrite

Take your most important piece of marketing copy (the headline, the first paragraph of your landing page, the opening of your pitch) and rewrite it using only words a five-year-old would understand. Not because your audience is unsophisticated. Because concrete, simple language activates broader neural networks than abstract jargon. George Miller's landmark 1956 paper established that working memory holds roughly four to seven chunks of information. Every piece of jargon that requires the reader to decode it consumes a chunk. By the time they've processed your third technical term, they've run out of working memory for the actual point you were trying to make.

The five-year-old rewrite isn't dumbing down. It's clearing the channel. "We use machine learning to predict churn" becomes "We tell you which customers are about to leave before they leave." The second version contains the same information. It consumes fewer cognitive resources to process. It leaves room in working memory for the listener to think, Do I have that problem?. Which is the thought you actually wanted them to have.

3. The Fresh-Eyes Audit

Designate the first two weeks of every new hire's tenure as a messaging audit period. Have them read every customer-facing piece of content and flag anything that confused them, required a second reading, or used a word they hadn't encountered before joining.

New employees are the closest thing you have to a customer's perspective from inside your organization. Their confusion is data. Within three months, the curse will have claimed them too. The window is brief. Use it while it exists.

4. The Reverse Pitch

After you pitch to anyone (an investor, a customer, a potential partner), ask them to pitch your product back to you. Not to evaluate whether they liked it. To hear what survived the translation. The words they use, the features they emphasize, the problem they describe: this is what actually transferred. Everything they left out is everything the curse ate.

The reverse pitch reveals the delta between your intention and the listener's reception. If you said "AI-powered predictive analytics platform" and they say back "it helps me figure out what's going to break before it breaks," you've just received the copy you should have written in the first place.

5. The Before-and-After Story

Replace your feature description with a concrete story about one customer's experience. A specific person, a specific problem, a specific change. "Sarah ran a bakery and was throwing out $400 of unsold bread every week. She couldn't predict which days would be busy. Now she checks her phone Monday morning and knows exactly how much to bake."

Stories bypass the curse entirely because they carry their own context. The character, the problem, and the resolution provide all the scaffolding a naive listener needs. Narrative doesn't require shared vocabulary. It builds shared experience, the same neural coupling mechanism Uri Hasson documented at Princeton, where a listener's brain synchronizes with the storyteller's.