The Tutor Handbook Vol No.0018 | The Reattempt — How a Tutor Uses a Second Attempt to Separate Recognition, Repair and Real Learning

The Tutor Handbook · Volume 0018 · Series ID THB-0018

The student tries again.

That sounds simple.

It is not.

On the first attempt, the learner may have misunderstood the question, chosen the wrong route, forgotten a prerequisite, made an execution error, run out of time, relied on a prompt, copied a familiar pattern or simply not known what to do.

Then something happens.

The tutor explains.

The learner corrects.

A model answer is seen.

A hint is given.

The weak step is repaired.

Now comes the dangerous moment.

The student attempts the work again and gets it right.

What exactly did the second attempt prove?

Perhaps the learner has genuinely learned.

Perhaps the learner recognised the page.

Perhaps they remembered the tutor’s sentence.

Perhaps they remembered that the answer was B.

Perhaps the repaired reasoning is available, but only on the exact old question.

Perhaps the second attempt is the first real evidence that something changed.

This is Volume 0018 of The Tutor Handbook, eduKate Sengkang’s long-form practical series on the real work inside tutoring.

Volume 0017, The Return, examined what a tutor can learn when evidence comes back from outside the lesson. This volume zooms in on one of the most useful things a tutor can do with that evidence:

Ask the learner to try again under deliberately chosen conditions.

The broader evidence cycle already has a canonical owner in eduKate Sengkang’s Evidence Loop | How We Know Learning Changed. The subject-specific decision about repeating an old PSLE Science paper already has its own owner in How to Decide Whether to Reattempt the Same PSLE Science Paper or Use a New One After Corrections. This handbook volume does not replace either page.

Its narrower job is tutor practice across subjects and levels:

How should a tutor design, read and interpret a second attempt so that recognition is not mistaken for repair, and repair is not mistaken for durable learning?

The tutor classification remains anchored in the Tutor Classification Model by eduKateSG, the Tutorial Class Control Tower and the Three Modes of Tuition. The Reattempt is not a new tutor class. It is an evidence operation available to every class.

Quick Read

• A second attempt is useful only when the tutor knows what it is meant to test.
• The first attempt gives a baseline; correction changes the learner’s state; the reattempt tests what became independently available afterwards.
• Getting the same question right again can show useful retrieval, but familiarity may inflate performance.
• A reattempt should record support conditions: independent, self-corrected, lightly prompted, example-supported or fully retaught.
• The tutor should distinguish recognition of the old answer from reconstruction of the underlying reasoning.
• Same-item success is not the same as transfer. A changed question is often the next test.
• Wrong again is useful evidence when the tutor can see whether the same failure signature returned.
• A different wrong answer can be progress if the original weak link has been repaired and a later one is now visible.
• The tutor should avoid endless same-item repetition. Once the repair is stable, widen the context.
• The long-term goal is a learner who can detect a failed attempt, repair it, reattempt independently and decide whether the improvement is real.

1. A Reattempt Is Not Merely “Do It Again”

“Try again” can mean several very different things.

• Repeat immediately while the explanation is fresh.
• Repeat after the learner has seen the answer.
• Repeat after a short delay without notes.
• Repeat after a longer delay.
• Repeat with one prompt removed.
• Repeat with the same question but no worked example visible.
• Repeat the reasoning on a changed question.
• Repeat the skill inside a mixed paper where the learner must identify when to use it.

These are not equivalent tests.

A reattempt becomes educationally useful when the tutor controls enough of the condition to know what the new performance means.

2. The First Attempt Is a Baseline, Not an Embarrassment

Students often want to erase the first wrong answer.

Tutors often want to replace it with a correct one.

That can destroy the most important comparison.

The first attempt tells the tutor:

• how the learner interpreted the task;
• which method was selected;
• where the first meaningful error entered;
• how far the learner could proceed independently;
• what checking occurred;
• whether the learner knew they were wrong;
• what kind of help became necessary.

Without the first attempt, the second attempt floats without a reference point.

Do not erase the state you are trying to change.

3. Correction Changes the State

Between Attempt 1 and Attempt 2, something enters the system.

It might be a tiny question:

What quantity are you actually comparing?

It might be a full explanation.

It might be a worked example.

It might be teacher feedback.

It might be the student noticing their own error.

The tutor should know what intervention occurred because Attempt 2 is partly a test of that intervention.

4. The Reattempt Is a Controlled Observation

A good reattempt asks one clean question.

