How Misconception Repair Works in Teaching | Surface the Wrong Model, Build a Better One and Prove the Old Idea No Longer Wins

A learner says that one-fifth is larger than one-half because five is larger than two.

The answer is wrong, but the important fact is not the wrong answer. The learner has used a rule that works perfectly well for whole numbers and carried it into a number system where the relation has changed. The response has an internal logic. Correcting the answer without correcting that logic may leave the wrong model ready to return on the next unfamiliar question.

This is what makes a misconception different from a random mistake.

Misconception repair works in teaching when the teacher identifies the learner’s active wrong model, creates evidence that exposes where that model fails, supplies a more accurate model that explains both the familiar and the troublesome case, and then verifies that the old model no longer controls later decisions.

The work is therefore not “tell the correct answer more clearly.” It is model replacement, model reorganisation or model boundary repair.

This article continues the eduKate Sengkang How Teaching Works series after How Feedback Works in Teaching. It owns the general teacher-side mechanism for misconception repair.

It does not replace the learner-side MindOS Refutation State, the subject-specific PSLE Science misconception-repair guide, or How to Improve Learning From Mistakes. Those pages solve narrower learner, subject and improvement jobs. This article concentrates on what the teacher has to detect, design, explain and verify.

The cases below are original teaching examples unless a source is explicitly identified. They are not records of actual students, official examination questions or experimental outcomes.

A route through misconception repair

Begin with what counts as a misconception, then move through diagnosis, surfacing the learner’s model, contrast and counterexamples, building a replacement model, and verification across transfer and delay. Worked cases cover fractions, algebra, science and English. Later sections address whole-class teaching, small groups, AI, parents, failure modes and evidence limits.

A misconception is a wrong model, not merely a wrong answer

A learner can be wrong without holding a misconception. The learner may misread a number, forget a fact, lose a sign, rush, guess, or make a one-off execution error. Those states can produce the same visible result as a misconception but need different teaching.

A misconception is more structured. The learner applies an idea, relationship or rule that feels coherent and often works in some familiar situations. The trouble appears when the model is incomplete, overgeneralised, mapped onto the wrong domain or built on a false causal mechanism.

Examples include:

• “A larger denominator means a larger fraction because the number is bigger.”
• “Multiplication always makes numbers larger.”
• “Anything heavier falls faster.”
• “Plants get their food from the soil.”
• “A lid stops heat from escaping completely.”
• “The main idea is always the first sentence.”
• “A longer answer earns more marks because it contains more information.”
• “If two variables move together, one must cause the other.”

Each statement contains an organising rule. That rule can generate many future answers. Repair therefore needs to operate at the rule or model level.

Do not diagnose a misconception from one wrong answer

A single response is evidence, not direct access to the learner’s mental model. The learner who writes one-half as smaller than one-fifth may have misunderstood fraction size, may have reversed the comparison sign, may have rushed, or may simply have guessed.

The teacher needs at least one discriminating check. Ask for a reason. Change the numbers. Ask for a drawing. Present a case where the suspected rule predicts the opposite result. Confidence can also be useful when compared with reasoning.

This is the same disciplined uncertainty used in How Diagnostic Teaching Works: form a working hypothesis, test it, then update it.

Wrong knowledge, missing knowledge and inaccessible knowledge are different states

Missing knowledge means the relevant idea is not yet available. The learner may need explicit teaching.

Weak retrieval means the learner may know the idea but cannot access it reliably. Retrieval practice may be more useful than another explanation.

Misconception means another model is actively competing with the correct one. Simply adding correct information may not be enough because the wrong model still explains the learner’s experience.

These states can coexist. A learner may have a misconception because necessary background knowledge was missing when the original idea formed. A learner may understand the correct model but fail to retrieve it quickly, allowing the older intuitive model to win under time pressure.

Misconceptions are often rational extensions of prior knowledge

Many misconceptions are not irrational. They are reasonable inferences from a learner’s prior experience.

Whole-number knowledge teaches that five is greater than two, so “one-fifth is greater than one-half” can feel natural before the learner understands that the denominator names the size of equal parts. Everyday experience can suggest that heavier things fall faster because air resistance, shape and materials vary together in uncontrolled ways. A child sees plants rooted in soil and water added to soil, so “soil is the plant’s food” can fit the visible story.

This matters because contempt is diagnostically useless. The teacher needs to understand why the wrong model feels plausible before designing a better one.

EEF’s discussion of prior knowledge and pupil misconceptions makes a similar point: teachers need to consider how existing knowledge shapes interpretation and how common misconceptions may arise. The teaching task is to build on prior knowledge without allowing an inappropriate rule to dominate a new domain.

The misconception-repair cycle

1. Notice: identify a response pattern worth investigating.
2. Elicit: ask the learner to predict, explain or represent the idea before correction.
3. Differentiate: separate misconception from slip, missing knowledge, retrieval failure or task misreading.
4. Name the model: state the learner’s current rule precisely enough to test it.
5. Find the boundary: identify where the rule works and where it fails.
6. Design a contrast: choose an example, non-example or counterexample that reveals the failure clearly.
7. Protect reasoning: let the learner commit to a prediction before revealing the result where appropriate.
8. Explain the conflict: make clear why the old model cannot account for the evidence.
9. Build the replacement: teach a more accurate model, mechanism or conditional rule.
10. Reconstruct: ask the learner to explain the new model in their own reasoning.
11. Test the old trigger: present a fresh case that would previously have activated the misconception.
12. Vary: change surface features and representations.
13. Delay: return after time has passed.
14. Monitor reappearance: check whether the misconception returns under speed, complexity or unfamiliar context.
15. Retire: stop special monitoring when the correct model remains stable across appropriate conditions.

