Here’s a small experiment you can run on yourself. Pick something you use every day — a zipper, a flushing toilet, a ballpoint pen — and rate on a scale from one to seven how well you understand how it works. One means you have no idea. Seven means you could confidently explain it to someone else in detail.
Most people, for most everyday objects, rate themselves around four or five. A reasonable sense of how this stuff works, give or take.
Now, for whatever you just rated, try actually explaining it. Step by step, how does a zipper close? What is the mechanism by which a flushing toilet empties and refills? What makes the ink come out of a ballpoint pen when you press it down and stop when you don’t?
If you’re honest, you’ll notice something uncomfortable. The explanations either don’t come, or they come in fragments, or they reach their limits embarrassingly fast. And if you re-rate your original understanding after trying to explain it out loud, most people drop by one or two points. What felt like understanding turned out to be, on closer inspection, more like recognition — the fuzzy sense that you know how something works, which you didn’t.
This gap has a name. It’s called the illusion of explanatory depth, and it was documented in the early 2000s by two psychologists named Leonid Rozenblit and Frank Keil at Yale. Their finding has turned out to be one of the more unnerving in cognitive psychology, partly because it generalises well beyond zippers.
What Rozenblit and Keil found
The original studies were disarmingly simple. Rozenblit and Keil would ask participants to rate their understanding of everyday objects. Then they’d ask them to write out detailed explanations of how the objects worked. Then they’d ask participants to re-rate their understanding.
The ratings dropped. Not slightly — substantially. People who had felt confident they understood zippers, helicopters, ballpoint pens or piano keys discovered, when actually asked to explain, that their understanding was thinner than they’d thought. The exercise of explanation collapsed the felt sense of knowing into the narrower reality of what could actually be articulated.
What made the finding striking was that the same pattern didn’t hold for other kinds of knowledge. If participants were asked to rate their knowledge of historical facts (the year of an event) or procedural skills (how to tie a shoe), the ratings held steady after testing. The illusion was specific to what researchers call explanatory knowledge — the sense of understanding why or how something works. Rozenblit and Keil’s interpretation was that people tend to confuse familiarity with an object or concept for understanding of it. Because you’ve seen zippers, used zippers, had zippers work and not work, it feels like you know zippers. But knowing how to use something and knowing how it works are different kinds of knowledge, and we regularly confuse them.
Why this extends far past zippers
The unsettling thing is that the same phenomenon applies to almost everything we claim to understand.
Ask yourself how well you understand how the electoral system of your country works. For most people, the confident answer is pretty well. Now try to explain, step by step, how a vote you cast actually translates into a person sitting in a legislature. Most people, doing this honestly, quickly discover they know the broad outlines but not the details. The system they thought they understood is, in fact, something they mostly trust others have explained correctly somewhere.
The same holds for economics (how does inflation actually reduce the value of savings?), for climate science (what exactly is the greenhouse effect, and why do some gases cause it and others not?), for biology (how does your immune system distinguish between a cold virus and a harmless protein?), for law (what would actually happen if you were sued tomorrow?). In each case, most of us carry around a sense of understanding that evaporates on request.
This wouldn’t matter if the illusion stayed private. But it doesn’t. In 2013, two political scientists — Philip Fernbach and Steven Sloman, building on Rozenblit and Keil’s work — ran an extension. They asked American participants to rate how well they understood six policy proposals (e.g., a carbon emissions cap, sanctions against Iran). Then they asked participants to explain the policies in detail. Then they asked them to re-rate their understanding, and also to rate how strongly they supported or opposed each policy.
Two findings emerged. People’s understanding ratings dropped, as expected. But their position ratings also moderated. Having discovered, through the exercise of explanation, that they didn’t really understand the policy as well as they’d thought, participants reported less extreme views on it. The illusion of explanatory depth was partly responsible for how confidently people had held their positions.
The educational version
A related phenomenon comes from research on learning. Robert Bjork, a cognitive psychologist at UCLA, has spent decades studying what he calls illusions of competence.
Bjork’s key finding is that many of the study techniques students prefer produce a feeling of learning without the reality. Re-reading a chapter, for example, feels productive — the material becomes familiar, the concepts feel increasingly known. But test performance after re-reading is often weak. Testing yourself, by contrast, feels harder and less productive in the moment but produces much stronger retention.
The student re-reading a chapter is, in Bjork’s framework, experiencing the illusion of competence. The familiarity is real; the learning is not. The student who closes the book and tries to remember the key points, by contrast, confronts the actual state of their knowledge — which is often humbler than the familiarity had suggested — and builds real learning from that confrontation.