For example:

• Can Alicia now identify the unknown before choosing the operation?
• Can Beatrice now distinguish cause from description?
• Can Ciara now carry the negative sign through expansion without a reminder?
• Can Denise now select the A-Math method without the topic label?
• Can Faith now include the scientific mechanism without copying the model sentence?

The cleaner the question, the easier the second attempt is to interpret.

5. Recognition, Repair and Real Learning Are Different States

State	What the learner may be doing	What the tutor should test next
Recognition	Remembering the answer, layout, wording or tutor cue	Remove familiar cues or change the surface
Repair	Reconstructing the corrected step on the old problem	Test after delay and on a changed item
Real learning	Selecting and using the repaired capability under less familiar conditions	Mix, delay, transfer and performance test

Recognition is not worthless.

Repair is not fake.

The mistake is calling an earlier stage a later one.

6. Familiarity Can Make Attempt 2 Look Stronger Than It Is

The learner remembers that the difficult question was on the bottom left.

They remember the diagram.

They remember that the tutor said “look at the denominator”.

They remember that Option C was wrong.

They remember the opening sentence of the corrected composition.

This familiarity can help the second attempt.

That does not make the reattempt useless. It makes its evidence narrower.

Same-item success says, “Something from the correction returned.” It does not yet say, “The learner owns the capability everywhere it matters.”

7. Ask the Learner What They Remember

After a correct reattempt, do not rush to the next page.

Ask:

Did you solve it again, or did you remember it?

A mature learner may say:

I remembered that my old answer was wrong, but I still had to reconstruct why.

That answer is valuable.

Memory of the old failure and reconstruction of the reasoning can coexist.

8. Immediate Reattempts Have a Job

Immediate reattempts are often dismissed as too easy.

They can still answer an important question:

Did the learner understand the correction well enough to reproduce it now?

If the student cannot reattempt successfully even while the correction is fresh, delay is not yet the main problem.

The repair itself may be incomplete.

9. Delayed Reattempts Have a Different Job

After time passes, the tutor can ask whether the learner can retrieve and reconstruct the repair when the original explanation is no longer occupying working memory.

There is no universal magic interval.

The principle is to create enough separation that the reattempt requires meaningful retrieval rather than simple continuation of the correction episode.

Highly memorable questions may need more delay or faster movement to changed contexts because the old surface remains a strong cue.

10. Close the Notes Before the Reattempt

If the purpose is to test independent reconstruction, the worked correction should not remain visible.

That sounds obvious.

In practice, students often glance at the corrected line, look back at the question, write a little, glance again and then believe they have “done it themselves”.

They have done useful supported practice.

They have not yet produced an independent reattempt.

11. Preserve the Old Attempt Beside the New One

Side-by-side comparison is powerful.

Attempt 1 shows the old route.

Attempt 2 shows the new route.

Ask the learner to identify the first line where the two attempts diverge.

That line often reveals the actual repair.

12. Ask “What Changed?” Before Saying “Correct”

A tutor can turn the second attempt into metacognitive training with one question:

What did you do differently this time?

Useful answers include:

• I identified the comparison before calculating.
• I checked what the pronoun referred to before answering.
• I wrote the scientific mechanism instead of jumping from condition to outcome.
• I selected the method before doing algebra.
• I stopped copying the model sentence and rebuilt the idea in my own words.

If the learner cannot name any changed operation, the tutor may need another test before concluding that the repair is understood.

13. Record the Support Condition on Attempt 2

Attempt 2 condition	Interpretation
Independent	Strongest immediate evidence from the same item
Self-corrected	Monitoring is becoming available
One attention cue	Repair is near but not yet self-triggering
Method named	Execution may be stronger than selection
Worked example used	Supported reconstruction
Full explanation repeated	The original repair may not yet be retrievable

A correct answer without a support record can be surprisingly hard to interpret.

14. Correct and Explainable Is Stronger Than Correct Alone

A learner gets the old question right.

Now ask why.

If the learner can explain the relevant relationship, the evidence becomes stronger.

If the learner says, “I remember this one,” the tutor has learned something different.

Do not punish answer memory.

Move to a changed question.

15. Wrong Again Can Be Excellent Evidence

The learner repeats the same mistake.

That can feel discouraging.

But the tutor now knows the correction did not alter the relevant operation strongly enough.

The next move is not necessarily “explain louder”.

Ask why the error returned.

• Was the explanation understood only while visible?
• Was the corrected rule too vague?
• Did the learner fail to recognise when to use it?
• Is an earlier prerequisite still unstable?
• Did the tutor fix the symptom rather than the first weak link?