Diagnosis begins by preserving the learner’s first model

Teachers often correct too quickly. The learner says something wrong, and the teacher immediately supplies the correct statement. The answer improves, but the teacher loses the chance to see what produced the error.

Before correcting a consequential misconception, ask the learner to make the rule visible. “What makes you think that?” “What would happen if the denominator became ten?” “Which part of the diagram supports your idea?” “What do you think the plant is using the soil for?”

The goal is not to interrogate the learner. It is to capture the structure of the current explanation before new information overwrites the evidence.

Prediction is one of the cleanest ways to surface a model

Ask what should happen before showing what does happen. A prediction commits the learner to a model that can later be compared with evidence.

In science: “Which object will reach the floor first if air resistance is negligible?” In mathematics: “Without calculating fully, which is larger: one-third or one-fifth?” In English: “Where would you expect the main idea to appear in this paragraph, and why?”

Prediction is particularly useful when the misconception concerns a causal mechanism or overgeneralised rule.

Confidence can help distinguish a misconception from a guess

A wrong answer given with high confidence is not proof of a misconception, but it is different evidence from a wrong answer given with uncertainty.

Ask learners how confident they are and why. High-confidence wrong reasoning may indicate a stable wrong model. Low-confidence responses may reflect weak knowledge or guessing. Correct low-confidence responses can reveal fragile understanding that needs reinforcement.

Confidence should never replace reasoning. It is one additional signal.

Use multiple cases before deciding what the learner believes

A misconception is a pattern. Test the suspected rule in more than one form.

If the learner thinks a larger denominator always gives a larger fraction, compare one-half with one-fifth, then three-quarters with three-eighths, then one-quarter with three-eighths. The third pair matters because it shows that a denominator shortcut cannot be used when numerators differ.

If the learner says “the first sentence is the main idea,” use a paragraph with a topic sentence at the end, then one with the central idea implied across several sentences. The teacher is mapping the rule’s boundary, not merely collecting more wrong answers.

Hinge questions can expose common misconception patterns efficiently

In whole-class teaching, a well-designed multiple-choice question can separate several plausible models quickly. Each option should reflect a meaningful reasoning path, not a random distractor.

EEF’s 2026 checking-for-understanding guidance recommends planning response options so that different answers reveal different understandings or misconceptions. This makes the question useful for deciding whether to pause and fix, adapt support or move on.

Do not assume each option corresponds to one unique mental state. Follow up important patterns with explanation or another discriminating question.

Contrast is powerful because it reveals the feature that controls the concept

A single correct example shows what belongs. A contrast pair can show why.

Compare a complete circuit with an open circuit. Compare a correlation claim with a causal claim. Compare two algebraic transformations, one preserving equality and one not. Compare a paragraph that merely quotes evidence with one that explains the evidence-to-claim relationship.

Ask the learner to identify the feature that changes the classification. This directs attention away from superficial similarity and towards the governing relation.

A counterexample is useful only when the learner can understand why it matters

Teachers sometimes produce a dramatic counterexample and expect the misconception to disappear. But surprise alone does not teach the replacement model.

If a learner believes “multiplication always makes numbers larger,” showing that one-half multiplied by one-half equals one-quarter creates a conflict. The teacher still needs to explain why multiplying by a number between zero and one scales the original quantity down.

The counterexample creates a problem for the old model. Explanation builds the new one.

Cognitive conflict without resolution can leave two competing models

A learner can see evidence that contradicts a belief and still preserve the belief by treating the evidence as an exception. “Usually heavier objects fall faster; this one is special.” “Normally the bigger denominator means the bigger fraction; this question is tricky.”

The teacher must therefore close the explanatory gap. Why did the prediction fail? What more general principle accounts for both the familiar and the surprising case?

The goal is not conflict for its own sake. It is conceptual reorganisation.

Refutation should name the old idea, explain why it fails and supply the better model

A useful refutation sequence is:

• State the tempting or common belief accurately.
• Identify the condition under which it fails.
• Provide evidence or a counterexample that makes the failure visible.
• Explain the correct relationship or mechanism.
• Apply the new model to both the original and a new case.

This structure is the teacher-side counterpart of the MindOS Refutation State. Telling the right answer is one part of repair; explaining why the wrong model fails is another.

The replacement model must do more explanatory work than the misconception

Wrong models survive because they are useful. They explain some experiences, simplify decisions and often produce correct answers in limited cases.

A replacement model should therefore be more powerful, not merely more official. It should explain why the old rule sometimes appeared to work, why it fails elsewhere, and how to decide correctly in a broader range of situations.

For fractions, “the denominator tells the size of equal parts of one whole” explains why one-fifth is smaller than one-half and why denominator-only comparisons sometimes work when numerators are the same. For heat, “insulation reduces the rate of energy transfer” explains why an insulated object can still change temperature rather than forcing the false binary choice of “heat escapes” versus “heat cannot escape.”