This explains a puzzle many teachers have observed. Students who feel most confident before an exam are not reliably the ones who do best. The most confident students are often those who spent time re-reading and highlighting — activities that produce fluency and familiarity, which feel like understanding. Students who spent the same time on flash cards or practice problems feel less confident but usually perform better. The confidence tracks the felt fluency, not the actual mastery.
The expert version of the same illusion
There’s one more twist worth knowing about, from the political scientist Philip Tetlock, whose decades of research on expert forecasting revealed a surprising pattern. Subject-matter experts — foreign policy analysts, economists, political scientists — are, on average, not much better than chance at predicting outcomes in their own fields. In some studies, they performed slightly worse than naive algorithms that just assumed the future would look like the recent past.
Tetlock’s interpretation is subtle. Experts have, generally, escaped the illusion of explanatory depth about their own field — they usually can explain how things work. But they’ve replaced it with a related illusion, the illusion of forecasting. Because they understand their field deeply, they feel they should be able to predict its movements. The feeling is misleading. Understanding how a complex system works and predicting what a complex system will do are different problems, and expertise in the first doesn’t reliably produce skill at the second.
This has implications for how we should listen to experts. Their explanations of how their field works are usually excellent. Their predictions about what their field will do next are usually unreliable. The confident expert predicting the future is in roughly the same epistemic position as you, trying to explain how a zipper works. The feeling of confidence is real. The accuracy behind it is often less than the feeling would suggest.
The counter-view: sometimes shallow is enough
A reasonable counter-argument, advanced by several researchers, is that the illusion of explanatory depth isn’t always a problem. Humans live in a world too complex to understand in detail. You don’t need to understand how a zipper works to use one. You don’t need to understand the physics of engine combustion to drive a car. You don’t need to understand immunology to decide whether to vaccinate. In most practical domains, shallow understanding is adequate — and demanding deep understanding would paralyse everyday life.
The cognitive scientist Hugo Mercier has made a version of this argument. Much of human knowledge is distributed — held across networks of experts, institutions and tools, rather than in any individual mind. We function by knowing enough to identify whom to trust and when to seek more information. This isn’t intellectual failure; it’s a rational division of cognitive labour.
So the honest framing is not that the illusion of knowing is always bad. Shallow understanding of most things is fine. The problem arises specifically when shallow understanding produces overconfident conclusions — particularly in domains where one’s conclusions affect one’s decisions or others’ lives. Voting without really understanding a policy. Investing based on a feeling that you grasp a market. Advising a friend on a medical question you’ve only half-read about. These are the situations where the illusion of knowing turns from harmless to costly.
How to notice it in yourself
The most useful practice from this research is simple. Before committing to a strong opinion on something, try explaining it. Not in your head — out loud, or in writing, to someone who will ask clarifying questions. Does the explanation flow? Does it break down at the second or third layer? Where do you find yourself saying “and then it sort of just works”?
The places where the explanation breaks down are the places where your understanding is shallower than it felt. This isn’t an indictment. It’s information. It lets you decide whether to go deeper, or to moderate your confidence, or to acknowledge that on this question, you should probably listen more than you speak.
A second practice, drawn from Bjork’s research, is to test yourself rather than familiarise yourself. When you want to know whether you really understand something, the reliable move isn’t to re-read it. It’s to close the book, look away, and try to reproduce the argument from memory. What you can reproduce is yours. What you can’t reproduce was only ever borrowed.
The question that remains
The illusion of knowing is, in the end, one of the most human things about us. Familiarity feels like understanding. Recognition feels like knowledge. Fluency feels like mastery. These impressions aren’t malicious or lazy — they’re the natural outputs of minds that had to make quick sense of the world without infinite time to check.
But the gap between what you know and what you feel you know is real, and knowing it exists is the beginning of a more honest kind of intellectual life.
The question worth carrying is one the Rozenblit and Keil research implicitly asks about everything:
Of the things you currently hold strong opinions about, how many could you actually explain, step by step, well enough that an intelligent stranger would be satisfied — and how many are you just familiar with?
Key research referenced: Leonid Rozenblit and Frank Keil’s illusion of explanatory depth research (Cognitive Science, 2002); Philip Fernbach and Steven Sloman’s political extension (Psychological Science, 2013); Robert Bjork’s research on illusions of competence; Philip Tetlock’s expert-forecasting research (Expert Political Judgment, 2005).