16. A Different Wrong Answer Can Be Progress

This is one of the most important reattempt readings.

Attempt 1 fails at Step 2.

Attempt 2 passes Step 2 and fails at Step 5.

The final answer is still wrong.

The learner has still moved.

When the first broken link is repaired, a later broken link becomes visible.

Do not collapse all wrong answers into one state.

17. A Faster Correct Answer Can Still Be Weak Evidence

The student answers instantly on Attempt 2.

Fast is not automatically fluent.

It may be recognition.

Ask for the reasoning or use a changed item.

True fluency should increasingly survive when superficial cues change.

18. A Slower Correct Answer Can Be Stronger Evidence

The learner hesitates, reconstructs the idea, checks a condition and then succeeds without help.

That may be better evidence than instant recognition.

The tutor should read the process, not only the stopwatch.

19. Reattempt the First Meaningful Failure, Not Necessarily the Whole Task

If an examination paper contains four informative errors, repeating the whole paper may spend an hour re-proving what was already stable.

A more efficient tutor may select:

• the earliest load-bearing error;
• one downstream question affected by it;
• one changed question that tests the repair;
• one mixed item later.

The purpose is evidence quality, not repetition volume.

20. Reattempt the Whole Task When the Whole Task Is the Skill

Sometimes the problem exists at task level.

• A composition loses coherence across forty minutes.
• A Mathematics paper collapses late because time allocation is poor.
• A Science paper shows stable knowledge but weak switching between MCQ and structured response.
• An oral response begins strongly but loses structure after follow-up questions.

In those cases, a larger reattempt can be justified because the system-level performance is itself being tested.

21. The Prompt Ladder Makes Reattempts More Informative

If Attempt 2 stalls, do not immediately supply the full correction again.

1. Wait.
2. Ask the learner to restate the task.
3. Ask where the old error occurred.
4. Direct attention to one relevant feature.
5. Give a small cue.
6. Name the method or principle.
7. Show a parallel example.
8. Reteach if necessary.

The level at which performance returns tells the tutor how available the repair has become.

22. Reattempts Should Make Support Visible

A learner can use notes, examples, parents, peers, answer keys, search or AI while practising.

The important tutoring question is what those supports did.

If the tutor wants Attempt 2 to measure independent reconstruction, those supports should be removed or explicitly marked.

If the tutor wants Attempt 2 to practise supported correction, those supports may remain.

The tool is not the problem. Ambiguous evidence is the problem.

23. The Reattempt Should Not Become a Memory Trap

Some students become very good at old questions.

They memorise:

• the opening line;
• the numerical route;
• the correct option;
• the tutor’s phrase;
• the location of the hidden trick.

This is why reattempts must eventually leave the old page.

24. Same Item, Changed Item, Mixed Item

Level	Main question
Same item	Can the learner reconstruct the repair?
Changed item	Can the learner carry the same reasoning across changed surface conditions?
Mixed item	Can the learner recognise when this reasoning is needed among competing possibilities?

This three-level ladder protects the tutor from declaring mastery too early.

25. Mathematics Reattempt: Separate Method Selection From Execution

Suppose Denise originally fails a differentiation problem.

The tutor discovers that she knows the chain rule but does not recognise the nested structure independently.

On reattempt, do not say, “Use the chain rule.”

Ask:

What is inside what?

If Denise identifies the nesting and differentiates correctly, selection has improved.

Now change the coefficients and form.

If she still succeeds, the evidence becomes stronger.

26. Mathematics Reattempt: Read the First Divergent Line

Attempt 1:

The student expands incorrectly.

Attempt 2:

The expansion is correct, but a later arithmetic slip appears.

The final answer remains wrong.

The repair still worked.

The tutor should teach the learner to compare lines, not merely answers.

27. English Comprehension Reattempt: Reclassify the Question Before Rewriting the Answer

Beatrice’s first answer describes an event when the question asks why it happened.

The tutor corrects the distinction between description and cause.

On reattempt, the first job is not writing a polished sentence.

It is:

What relationship is this question asking for?

If Beatrice correctly classifies the relationship but selects weak evidence, the first weak link has moved.

28. English Writing Reattempt: Do Not Let the Tutor Become the Ghostwriter

A tutor can improve a composition so thoroughly that the reattempt becomes meaningless.

Instead of asking the learner to reproduce the tutor’s paragraph, isolate one writing operation.

• Rewrite the opening with the same idea but without the model sentence.
• Reorder one paragraph for causality.
• Replace vague verbs with precise ones in a new paragraph.
• Repair one sentence pattern, then use it in a different context.