Teach conditional rules instead of replacing one overgeneralisation with another

Misconception repair can fail when the new teaching becomes another slogan.

“Smaller denominator means larger fraction” is valid only when positive fractions have the same numerator and refer to equal wholes. If learners memorise the shortcut without the conditions, the repair creates a new future misconception.

Teach the condition with the rule. Ask learners to state when the rule applies and when it does not.

Representations can make the replacement model visible

Some misconceptions persist because symbolic language hides the relation. A well-chosen representation can externalise it.

Number lines and fraction strips can show that one-half is to the right of one-fifth. Balance representations can make equation equivalence visible. Energy-flow diagrams can show transfer rather than treating “heat” as a substance stored inside an object. Evidence maps can show the link between a claim and supporting details.

The IES fractions practice guide and its later teaching toolkit use representations as central tools for addressing common fraction misunderstandings. The important principle is not that pictures are automatically better. The representation must preserve the relevant relationship accurately.

Move between representations so the repair does not depend on one picture

A learner may succeed with a fraction strip and fail with symbols. Another may understand a graph but not the verbal relationship. Repair is stronger when the learner can translate.

Ask the learner to explain the same idea with words, symbols, a diagram and a fresh example where appropriate. Translation reveals whether the relationship has become portable.

Analogies help only when the mapping and its limits are taught

An analogy can connect a new model to familiar knowledge, but it can also generate the next misconception.

For every analogy, identify what maps and what does not. A water-flow analogy may help introduce aspects of electric current but should not imply that charge is consumed like water in a leaking pipe. A “memory as storage box” analogy may help with capacity ideas but can distort retrieval and reconstruction.

A mature repair eventually moves beyond the analogy to the disciplinary model.

Worked incorrect examples can turn mistakes into objects of analysis

Teachers can present a deliberately incorrect solved problem and ask learners to locate the first invalid step, explain why it fails and repair it.

The IES algebra practice guide recommends using solved problems to engage students in analysing algebraic reasoning, including comparisons between correct and incorrect solutions. Its examples ask learners to explain why errors lead to wrong answers and how the solution should be checked.

Incorrect examples should be clearly labelled when learners might otherwise encode the error. The task requires enough prior knowledge to evaluate the reasoning.

Ask learners to explain why the wrong answer is tempting

This is a powerful advanced move. “Why might someone think this?” turns the misconception into an object rather than an identity.

For the fraction misconception, learners can explain that whole-number magnitude is being applied to denominators. For a science misconception, they can identify the everyday observation that makes the wrong causal story plausible.

Understanding the temptation strengthens discrimination. The learner can recognise the trigger when it returns.

Do not rehearse the misconception more strongly than the correction

A long lesson about every wrong version can accidentally make the wrong idea highly familiar. The misconception should be surfaced clearly enough to diagnose and refute, then attention should move to the correct model and its use.

Repeat the correct relationship across varied examples. When the old misconception reappears, use it diagnostically rather than turning the lesson into a catalogue of errors.

Worked case: “One-fifth is bigger than one-half because five is bigger than two”

Begin by preserving the learner’s reasoning. Ask: “If the same pizza is cut into five equal pieces and another is cut into two equal pieces, which individual piece is larger?”

If the learner still chooses the fifth, use equal-whole fraction strips or a number line. Make sure the representations are genuinely comparable. The visual should show that one-half covers more of the whole and lies farther to the right than one-fifth.

Then explain the model: the denominator is not simply a whole number being compared with another denominator. It tells how many equal parts make the whole. More equal parts means each part is smaller when the whole remains fixed.

Now test the boundary. Compare three-quarters and three-eighths. With the same numerator, the larger pieces make three-quarters greater. Then compare one-quarter and three-eighths. The denominator-only shortcut no longer decides the problem because the numerators differ. One-quarter equals two-eighths, so three-eighths is larger.

The learner’s replacement model should now be able to explain all three comparisons, not just memorise that “two beats five.”

Return later with unfamiliar fractions and a number-line placement task. If the old whole-number rule returns under time pressure, the misconception has weakened but not yet disappeared.

Worked case: “Moving a term across changes its sign”

Students often learn the useful classroom shorthand “move it across and change the sign.” The shorthand can produce correct answers while hiding the governing principle.

Consider x + 5 = 12. The learner writes x = 12 – 5. Ask why the sign changed. If the explanation is “because it crossed the equals sign,” present 3x = 12 and ask what “crossing” would mean for the 3.

Build the replacement model: an equation states equality. We preserve equality by performing equivalent operations on both sides. Subtracting 5 from both sides gives x + 5 – 5 = 12 – 5. Dividing both sides of 3x = 12 by 3 gives x = 4.

The phrase “move across” can remain as compressed expert shorthand later, but it should be attached to the operation that preserves equality. Otherwise the learner may apply sign-changing rituals incorrectly in more complex algebra.

Test transfer with fractions, brackets and variables on both sides. Ask the learner to justify one transformation without using “move across.”

Worked case: “Plants get their food from the soil”

This misconception is plausible because plants grow in soil, roots absorb materials from soil, and fertiliser affects growth.

First separate what the learner means by “food.” Ask what material becomes new plant mass and where the carbon in plant tissues comes from. The learner may use “food” to mean mineral nutrients, which reveals a vocabulary and mechanism issue rather than a simple factual mistake.

