Why your notes feel easy but the exam does not: The illusion of competence

A biology chapter can make perfect sense while it is open. The diagram labels each step, the bold terms look familiar, and the explanation seems obvious. Close the book and try to describe the same process from memory, though, and the answer may suddenly become much thinner.

That blank is confusing because nothing felt wrong during the study session. The page was easy to follow, so it seemed as if the material had been learned. In reality, the headings, diagrams, and sentences were supplying cues that may not be available when the exam asks for unaided recall or application.

Researchers use illusion of competence for cases in which study conditions make future performance feel more secure than it proves to be. It is one possible error in judging learning, not a claim that confidence is always wrong. Confidence gathered with the answer in view simply needs a better check.

Why familiar material can feel learned

Recognition lets a learner identify previously encountered information among cues or alternatives. Recall requires producing information from a cue or from memory. A course may instead require an explanation, a comparison, or an application, so the check should match that task.

Rereading keeps the source available and does not by itself require recall. A second pass can improve learning, especially when it is spaced or follows a retrieval attempt, but easier processing is not sufficient evidence that the idea will remain accessible later (Roediger and Karpicke, 2006; Rawson and Kintsch, 2005).

In paired-word experiments, Asher Koriat and Robert Bjork found selective overestimation when people judged future recall just after studying an intact pair. Some pairings looked meaningfully connected when both words had just been presented, yet the cue alone did not reliably bring back the target on a test about a minute later. The participants were not uniformly overconfident, and the laboratory task was much narrower than textbook study. The practical warning is narrower too: conditions that help an answer feel accessible during study may not support the task required later.

Highlighting leaves a similar evidence gap when marking replaces a closed-book check. Marking a definition may help organize a later review, but the mark itself does not show whether the definition can be explained with the page closed (Dunlosky and colleagues, 2013).

The problem is the missing test, not the feeling itself

Confidence is useful when it comes from relevant evidence. Solving several problems without hints, explaining a process accurately, or recalling an idea after some time has passed gives confidence a stronger basis. Even then, self-assessment is not infallible, and the practice task needs to resemble what the course will ask (Yang and colleagues, 2022).

The trouble starts when the evidence is only that a page looked clear. Psychologists call the skill of judging one's own learning metacognition. Good metacognition is not constant self-doubt. It is the ability to make a reasonably accurate call about what is secure, what is shaky, and what still needs work.

That accuracy is called calibration. A well-calibrated student might predict an 80 percent quiz score and earn something close to it. Poor calibration can run in either direction: expecting an A and discovering large gaps, or knowing the material but feeling certain that failure is coming.

The goal, then, is not to stop trusting every feeling. It is to compare the feeling with performance often enough that confidence becomes more informative.

Why a smooth study session can mislead

Some methods make a session feel smooth because the material stays visible. Other methods make the learner retrieve, connect, and explain, so the gaps become noticeable. A smooth session can reflect genuine understanding, help from visible cues, or both; current performance is not always evidence of learning that will last (Soderstrom and Bjork, 2015).

That difference between a smooth session and lasting learning has appeared in controlled studies. Jeffrey Karpicke and Janell Blunt asked students to study science texts with retrieval practice or concept mapping. Students made concept maps while viewing the text; the retrieval condition involved free recall, restudy, and another recall, with study time matched. Students underestimated retrieval practice: in the first experiment they predicted repeated study would work best, and in the second most expected concept mapping to work at least as well as retrieval. On tests given a week later, the retrieval-practice groups remembered more and performed better on questions that required inference as well as direct recall.

The result does not prove that concept maps are always weak or that retrieval wins in every course. A 2025 meta-analysis of retrieval and elaborative study found only a small average retrieval advantage, with feedback as an important moderator. Retrieval can expose uncertainty and strengthen later access, but it should include corrective feedback and can be combined with relevant explanation or elaboration.

Research on interleaving makes the same point in a different setting. In a study of artists' painting styles, participants generally thought studying one artist at a time worked at least as well, even though mixing, and therefore spacing, each artist's examples improved classification of new paintings on a test that began after a 15-second distractor. A later meta-analysis found that interleaving effects differed sharply by material, so the painters result is not a rule for every mixed study session.

A five-step check for what you actually know

The most useful correction is a short study loop that produces evidence. It can work with a practice problem, a flashcard, a blank page, or an open-ended question.

1. Decide what the exam will ask you to do

"Know chapter six" is too vague to test. A better target might be:

• explain how an action potential travels down an axon
• compare mitosis with meiosis
• calculate acceleration from a velocity graph
• use a case's facts to identify the relevant legal rule

The form matters. Recalling a definition does not establish that a new problem can be solved, and solving one familiar problem does not establish that the underlying idea can be explained.

2. Remove the helpful cues

Close the note, hide the answer, or move to a blank page. Then attempt the same kind of response the exam will require.

A complete answer is not required on the first attempt. The point is to make the gap visible while there is still time to fix it. If nothing comes to mind, that is useful information, not a verdict on intelligence.