The artifact may be less polished.

The learner may be more visible.

29. Science Reattempt: Rebuild the Mechanism, Not the Model Answer

Faith’s original answer jumps from condition to outcome.

The tutor supplies the missing mechanism.

On reattempt, Faith should not be asked to reproduce the exact sentence.

Ask her to reconstruct:

condition → relevant concept → mechanism → outcome.

Then change the objects.

If the mechanism survives the new surface, the repair is travelling.

30. Science Reattempt Has a Subject-Specific Owner Too

For Primary 5/6 Science, the detailed decision about whether to repeat the same paper or use a fresh one belongs to How to Decide Whether to Reattempt the Same PSLE Science Paper or Use a New One After Corrections.

That guide distinguishes same-paper retrieval from changed-question transfer and full mixed-paper recognition.

This Tutor Handbook volume keeps the cross-subject tutor principle:

Choose the reattempt condition to match the claim you are trying to test.

31. Examination Reattempt: Reproduce the Condition, Not the Drama

If the original failure happened under time pressure, an untimed second attempt answers only part of the question.

Start untimed if you need to separate knowledge from performance.

Then restore timing gradually.

A Performance Coach may use this progression:

untimed repair → bounded timing → mixed timed set → full performance condition.

The learner should not be thrown back into the original stressful condition before the mechanism is understood.

32. Homework Reattempt: Ask What Happened Between Attempts

A student corrects homework at home and returns with the right answer.

The tutor asks:

• Did you see the answer key?
• Did a parent explain it?
• Did you use a worked example?
• Did you ask AI?
• Did you close everything and try again?
• How long after the correction did you reattempt?

This is not policing.

It is evidence reconstruction.

33. Class 0 · Homework Helper: Reattempt the Start

A Class 0 tutor may not need to reattempt the academic question first.

The first failure may be initiation.

Attempt 1: the learner waited for a parent to organise the task.

Attempt 2: the learner opens the correct book, identifies the next task and begins alone.

That is the reattempt that matters.

34. Class 1 · Explainer: Reattempt the Idea Without the Explanation

The Explainer’s central question is:

Can the learner reconstruct the idea after my explanation disappears?

Ask for a fresh explanation in the student’s own words.

Then apply it.

35. Class 2 · Drill Builder: Reattempt Until the Error Signature Changes

Drill is not “keep doing questions until the page is full”.

The Drill Builder watches whether:

• the target error becomes less frequent;
• self-correction appears earlier;
• speed rises without accuracy collapse;
• the repair survives mixed examples;
• the learner no longer needs the explicit cue.

Once the error signature changes, the drill should change too.

36. Class 3 · Diagnostic Tutor: Reattempt to Discriminate Between Explanations

The Diagnostic Tutor uses Attempt 2 as an experiment.

Suppose two hypotheses remain:

• the learner cannot select the method;
• the learner cannot execute the method.

Name the method but do not demonstrate it.

If performance returns, selection becomes the stronger explanation.

If execution still collapses, the diagnosis changes.

37. Class 4 · Route Designer: Decide Where the Reattempt Belongs

The Route Designer decides:

• immediate or delayed;
• same item or changed item;
• isolated or mixed;
• untimed or timed;
• supported or independent;
• one item or whole task.

The reattempt is not just a question.

It is a scheduled evidence event.

38. Class 5 · Performance Coach: Reattempt Under the Condition That Broke

If knowledge is available but performance collapses under timing, mixing, pressure, answer constraints or long-paper fatigue, the Performance Coach restores those conditions strategically.

Do not keep proving that the learner can succeed in ideal conditions.

Bring back the condition that matters.

39. Class 6 · Learning Architect: Reattempt the System Operation

A Learning Architect may be testing a higher-level operation.

Can the learner independently:

• select which old error deserves reattempt;
• choose when to delay it;
• remove unnecessary support;
• decide when same-item success is no longer informative;
• build or request a changed test;
• record what the result means?

Now the learner is beginning to operate the reattempt machinery themselves.

40. Repair Mode: Reattempt the Load-Bearing Link

In Repair mode, the tutor should test the exact prerequisite that was rebuilt.

If fractions were repaired, do not stop with an isolated fractions worksheet.

Reattempt the downstream ratio or algebra task that used to fail because of fractions.

The strongest repair evidence is often downstream relief.

41. Alignment Mode: Reattempt Inside the Current School World

In Alignment mode, the learner should eventually use the repaired capability on current school work.