Build the replacement model: green plants manufacture glucose during photosynthesis using carbon dioxide and water, with light energy captured by chlorophyll. Mineral salts from soil are important nutrients but are not the plant’s manufactured food in the same sense.

Use a representation that traces matter sources rather than a slogan. Carbon dioxide contributes carbon to carbohydrates; water contributes hydrogen and oxygen; roots supply water and mineral ions. Ask where each input enters the plant and what role it plays.

Then test a new case: a plant grown hydroponically without soil. If the learner says it cannot make food because there is no soil, the old model is still active. If the learner can explain how water, mineral ions, carbon dioxide and light support the plant without soil, the replacement model has broader reach.

For the PSLE-specific mechanism, use the separate subject owner How to Repair a PSLE Science Misconception by Replacing the Wrong Mechanism, Not Memorising an Exception.

Worked case: “A lid stops heat from escaping”

Suppose an uncovered container and an insulated container both begin at 70°C. Ten minutes later, the uncovered container is 54°C and the covered container is 61°C.

The statement “the lid stops heat from escaping” conflicts with the covered container’s 9°C temperature decrease. Use that evidence to challenge the absolute term stops.

Build the replacement: insulation can reduce the rate of energy transfer; it does not usually eliminate transfer completely. The model explains both why the covered container cools more slowly and why it still cools.

Then change the context. Ask about a thermal flask, a jacket or a cool box. The learner should use the same principle while respecting the different mechanisms and conditions in each case.

Worked case: “Heavier objects fall faster”

Everyday experience can make this belief feel obvious because paper, stones, leaves and balls differ in shape and air resistance as well as mass.

Ask the learner to predict what happens when two objects of different mass have similar shape and air resistance, or discuss the idealised case where air resistance is negligible.

The replacement model must distinguish gravitational acceleration from effects of air resistance. A heavy stone and a light compact object can accelerate similarly under gravity when drag differences are negligible; a sheet of paper may fall slowly because of its interaction with air.

Do not turn the repair into another slogan such as “everything falls at the same speed.” Speed changes during the fall, and air resistance matters in real situations. Teach the conditions attached to the model.

Worked case: “The main idea is always the first sentence”

This classroom rule may begin as helpful scaffolding because many instructional paragraphs place the topic sentence near the beginning. It becomes a misconception when the learner treats position as definition.

Present three short paragraphs: one with the main idea first, one with it last, and one where the central idea must be inferred from several details.

Ask what makes an idea “main.” Build the replacement model around explanatory coverage: the main idea captures what the paragraph is mainly saying and accounts for the supporting details. Sentence position can be a clue, but it is not the governing rule.

Then use an unfamiliar paragraph with a misleading first sentence. If the learner still chooses by position, more varied practice is needed.

Worked case: “A longer answer gets more marks”

Students can form this model because longer model answers often contain more relevant reasoning. They may reverse the relationship and assume length itself causes quality.

Compare a long repetitive answer with a shorter answer that directly satisfies the command word, uses relevant evidence and explains the necessary relationship.

The replacement model is task alignment: marks or quality depend on meeting relevant criteria, not producing maximum volume. More words help only when they carry needed information.

Ask the learner to improve an answer by removing one irrelevant sentence and adding one missing reasoning link. The repair should demonstrate that quality can increase while length decreases.

Misconceptions in grammar can come from overgeneralised school rules

“Never start a sentence with ‘And’,” “never use the passive voice,” or “a paragraph must have exactly five sentences” can begin as simplified teaching advice and later become brittle rules.

Repair should distinguish an instructional heuristic from a grammatical law. Show when the advice is useful, what problem it was meant to prevent, and examples where skilled writing violates the simplified rule for a clear reason.

The replacement is conditional judgement, not rejection of all structure.

Misconceptions about studying deserve the same repair logic

Learners can hold misconceptions about learning itself: “If the page looks familiar, I know it,” “highlighting is revision,” “more hours always mean more learning,” or “if I cannot answer immediately, I never learned it.”

Use prediction and evidence. Ask the learner to predict recall after rereading and after retrieval practice, then compare later performance. The point is not to humiliate the learner’s study method but to give the learner better evidence about how memory behaves.

Study-strategy misconceptions often survive because familiarity is subjectively convincing. A replacement model needs a different signal of learning: retrieval, explanation, transfer and delayed performance.

Some misconceptions are intuitive and may never disappear completely

Learning the correct model does not always erase the intuitive response. Under speed, distraction or unfamiliar framing, an older answer can reappear.

This matters because conceptual change can involve control as well as replacement. The learner may need to recognise a familiar intuitive trigger, pause and apply the disciplinary model deliberately.

The EEF’s recent Stop and Think evaluation is useful precisely because it gives a nuanced result. The programme trained pupils to slow down on counterintuitive maths and science items. In a large effectiveness trial, science attainment improved on average while mathematics showed no additional progress overall. That is a warning against treating “stop and think” as a universal misconception cure.

The educational principle is narrower: some counterintuitive problems may benefit from deliberate inhibition of an initial response, but the effect depends on domain, content, implementation and the quality of the replacement knowledge.