3. Compare the attempt with a trustworthy source

Open the lecture note, textbook, answer key, or instructor-provided material and check the response. Look for three things:

• accuracy: Which statements were correct or incorrect?
• completeness: Which important steps or conditions were missing?
• certainty: Which parts were guesses, even if they happened to be right?

This comparison is what turns retrieval into feedback. Repeating an incorrect answer without checking it can reinforce the error.

4. Try again in a relevant form

After correcting the answer, attempt it again without looking. Later, return to it after other material has intervened, or use a different question that requires the same concept.

No single delay works for every exam or every topic. The important change is that the answer is tested beyond the moment when it has just been read. A later attempt gives a better sample of whether the idea remains accessible.

5. Record the result, not a mood

A simple calibration log can be enough:

After a few rounds, patterns become visible. Maybe high confidence is usually accurate for definitions but unreliable for multi-step problems. Maybe low confidence reflects anxiety rather than weak recall. Either result is more useful than assuming every feeling means the same thing.

How to run the check in Bananote

Bananote accepts a lecture recording, supported audio or video, a PDF, pasted text, or a YouTube link and can generate a note, flashcards, and a practice quiz from that source. That automates the initial flashcard and quiz setup, but the generated material still needs to be treated as a study aid rather than an unquestionable answer key.

Start with one note that has been checked against the original source. Then use its flashcards or quiz with the note closed. A quiz score is one sample from one set of questions, so use the missed and uncertain answers to choose what to review instead of treating the number as a complete measure of the course.

For an open-ended check, Bananote's chat can ask questions about the note. A useful prompt is:

> Ask me one open-ended question at a time about this note. Wait for my answer. Then identify what I explained accurately, what I missed, and what needs verification against the note. Do not show the answer before I attempt it.

This prompt asks for an attempt before more explanation, but check whether the chat followed the instructions and verify its feedback against the note and original source.

The blurting method offers a lower-tech version of the same check: write everything remembered about one topic, then compare the page with the note. Trying to explain a concept in your own words can also reveal where the explanation breaks down; the Feynman technique guide shows one way to structure that attempt.

What better calibration sounds like

Imagine an anatomy exam is two weeks away. "I've looked through every lecture twice" says what happened during studying, but it does not say what is ready for the exam.

A calibrated update sounds more like this:

> I can label the heart's chambers and trace normal blood flow without the diagram. I keep mixing up two valve disorders in case questions, and I cannot yet explain why their murmurs occur at different points in the cardiac cycle.

That description is useful because it names both the secure knowledge and the next task. It also leaves room for revision. After more practice and feedback, the same student may discover that the explanation is now solid but a new type of case still causes trouble.

Frequently asked questions

What is the illusion of competence?

It is the mismatch between feeling that material has been learned and being able to recall or use it under the conditions that matter. The feeling often develops when helpful cues, such as the open textbook or visible answer, make information easy to recognize during study.

Does rereading never work?

Rereading can help with orientation, clarification, or reviewing a passage after a failed attempt. The limitation is that rereading alone gives little evidence about unaided recall or application. Pair it with a closed-source attempt that matches the kind of response the course requires.

Is highlighting useless?

Highlighting can mark structure or flag material for later review, but covering a page in color does not test learning. The highlighted material still needs to be recalled, explained, or applied without the page providing the answer.

How do I know if I actually know something?

Choose a task that resembles the exam, attempt it without notes, compare the response with a reliable source, and try again after feedback. Several attempts across relevant questions give a better estimate than one feeling or one quiz score.

Should studying feel difficult to work?

Not automatically. Difficulty helps only when it serves the learning goal and the learner can make a useful attempt with feedback. Confusing instructions, missing background knowledge, and impossible questions can feel hard without producing better learning (Bjork, Dunlosky, and Kornell, 2013).

---

Familiarity is a useful first signal, but it cannot finish the diagnosis. The next time a chapter feels easy, close it and ask for the kind of answer the exam will require.

Try Bananote to turn one checked lecture note into flashcards and practice questions, then use the results to decide what needs another pass.

Sources

• Koriat and Bjork: Illusions of competence in monitoring one's knowledge during study (2005)
• Karpicke and Blunt: Retrieval practice produces more learning than elaborative studying with concept mapping (2011)
• Kornell and Bjork: Learning concepts and categories (2008)
• Roediger and Karpicke: Test-enhanced learning (2006)
• Rawson and Kintsch: Rereading effects depend on lag and test delay (2005)
• Bisra and colleagues: Inducing self-explanation (2018)
• Dunlosky and colleagues: Improving students' learning with effective learning techniques (2013)
• Soderstrom and Bjork: Learning versus performance (2015)
• Yang and colleagues: Improving students' text metacomprehension accuracy (2022)
• Goncalves, Muniz, and Jaeger: Retrieval practice versus elaborative encoding (2025)
• Brunmair and Richter: Interleaving meta-analysis (2019)