The reattempt may be:

• a new teacher worksheet;
• a school correction;
• a class quiz;
• a current chapter problem;
• an upcoming assessment condition.

Tutoring should not become a sealed environment where every reattempt succeeds only on tutor-selected pages.

42. Frontier Mode: Reattempt by Widening the Choice

A strong learner may not need more repetition.

They may need less signalling.

Remove the topic label.

Offer two valid methods and ask for comparison.

Ask the learner to find a counterexample.

Ask where the method stops working.

The frontier reattempt tests intellectual choice, not merely reproduction.

43. The Three-Pax Class Makes Reattempts Visible Without Making Them Public Failures

In a three-student class, one learner can reattempt independently while the tutor works briefly with another.

This creates an important condition:

Can the learner run Attempt 2 when the tutor is not hovering?

The small group can also compare strategies after the work is complete.

Not answers as a leaderboard.

Strategies as visible learning operations.

44. Feedback Is Incomplete Until the Learner Acts on It

Feedback can be accurate, detailed and beautifully phrased.

The reattempt asks whether it changed the next performance.

The Education Endowment Foundation’s August 2026 material on feedback and developing independent learners emphasises giving specific next steps and creating structured opportunities for learners to act on feedback through activities such as redrafting, correcting errors or completing similar tasks.

The reattempt is one of those opportunities.

Feedback becomes evidence of learning only when a later action changes.

45. Retrieval Practice Helps, but Reattempting Is Not One Single Technique

Research on retrieval practice supports the value of actively bringing learned material back from memory rather than relying only on restudy. A 2021 systematic review by Agarwal, Nunes and Blunt examined applied retrieval-practice research in schools and classrooms and found broadly consistent learning benefits across many settings.

But a tutor should not turn that broad finding into a simplistic rule that repeating the same old question proves everything.

The reattempt may involve retrieval.

It may also involve recognition, feedback use, method selection, transfer, fluency, self-monitoring and performance conditions.

The tutor still needs to identify which job is being tested.

46. Transfer Is a Separate Test

A learner can retrieve an old answer without transferring the underlying reasoning to a new context.

That is why the sequence matters:

repair → same-item reattempt → changed item → mixed recognition → performance condition.

Do not skip same-item reattempt if the repair is fragile.

Do not stop there if the educational goal requires flexible use.

47. The Reattempt Builds Metacognitive Calibration

Before Attempt 2, ask the learner to predict:

• Will I remember the method?
• Where might I still get stuck?
• Will I need a cue?
• How confident am I?

Then compare prediction with reality.

The EEF’s August 2026 material on metacognition and self-regulation emphasises planning, monitoring and evaluating learning so that these processes can become increasingly internal over time.

A well-run reattempt gives the learner a concrete event to monitor and evaluate.

48. Attempt 2 Should Often Be Smaller Than Attempt 1

The first task may have been a full page.

The tutor only needs one question to test whether the repair returned.

Smaller reattempts reduce noise.

They also reduce punishment-by-repetition.

49. Attempt 2 Should Sometimes Be Harder Than Attempt 1

Once the original repair is stable, keeping the exact old difficulty can become uninformative.

The tutor can change one dimension:

• different numbers;
• different objects;
• different wording;
• different representation;
• less cueing;
• mixed topics;
• shorter timing;
• greater explanation demand.

Change one thing first if you want a clean interpretation.

50. Do Not Change Everything at Once

If Attempt 1 was untimed, familiar, scaffolded and single-topic, and Attempt 2 becomes timed, unfamiliar, unscaffolded and mixed, failure tells the tutor very little about which change caused the collapse.

A Diagnostic Tutor changes conditions deliberately.

One useful variable at a time when possible.

51. The Third Attempt Should Not Automatically Exist

Students can become trapped in correction loops.

Wrong.

Correct.

Repeat.

Repeat again.

At some point the old item stops providing new evidence.

The tutor should move to a changed context or retire the item to delayed review.

52. Know the Stop Rule

A same-item reattempt has done its job when the learner can:

• reconstruct the repaired reasoning without the original support;
• explain what changed from Attempt 1;
• identify the old error;
• complete the item with acceptable independence.

Then widen the test.

Do not keep polishing familiarity.

53. The Reattempt Can Reveal That the Original Correction Was Wrong

Tutors can misdiagnose.

A correction may fix one example without fixing the mechanism.

If the learner reattempts and the same downstream failure returns, ask whether the tutor repaired the wrong layer.

A mature tutor lets Attempt 2 revise the tutor’s model too.