Speed can reactivate a repaired misconception

A learner may answer correctly in an untimed explanation and revert to the old rule during a timed paper. That does not necessarily mean the repair was imaginary. The correct model may still be slower or less automatic.

Build fluency after conceptual repair. Use gradually faster retrieval and mixed problems only after the learner can explain the governing relation accurately.

Do not train speed on the misconception itself.

Complexity can reactivate a repaired misconception

A learner may understand the corrected idea in isolation but revert when the problem has several steps. Working-memory demand can cause the learner to fall back on a familiar shortcut.

Test the model first in a simple context, then embed it inside more complex tasks. If the misconception appears only under high load, the next intervention may involve fluency and task coordination rather than another conceptual lecture.

Repair is not proven by one corrected answer

The learner can often repeat the teacher’s correction immediately. That shows access to the new information, not stable conceptual change.

Verification should include several conditions:

• a fresh problem that triggers the old misconception;
• a different surface context;
• a different representation;
• an explanation of why the tempting wrong rule fails;
• a delayed return after the original teaching is no longer fresh;
• mixed practice where the learner must decide which rule applies.

The misconception is becoming weaker when the learner can recognise the trigger, select the correct model and explain the boundary without teacher prompting.

Use the old trigger deliberately in the verification task

A generic easy question may not test the repair. The new task should contain the feature that previously activated the wrong rule.

If the misconception concerned denominator size, include a comparison where whole-number intuition is tempting. If it concerned “correlation means causation,” use a new graph with a strong association but no experimental evidence. If it concerned “main idea equals first sentence,” use a paragraph with a deliberately attractive opening detail.

The teacher is testing control over the old model, not merely recall of the correction.

Near transfer and far transfer answer different questions

Near transfer changes a small feature. Farther transfer changes the context or representation more substantially.

After repairing a fraction misconception with bars, near transfer may use a number line. Farther transfer may use ratio or probability contexts where fractional magnitude must be interpreted without visual partitions.

Do not jump to far transfer so quickly that failure becomes uninterpretable. Build evidence step by step.

Delayed return matters because old models can recover

The old misconception may feel weaker immediately after a strong explanation because the correct model is highly accessible. After time passes, the older intuitive rule can regain influence.

Return later without first displaying the correction. If the learner succeeds, confidence in the repair increases. If the misconception returns, re-enter at the point of failure rather than simply repeating the entire original lesson.

Ask learners to generate their own counterexample

Generating a counterexample is stronger than recognising one because the learner must understand what property would make the old rule fail.

“Give me two fractions that show why larger denominator does not always mean larger value.” “Invent a case where two variables correlate but neither causes the other.” “Write a paragraph where the main idea is not the first sentence.”

This turns misconception repair into concept-boundary knowledge.

Ask learners to teach the repaired idea

Have the learner explain the tempting misconception, why it fails, and the correct model to an imagined younger learner.

Teaching the repair forces organisation. It also reveals whether the learner has merely memorised a correct phrase or can distinguish the two models.

Keep the explanation bounded. The learner should not repeat the misconception so extensively that the wrong wording becomes the dominant memory trace.

Whole-class misconception repair begins with broad evidence, not the loudest volunteer

If a teacher hears one wrong answer, it is risky to assume the whole class shares the misconception. If a confident learner gives the right answer, it is equally risky to assume everyone understands.

Use all-pupil response methods where practical: mini-whiteboards, short written predictions, polls or carefully designed multiple-choice questions. EEF and AERO current guidance both emphasise gathering evidence broadly enough to guide the next teaching decision.

If many learners share the wrong model, pause and repair publicly. If only a few show it, use targeted support without forcing the entire class through unnecessary reteaching.

A whole-class repair should use anonymous reasoning, not public humiliation

Teachers can write a common misconception on the board without naming the student who produced it. “A common answer was…” turns the idea into an object for analysis.

Ask what makes the answer tempting, where it fails, and what model explains the evidence better. The classroom message should be that errors are inspectable reasoning, not social identities.

Small-group teaching makes misconception pathways unusually visible

In a three-student group, the teacher can compare reasoning across learners in real time. One learner may hold the misconception strongly, another may be uncertain, and the third may have the correct model but poor language for explaining it.

Collect individual predictions before discussion. Otherwise the most confident learner can overwrite the evidence from the others.

Then use shared contrasts. One learner explains the tempting rule, another identifies the counterexample, and another states the replacement principle. Rotate roles so no learner becomes permanently “the weak one” or “the explainer.”

Finish with fresh individual checks. Group agreement is not the same as individual model change.

Misconception repair should preserve learner dignity

Wrong models are normal products of learning. If learners are punished socially for exposing them, they become less willing to reveal reasoning.

Use language such as “That rule works in this situation but breaks here,” or “I can see why that interpretation is tempting.” Then move quickly into evidence and explanation.

Empathy does not require pretending the misconception is correct. It requires separating the person from the model and giving the learner a credible route to change it.

Do not reward confident wrongness, but do reward exposed reasoning

A learner who explains an incorrect model clearly has given the teacher valuable evidence. The reasoning still needs correction, but the act of exposing it should be safe.

Say, “That explanation helps us see exactly where the model changes,” rather than praising the misconception itself. Then use the visible reasoning to teach the boundary.

Misconceptions can be teacher-made

Simplifications, shortcuts and analogies can create future errors when their limits are not taught.

