9 min read

The Feeling of Knowing

metacognitionlearning sciencenorudit

You shut the book after a long session and something quietly settles: you've got this. Then the exam asks the question and the certainty is just gone, and you're sitting there looking at a blank you were completely sure you could fill. This post is about that gap, why the feeling lied, and the three tests that close it before an exam gets to close it for you.

It happens to everybody and it isn't a character flaw. The feeling of knowing is a real sensation your mind produces, and the problem is that most of the time it's measuring the wrong thing entirely.

What rereading is actually training

Think about what the third pass through a chapter feels like compared to the first. First time the words are stiff, you're stopping, going back, rereading sentences. By the third time it glides. You recognise the shape of a paragraph before you've finished it, your eyes move faster, nothing catches you out.

That smoothness feels exactly like learning. What's actually got easier, though, is reading the page, and reading the page and retrieving the idea are two different skills. Practising the first one does remarkably little for the second. This has a name, the illusion of competence, and it's the habit of treating the ease of taking words in as evidence you could get them back out again (Koriat and Bjork, 2005).

Highlighting is the same thing wearing a brighter coat. Dragging a yellow line across a sentence feels like doing something, and it is doing something, it's marking. It just isn't remembering. Then you come back to the page later and the colour does the recognising on your behalf, and that little flash of familiarity gets filed as mastery.

The core mistake

We confuse recognising the page with knowing the idea. The page feels easy because we've seen it before, not because we could rebuild it from nothing.

Which explains something that otherwise looks perverse: the study methods that feel best are consistently the ones that work worst. Rereading is comfortable. Highlighting is soothing. Sitting there trying to drag an answer out of your own head is unpleasant, and the unpleasantness is the part that's working.

Your confidence and your actual knowledge come apart

If the feeling of knowing tracked real knowing, none of this would matter. You'd trust your gut, study until it felt done, and be right. But the two come apart, and they come apart in one predictable direction, which is that we consistently think we know more than we can actually produce.

Ask people to predict how much they'll remember and then properly test them, and the prediction runs high, sometimes a long way high (Koriat, 1997). And it gets worse under exactly the conditions rereading sets up. In one study, students who kept restudying material got steadily more confident they'd remember it, then a week later remembered less than the students who'd been testing themselves instead. All those extra passes moved the confidence and left the knowledge where it was (Roediger and Karpicke, 2006).

Left to pick for themselves, people also take the comfortable route, dropping a topic at the point it starts feeling familiar rather than the point it's actually stuck (Kornell and Bjork, 2007). So the illusion isn't just inaccurate, it steers. It sends you back toward the material you've already got and away from the gaps that are going to cost you marks.

And that shows up in results. Students who are overconfident about what they've learned stop studying too early, call it done while it's still half built, and score lower for it (Dunlosky and Rawson, 2012). Overconfidence isn't a harmless bit of self-belief, it's the feeling of having finished turning up before the work has.

The illusion of competence doesn't just misjudge. It steers you toward what you've already got and away from the gaps that will cost you.

So what do you trust instead?

Not another feeling. You trust what you can produce.

There's really only one way to find out whether something is in your head, and it's to try and get it out with nothing at all in front of you. No page, no highlight, no multiple-choice list quietly holding the answer up where you can see it. Can you say it in plain words to someone who doesn't already know it? Can you rebuild it from a blank page? Can you still use it next week when the question is worded in a way you've never seen before?

Those are three separate questions and they fail in three separate ways. You can understand something and not remember it. You can remember it word for word and not understand a bit of it. You can have both today and have lost them by next Thursday. The illusion can hide in whichever of those gaps you left open, which is why one test was never going to be enough.

The three tests, in order

This chain is the one part of Norudit that isn't optional. Everything else you can wander through in whatever order suits the day. This you can't skip, because it sits precisely where "I've seen it" gets mistaken for "I know it".

First, explain it simply. This is the Feynman gate, named after Richard Feynman and the line usually attributed to him about not understanding what you can't explain simply. The gate makes you teach the idea back in your own plain words, and it's where the illusion usually dies, because plain words are extremely hard to fake. Memorised phrasing doesn't survive being translated. The moment you have to rebuild an idea in language a beginner could follow, either the understanding is there or the sentence comes apart in your hands halfway through, and that collapse is the illusion getting caught in the act. It has to be a genuine explanation rather than a smooth-sounding one, which is a trap worth its own post. This comes first because everything after it assumes the model in your head is correct, and drilling a misunderstood idea just makes the misunderstanding more durable.

Second, rebuild it from nothing. Passing the gate proves you understand it today. It doesn't prove a memory formed, and understanding you can't retrieve isn't much use to you in an exam hall. So once the gate opens the blank page starts: material closed, whole topic reconstructed, no prompts and no cues. There's nowhere for the illusion to hide here because there's no page in front of you to feel smooth about. What comes back is what you know, and what doesn't come back is where you go next.

Third, keep it while the wording keeps moving. Understanding and memory can both be genuinely real today and quietly rot by next week, and ordinary flashcards let a completely new illusion in through the side door: answer the same fixed card enough times and you stop knowing the concept and start recognising the card. So Norudit rewrites the card every time you answer it. Concepts come back reworded, facts come back as questions that make you use them. You never meet the same surface twice, so there's nothing available to recognise, and the only thing that can survive round after round is the idea itself.

The whole thing in a line

Understand it, prove the memory formed, then keep it while the wording keeps changing. Three tests in that order, and the illusion of competence has nowhere left to sit.

Understanding, then memory, then durability, and each one is fairly meaningless without the one before it. It's the same trio the whole learning sequence is built around, just looked at from the inside.

Two things that make those three trustworthy

The first is measuring instead of guessing. Behind all of it Norudit keeps a running estimate of how well you know each individual concept, updated every time you answer anything. The method is Bayesian knowledge tracing, which works out the odds that a skill is genuinely learned from your pattern of right and wrong answers rather than from how confident you claim to be (Corbett and Anderson, 1994). It never asks how ready you feel. It watches what you produce and estimates from that, then brings your weakest concepts back on its own, so you're not relying on willpower to make yourself pick the hard material. That estimate is the map the rest of the personalisation gets drawn on.

The second is grading that doesn't go soft on you. None of the three tests is worth much if the marking gives you the benefit of the doubt, because a generous grader just re-inflates the exact confidence the blank page was there to deflate. So the grading reflects what you actually wrote rather than what you meant to write. That stings sometimes, and the sting is doing a job.

Next time a session ends with that warm certainty, treat it as a hypothesis rather than a verdict. Shut the book and try saying the whole thing out loud, plainly, to nobody in particular. If it comes out, good, you knew it. If it doesn't, also good, because you've found that out here rather than in the exam hall.

Start your first class →

Sources

Read more