54. The Reattempt Can Reveal That the Student Needed Less Help Than Expected

Ciara fails Attempt 1.

The tutor plans a full explanation.

Before explaining, the tutor asks one discriminating question.

Ciara notices the error and completes the reattempt herself.

The correct lesson is not “explain more clearly”.

It may be “wait longer and preserve self-correction”.

55. The Reattempt Can Reveal That the Student Needed More Help Than Expected

A learner nods through correction.

The tutor assumes understanding.

Attempt 2 immediately collapses.

Good.

Now the tutor knows the explanation did not become usable.

Re-represent the idea.

Use a simpler example.

Check the prerequisite.

56. Parent Help Should Be Visible on Reattempts

For younger learners, parent support can be appropriate.

But the tutor needs to know what belonged to whom.

A useful note can be simple:

Parent reminded her to begin. Student chose the method and completed the reattempt independently.

Now the evidence remains interpretable.

57. AI Help Should Be Visible Too

A learner may use AI to explain the failed question.

The reattempt should then test what remains after the tool closes.

• Can the learner restate the principle?
• Can they solve the old item without the generated answer?
• Can they solve a changed item?
• Can they detect an incorrect AI explanation if one appears later?

Current EEF EdTech research priorities in 2026 include concern that generative AI may offload cognitive work such as recall, planning, reasoning or drafting. That makes support visibility especially important when tutors interpret reattempts.

58. The Student Should Predict Attempt 2 Before Starting

Ask:

What do you think will happen this time?

The learner may say:

I think I’ll remember the first step, but I may still lose the sign later.

Now the reattempt teaches two things:

• the academic skill;
• the accuracy of the learner’s self-model.

59. The Student Should Review Attempt 2 Before the Tutor Does

After finishing, ask:

What do you think this second attempt tells us?

The learner may notice:

• I remembered the answer but not the reason.
• I could do it without the tutor this time.
• I fixed the first problem but made a later one.
• I still needed the same cue.
• The old question is too familiar now.

That is metacognitive progress.

60. Failure Mode: Calling a Copied Correction a Reattempt

The learner looks at the model answer and rewrites it neatly.

That may be part of correction.

It is not an independent second attempt.

61. Failure Mode: Repeating the Same Item Until Memory Wins

Attempt 2.

Attempt 3.

Attempt 4.

Eventually the learner can reproduce the page.

The tutor mistakes familiarity for transfer.

Once the old item has done its job, leave it.

62. Failure Mode: Moving to a New Question Too Soon

The opposite mistake also exists.

The learner has not yet reconstructed the corrected step on the original problem.

The tutor introduces a harder unfamiliar problem.

Now the evidence becomes noisy.

Stabilise the repair enough to test it cleanly before widening the context.

63. Failure Mode: Treating Every Second Attempt as a Test

Sometimes Attempt 2 is still practice.

The tutor may deliberately keep a scaffold visible.

That is legitimate.

Just do not interpret supported practice as independent mastery.

64. Failure Mode: Punishing the Learner for Failing Attempt 2

If students believe the second attempt must be perfect, they will hide uncertainty.

The tutor loses the evidence.

Attempt 2 should be safe enough to fail honestly.

The expectation is effort, transparency and learning—not theatre.

65. Failure Mode: Using Only Final Answers to Compare Attempts

Attempt 1: wrong.

Attempt 2: wrong.

Looks unchanged.

But the first error may have moved three steps later.

Compare the route.

66. Failure Mode: Giving the Same Prompt Again

If the learner only succeeds when the tutor repeats the exact same prompt, the prompt may have become part of the task.

Fade it.

Change it.

Or ask the learner to generate the prompt themselves.

67. Failure Mode: Reattempting Without a Claim

“Do it again” is not enough.

The tutor should know what a success or failure would mean.

For example:

If Alicia can now build the bar model without prompting, the representation repair is probably available. If she still needs the representation named, selection remains the weak link.

That is a testable claim.

68. Alicia: The Reattempt Moves the Failure Downstream

Alicia’s first word problem fails immediately because she does not identify what must be found.

The tutor repairs one routine:

Unknown first. Relationship second. Calculation third.

On Attempt 2, Alicia identifies the unknown and draws the correct representation.

She then makes an arithmetic error.

Wrong answer.

Better learner state.

69. Beatrice: The Reattempt Exposes Answer Memory

Beatrice reattempts an old comprehension item and produces the corrected answer almost word for word.

The tutor asks why the answer is causal rather than descriptive.

Beatrice cannot explain.

The tutor does not mark the reattempt as failure.