“Move it across and change the sign,” “plants take food from the soil,” “current gets used up,” or “keywords get marks” can become misleading models when used without conditions.

When a class shares a misconception, inspect the teaching history as well as the learners. The repair may require changing the teacher’s language, examples or sequence.

Textbooks and model answers can also stabilise misconceptions

A compressed worked solution can hide why a method works. A model answer can make a surface phrase look like a required keyword. A diagram can imply a false scale or direction.

Teach learners to use models as evidence of decisions, not as scripts. The existing How Studying From Model Answers Works route develops that learner-side boundary.

Mark schemes can create “magic keyword” misconceptions

Students can conclude that marks are awarded for inserting certain words rather than demonstrating the required relationship.

Feedback should reconnect terms to meaning. A scientific term is useful because it expresses a concept precisely. A quotation is useful because it supports an interpretation. A mathematical notation is useful because it represents a relationship.

When a learner asks “What keyword do I need?”, ask what idea the word would need to communicate.

Misconceptions can arise from language, not only concepts

Words such as “work,” “power,” “theory,” “average,” “function,” “significant,” and “energy” can have everyday meanings that differ from disciplinary meanings.

A learner may understand the underlying phenomenon but map the wrong everyday meaning onto a technical term. Repair then needs semantic contrast as well as subject explanation.

Use the everyday meaning, the technical meaning and examples where confusing them would change the conclusion.

Bilingual learners may carry a translation-shaped misconception

A one-to-one translation can imply that two terms are exact equivalents when their usage differs. The learner may then apply a concept boundary from one language to another.

Use examples in context and compare meanings rather than relying on isolated word substitution. Diagnose whether the problem is conceptual, linguistic or both.

Examination pressure can make misconceptions look stronger than they are

Under time pressure, learners may revert to fast familiar rules. The correct model may still exist but lose the competition for retrieval.

After conceptual repair, practise recognition under gradually increasing time demand. Do not infer complete conceptual failure from one timed relapse without checking untimed reasoning.

Repeated correction without model repair creates error whack-a-mole

The teacher corrects one answer, then the same principle fails in a new context. Another answer is corrected. The cycle repeats because each surface error is treated separately.

When a pattern recurs, move upward to the generative rule. Ask what common idea could produce all of these errors. Repairing the model can eliminate several surface mistakes at once.

Misconception repair should not become a hunt for hidden defects

Not every incorrect answer needs deep conceptual analysis. Overdiagnosis wastes time and can lead teachers to invent elaborate explanations for ordinary slips.

Use the simplest explanation consistent with the evidence, then test it. If a one-off arithmetic error disappears on the next question and the learner explains the principle correctly, move on.

Positive conceptions should be reinforced too

When a learner uses the correct model under a tempting condition, identify what they did well.

“You ignored the denominator shortcut because the numerators were different and converted to eighths instead.” “You described a strong correlation but did not claim causation without causal evidence.”

Positive feedback helps stabilise the decision rule that replaced the misconception.

Use misconception maps cautiously

Teachers benefit from knowing common misconceptions, but a list can become a confirmation trap. If the teacher expects a particular error, ambiguous student work may be forced into that category.

Use common-misconception knowledge to design good questions, not to skip diagnosis. The learner’s actual reasoning remains the evidence.

Build a teacher library of discriminating questions, not just misconception labels

The most useful professional asset is not a list of “common misconceptions.” It is a set of small tests that separate competing explanations.

For fractions: equal numerators, unequal numerators, number-line placement. For causation: observational association versus controlled manipulation. For main idea: first, last and implied topic sentences. For algebra: equivalent operations versus sign-moving shorthand.

These questions shorten diagnostic time and improve the precision of repair.

A misconception registry should be temporary and evidence-based

A useful note can say: “Overgeneralises whole-number magnitude to unit fractions; corrected with equal-whole representations; succeeds on same-numerator comparisons; mixed-numerator transfer still unstable.”

This is better than “bad at fractions.” It identifies a model, evidence and next test.

Retire the entry when the learner remains stable across relevant tasks. An old misconception should not become a permanent identity.

Feedback and misconception repair are not identical

Feedback tells learners something about their performance relative to a goal. Misconception repair is a special case where the performance is generated by a wrong model that needs conceptual reorganisation.

A short feedback cue can be enough when the correct model is already present. A misconception requires more: surface the model, refute or bound it, explain the replacement and test the old trigger.

This is why the current article follows How Feedback Works in Teaching but remains a distinct owner.

Questioning and misconception repair form a loop

Questions surface the model. Contrast questions test its boundary. Explanation questions reveal whether the replacement has formed. Transfer questions test whether the old model still wins elsewhere.

This connects directly to How Questioning Works in Teaching. The best misconception questions are designed to distinguish models, not merely mark responses right or wrong.

Worked examples can repair misconceptions when comparison is explicit

A correct worked example beside an incorrect one can expose the first point where reasoning diverges.

Ask what both solutions agree on, where they separate, which principle controls that step and how the result can be checked. The learner should not merely copy the correct route.

The teacher-side design principles appear in How Worked Examples Work in Teaching.

Guided practice is where the new model becomes usable

After conceptual explanation, learners need supported opportunities to apply the replacement model before independence.