It is recognition evidence.

The next move is a changed passage with the same relationship demand.

70. Ciara: The Reattempt Shows the Prompt Is Still External

Ciara can correct signs whenever the tutor says, “Check the sign.”

On Attempt 2 without that phrase, the sign error returns.

The tutor’s next job is not more sign explanation.

It is transferring the check trigger.

Ciara begins using a self-generated checkpoint after expansion.

71. Denise: The Reattempt Separates Knowledge From Time Cost

Denise solves the old A-Math question correctly on Attempt 2.

Untimed, it takes twelve minutes.

The examination allocation would make that expensive.

The repair is conceptually stronger.

Performance is not yet ready.

The tutor moves from Class 3 diagnostic work toward Class 5 performance coaching.

72. Emily: The Reattempt Becomes Too Easy

Emily repairs an advanced problem and reattempts it correctly.

The tutor could assign three more near-copies.

Instead, Emily is asked:

What would have to change for this method to stop working?

The reattempt has finished.

The frontier opens.

73. Faith: The Reattempt Shows Mechanism Transfer

Faith’s old Science question is about evaporation.

She repairs the missing mechanism.

Attempt 2 on the same item is correct.

The tutor then gives a different phenomenon with the same reasoning demand: condition, mechanism, outcome.

Faith reconstructs the chain.

Now the evidence is no longer only about remembering the old Science sentence.

74. A Simple Reattempt Board

Field	Question
Attempt 1	Where did the first meaningful failure occur?
Intervention	What changed between attempts?
Attempt 2 condition	Independent, prompted, supported, delayed, timed?
First divergence	Where did Attempt 2 differ from Attempt 1?
Recognition risk	Could familiarity be carrying the answer?
Repair status	Can the learner reconstruct what changed?
Transfer status	Does it survive a changed question?
Next move	Repeat, fade support, change context, mix, time or retire?

75. A Five-Minute Reattempt Routine

1. Recall: ask what went wrong last time before showing the correction.
2. Predict: ask what the learner expects to happen now.
3. Reattempt: use the chosen support condition.
4. Compare: find the first line or decision that changed.
5. Decide: same item again, changed item, mixed item, delayed return or move on.

That routine is short enough to become normal tutoring practice.

76. A One-Minute Reattempt Routine

What failed last time? Try it again without the old help. What changed?

Even this simple version preserves more evidence than merely showing the answer again.

77. Twelve Questions Every Tutor Should Ask About a Reattempt

1. What did Attempt 1 reveal?
2. What intervention occurred before Attempt 2?
3. What exactly is Attempt 2 meant to test?
4. How much time has passed?
5. What support remains visible?
6. Could familiarity or answer memory carry performance?
7. Where does Attempt 2 first differ from Attempt 1?
8. Can the learner explain what changed?
9. Did the same error return?
10. Did a new downstream error appear?
11. Should the next test use a changed or mixed context?
12. What support can now be reduced?

78. Twelve Questions Learners Can Ask Themselves

1. What did I do wrong the first time?
2. Do I know why it was wrong?
3. What did the correction teach me?
4. Can I close the correction and try again?
5. Am I solving or remembering?
6. Where did I hesitate this time?
7. Did I need the same prompt?
8. What did I do differently?
9. Can I explain the repair?
10. Can I do a changed version?
11. Can I recognise the skill in mixed work?
12. What part can I now do without help?

79. Twelve Questions Parents Can Ask Without Taking Over

1. Did my child attempt before seeing the answer?
2. Was the original mistake preserved?
3. What changed between attempts?
4. Did my child reattempt independently?
5. Did I supply a method or only protect the work time?
6. Was AI or an answer key used?
7. Could my child explain the corrected idea afterwards?
8. Did the tutor compare the two attempts?
9. Did the tutor move to a changed question when the old one became familiar?
10. Is the same prompt still required?
11. Is support reducing over time?
12. Is my child becoming better at deciding what a second attempt actually proves?

80. Frequently Asked Questions

Should students always reattempt wrong questions?

No. Reattempts are most useful when the error is informative and the tutor knows what the second attempt will test. Some errors need direct reteaching first; some old questions become too familiar; some stable skills are better tested on changed work.

Should the reattempt happen immediately?

An immediate reattempt can show whether the correction was understood well enough to reproduce. A delayed reattempt asks whether the repair remains retrievable later. Both can be useful for different reasons.

How long should we wait?

There is no universal interval. Use enough separation to answer the intended question. Highly memorable items may need greater delay or faster movement to changed contexts.

If the student gets the same question right, is it mastered?

Not necessarily. Same-item success is useful evidence that something from the repair returned. Stronger claims require changed, delayed, mixed or performance-context evidence.

What if the student remembers the answer?

Do not treat remembering as failure. Ask for the reasoning, then move to a changed item if the old answer is too familiar to test the underlying capability.

What if the student fails the same way twice?

Re-diagnose. The correction may not have become usable, the learner may not recognise when to apply it, or an earlier prerequisite may still be unstable. Repeating the same explanation is only one possible response.

What if the second answer is still wrong but the working is better?

That can be meaningful progress. Compare where the first error occurs. A repaired early step can expose a later weakness that was previously hidden.

Should a tutor use a new question immediately?

Not always. First ensure the learner can reconstruct the repair with enough stability to make the changed question interpretable. Then widen the context.

What if AI was used between attempts?

Make the help visible. Then close the tool and test what the learner can reconstruct, explain and transfer independently.

Should parents correct a reattempt?

Only according to the agreed support condition. If the goal is independent evidence, let the learner finish first and preserve the failed point for the tutor.

How do we know when to stop repeating an old item?

When the learner can reconstruct the repair independently and explain what changed, the old item may have done its job. Move to changed, mixed or delayed testing.

What is the final goal of reattempting?

The learner eventually becomes able to inspect a failed attempt, understand the correction, try again honestly, distinguish recognition from learning and decide what evidence is needed next.

81. Evidence and Further Reading

• Education Endowment Foundation — Feedback and Developing Independent Learners (24 August 2026)
• Education Endowment Foundation — Metacognition and Self-Regulation (24 August 2026)
• Education Endowment Foundation — Modelling Independence: The Seven-Step Model (28 April 2026)
• Education Endowment Foundation — Research Agenda Theme: EdTech and Generative AI
• Agarwal, Nunes & Blunt — Retrieval Practice Consistently Benefits Student Learning: A Systematic Review of Applied Research in Schools and Classrooms
• Pan & Rickard — Transfer of Test-Enhanced Learning: Meta-Analytic Review and Synthesis
• eduKate Sengkang — How to Decide Whether to Reattempt the Same PSLE Science Paper or Use a New One After Corrections
• eduKate Sengkang — The Evidence Loop | How We Know Learning Changed
• eduKate Sengkang — Prompt Dependence | When Correct Work Is Not Yet Independent
• eduKate Sengkang — Confidence Calibration
• eduKate Sengkang — How We Know Learning Has Really Held
• eduKate Sengkang — How Independent Learning Works
• eduKateSG — Tutor Classification Model
• eduKateSG — Tutorial Class Control Tower
• eduKateSG — The Three Modes of Tuition
• The Tutor Handbook Vol No.0017 — The Return

82. Evidence Boundary

Research supports retrieval practice, feedback use, metacognitive monitoring, scaffold fading and opportunities to act on correction.

It does not produce one universal rule for how many times every learner should repeat every question.

The right reattempt depends on what failed, what was repaired, how familiar the item is, how much support remains and what educational claim the tutor needs to test next.

That is why the reattempt belongs inside diagnosis and route design rather than becoming a mechanical rule.

83. The Reattempt Operating Cycle

Preserve Attempt 1 → locate the first meaningful failure → repair one useful layer → choose the support condition → reattempt → compare the first divergence → ask what changed → decide whether the evidence is recognition, repair or stronger learning → move to delay, changed context, mixed recognition or performance conditions → reduce support when the learner can carry more.

This is how correction stops being cosmetic.

The learner does not merely replace a wrong answer with a right answer.

The tutor tests whether the operation that produced the wrong answer has changed.

84. Final Compression

The student tries again.

Do not ask only whether the answer is now correct.

Ask what survived from the correction.

Ask what support remained.

Ask whether the learner solved or recognised.

Ask where the two attempts first diverged.

Ask whether the original error disappeared.

Ask whether a later error became visible.

Ask whether the same prompt is still required.

Ask whether the repair survives delay.

Then change the surface.

Change the representation.

Mix the skill with other skills.

Restore the performance condition when appropriate.

And when the learner can do the old question because the old question is now simply old, stop asking it.

A second attempt is not proof that learning happened. It is a carefully chosen opportunity to discover what changed after the first failure.

That is the discipline of the reattempt.

That is Tutor Handbook Volume 0018.

Next in the series: The Tutor Handbook Vol No.0019 | The Changed Question — How a Tutor Tests Whether Learning Travels When the Surface Changes.