Early practice can use close cases and prompts that direct attention to the critical feature. Later prompts fade, problem types mix, and the learner must recognise the old trigger independently.

See How Guided Practice Works in Teaching.

Retrieval practice should retrieve the correct model, not only the answer

After repair, ask learners to retrieve the principle and its boundary. “Why is one-fifth smaller than one-half?” “When can a smaller denominator indicate a larger fraction?” “Why does correlation not prove causation?”

This helps the replacement model become accessible without the original contrast present.

Spacing matters because conceptual change must survive time

Return to the misconception after days or weeks rather than assuming immediate correction is durable.

The return should be short and diagnostic. One well-chosen problem can reveal whether the old model has recovered.

Interleaving tests whether the learner can choose between competing rules

Blocked practice tells learners which method or concept is active. Mixed practice forces discrimination.

After repairing a percentage-base misconception, mix questions involving percentage change, percentage points and simple proportions. After repairing correlation-causation confusion, mix observational and experimental evidence.

The learner must now decide which model applies.

Misconception repair can fail because the correct model is too abstract

A technically accurate explanation may not connect to anything the learner can use.

Begin with a concrete or representational case when needed, then connect it to formal language. For fractions, equal-whole visual models can precede symbolic generalisation. For electrical concepts, observable circuit behaviour can precede more abstract current and potential models.

Do not remain concrete forever. The learner should eventually use the abstract model without depending on the original demonstration.

Misconception repair can fail because the counterexample adds a second difficulty

If the “counterexample” requires unfamiliar vocabulary, complex arithmetic or a new representation, failure may reflect the added demand rather than the misconception.

Use the simplest case that isolates the conceptual conflict. Once the boundary is understood, increase complexity.

Misconception repair can fail because the new model is taught as a slogan

“Heat moves from hot to cold.” “Correlation is not causation.” “The denominator tells the size of the pieces.” These can be useful summaries but weak replacements if learners cannot explain or apply them.

Ask learners to use the statement in a case, explain why it changes a prediction, and identify a boundary. A slogan should compress understanding, not substitute for it.

Misconception repair can fail because the learner memorises an exception

The learner may think, “One-half is bigger than one-fifth because that example is special,” while retaining the denominator rule elsewhere.

Use multiple examples that require one general model. Ask what all of them have in common. The replacement should explain a family of cases, not one exception.

Misconception repair can fail because teacher confirmation arrives too soon

If the teacher approves every step immediately, the learner may follow cues rather than reconstruct the model.

After initial support, delay confirmation and ask the learner to justify the decision. Then use a fresh problem without the teacher’s facial or verbal cues.

Misconception repair can fail because the learner never meets the old trigger again

If later practice contains only obvious correct cases, the teacher never discovers whether the old model remains dormant.

Design occasional “temptation tests” that resemble the conditions where the misconception used to win. Use them as diagnostic checks, not tricks.

Misconception repair can fail because a whole-class correction misses individual models

Several learners can give the same wrong answer for different reasons. One class explanation may repair one model while leaving another intact.

After whole-class repair, use a short individual check that requires explanation. If different reasoning persists, target the remaining model.

Misconception repair can fail because the teacher never checks the replacement model’s limits

The new model may itself be overgeneralised. “Larger denominator means smaller fraction” can become another misconception when numerators differ.

Every replacement should include at least one boundary case. Teach not only the rule but its conditions.

A practical teacher sequence for misconception repair

1. Preserve the first response.
2. Ask for the learner’s reason or prediction.
3. Test whether the pattern repeats on a second case.
4. State the suspected misconception provisionally.
5. Select a simple case that the misconception predicts incorrectly.
6. Ask the learner to commit to a prediction where appropriate.
7. Reveal the evidence or work through the counterexample.
8. Ask why the old rule fails.
9. Teach the replacement model explicitly.
10. Connect the replacement to prior knowledge.
11. Use a representation if it makes the relation visible.
12. Ask the learner to explain the new model.
13. Apply it to the original case.
14. Apply it to a near-transfer case.
15. Remove prompts.
16. Use a mixed or farther-transfer case.
17. Return after delay.
18. Record whether the old model reappears.
19. Reduce special monitoring when the new model remains stable.

A compact misconception-repair record

A useful record can fit in one line:

Trigger → learner model → discriminating evidence → replacement model → fresh test → delayed state.

Example: “Unit fractions → larger denominator treated as larger value → equal-whole strips and one-half vs one-fifth prediction → denominator controls part size when the whole is fixed → mixed comparisons independent → recheck in two weeks.”

The record should describe a current instructional state, not a permanent property of the learner.

AI can help generate contrasts, but it can also invent misconceptions

An AI system can produce examples, non-examples, diagnostic questions and alternative explanations quickly. These capabilities can help teachers prepare misconception checks.

But AI can confidently claim that a wrong answer “shows” a particular misconception when the evidence is ambiguous. It can also generate a counterexample that changes several variables at once or a technically incorrect explanation.

Use AI output as draft material. Verify the subject content, solve the examples, check the counterexample, and avoid uploading identifiable learner information unnecessarily.

When learners use AI, ask them to state their own prediction before seeing the generated explanation. After the explanation, close it and use a fresh task. Otherwise the tool may repair the current answer while the learner’s original model remains untested.

The broader boundary remains How AI-Assisted Study Works.

For parents: ask what rule the child is using before supplying the correction

If a child gives a surprising answer, begin with, “Show me how you decided.” The explanation may reveal a sensible but misplaced rule.

Do not turn the conversation into a debate about intelligence or effort. Work with the model. Use one simple contrast or example. If the issue depends on subject knowledge you are unsure about, record the child’s reasoning and ask the teacher rather than inventing a marking rule.

After helping, ask for a fresh case. A child who can repeat the parent’s correction may still hold the original model.

For teachers: do not confuse a misconception clinic with the whole curriculum

Misconception work is valuable because it repairs predictable failure points. It should not become a curriculum organised entirely around errors.

Teach the subject positively and coherently. Use misconception checks at points where they clarify boundaries, correct emerging errors or protect later learning.

A learner needs a rich correct model, not an encyclopaedia of wrong ones.

Common teacher failure modes in misconception repair

• Correcting before eliciting: the teacher loses evidence about the learner’s model.
• Diagnosing from one answer: a slip becomes a supposed misconception.
• Using a dramatic counterexample without explanation: surprise replaces teaching.
• Teaching an exception: the old rule survives elsewhere.
• Replacing one slogan with another: conditions and mechanisms remain hidden.
• Using a counterexample with too much extra difficulty: failure becomes uninterpretable.
• Over-rehearsing the wrong idea: the misconception becomes more familiar than the correct model.
• Publicly attaching the misconception to a student: reasoning becomes a social identity.
• Assuming class agreement means individual repair: peer answers hide remaining models.
• Testing only the original example: transfer remains unknown.
• Never returning after delay: the old model may recover unnoticed.
• Never testing under complexity or time: fragile control is mistaken for stable change.
• Using common-misconception lists as diagnosis: teacher expectation replaces evidence.
• Ignoring teacher-created misconceptions: problematic shorthand remains in instruction.
• Keeping the label after repair: an old state becomes a permanent learner identity.

A strong repair should answer five questions

• What model was the learner using?
• Why was that model plausible?
• What evidence reveals its limit?
• What better model explains both the easy and difficult cases?
• What fresh evidence shows the new model now controls performance?

If the teacher cannot answer the first question, diagnosis may be premature. If the learner cannot answer the fourth, conceptual replacement may be incomplete. If the fifth is missing, repair remains unverified.

Evidence, interpretation and limits

Current educational guidance strongly supports checking for understanding, identifying misconceptions early and adapting teaching in response. AERO’s Monitor Progress guide, updated 14 May 2026, recommends regular checks for understanding and additional instruction, guidance or feedback where necessary. Its At a glance: Formative assessment resource, published and updated in July 2026, explicitly describes formative assessment as a way to reveal misconceptions so they can be corrected early.

EEF’s 2026 adaptive-teaching resources similarly frame misconceptions as information for deciding whether to pause and fix, adapt support or extend. The January 2026 checking-for-understanding guidance emphasises designing diagnostic options around specific misconceptions rather than collecting answers for their own sake.

Older IES / What Works Clearinghouse practice guides remain useful for domain-specific mechanisms. Developing Effective Fractions Instruction identifies common fraction misconceptions and recommends conceptually meaningful representations and explanations. Teaching Strategies for Improving Algebra Knowledge recommends analysing solved problems and explaining why errors lead to incorrect results. These sources are domain-specific and should not be treated as proof of one universal misconception protocol.

The 2025 EEF Stop and Think effectiveness trial provides a useful caution about generalisation. The trial involved 173 schools and 14,645 pupils. EEF reported no additional months of maths progress overall, while science showed two additional months on average; the evaluation received a 3-out-of-5-padlock security rating after attrition reduced the rating. The result does not prove that all misconception inhibition methods work in science or fail in mathematics. It shows why domain, implementation and outcome matter.

The original examples, repair cycle and teaching record in this article are editorial syntheses. They have been checked for internal logic but have not been evaluated as one complete intervention. No fixed number of counterexamples, universal timing interval or guaranteed learning gain is claimed.

Sources for this edition were reviewed on 13 September 2026. Later revisions should recheck current source versions before repeating time-sensitive dates, results or implementation claims.

Selected sources

Australian Education Research Organisation: Monitor Progress; At a glance: Formative assessment; and Monitor progress: Primary and secondary.

Education Endowment Foundation: How checking for understanding can guide your teaching in the moment; Checking for understanding that leads to action; and Stop and Think: Learning Counterintuitive Concepts – second trial.

Institute of Education Sciences / What Works Clearinghouse: Developing Effective Fractions Instruction for Kindergarten Through 8th Grade; Teaching Fractions Toolkit; and Teaching Strategies for Improving Algebra Knowledge in Middle and High School Students.

The misconception-repair standard: the old model must lose where it used to win

Return to one-fifth and one-half.

The repair is not complete because the learner can now say that one-half is larger. The repair is stronger when the learner can explain why, identify why the whole-number intuition was tempting, state when denominator comparison is valid, solve a changed case, place the fractions on a number line, and still use the correct model after time has passed.

That is the real test of misconception repair.

The old model must encounter the situations where it once controlled the answer—and lose to a better model for the right reason.

Continue through How Teaching Works, revisit How Feedback Works in Teaching, or use the learner-side Refutation State.