Not every error deserves the same repair time. A copied digit may cost one answer once; an unstable idea about what the question is asking can damage several steps, subjects and future papers. The Error-Repair Priority Score helps a G2 SEC learner rank mistakes by impact, recurrence and dependency before deciding what to practise next.
Use it now. For each reviewed mistake, record three ratings from 0 to 2: impact on available credit, recurrence across attempts, and dependency—how many later decisions rely on it. Start with the highest combined score, but use the ratings as a learning guide rather than an examiner rule. Repair the earliest controllable cause, test it on a changed item, and stop once independent transfer is visible.
The 2027 G2 directory lists English Language K200, Mathematics K210 and Science K223/K224/K225. Their tasks and credit structures differ, so this cross-subject routine never replaces a current question, mark scheme, specimen material, teacher feedback or school instruction. MOE states that the SEC replaces N- and O-Level examinations from 2027 while students sit subjects at G1, G2 or G3.
Official source route: SEAB 2027 G2 syllabus directory; applicable 2027 official syllabus; MOE SEC transition context. Recheck these documents for a later examination year.
Choose your route
Route 1 — Understand: Chapters 1–4 establish the governing distinction.
Route 2 — Execute: Chapters 5–9 build a repeatable method.
Route 3 — Diagnose: Chapters 10–13 expose nearby traps and add variation.
Route 4 — Transfer: Chapters 14–15 support adults and independent use.
Chapter contents
1. Separate symptom from cause
3. Measure recurrence across attempts
6. Keep controllability visible
7. Distinguish knowledge from execution
9. Design the smallest decisive drill
11. Update the score after evidence
12. Protect strengths while repairing
13. Use neutral error language
14. Parents and tutors should review patterns
15. Transfer the score across subjects
Series and learning links
Continue with eduKateSG Examinations and Assessment Hub, the canonical owner G1, G2 and G3 Assessment and Feedback owner guide, and Vol 0296 Mark-Value Response-Depth Calibration. This guide adds a narrow diagnostic and practice job rather than duplicating the owner.
For broader habits, use Examination Craft, the earlier PSLE Learner’s Guide series and the immutable eduKateSG small-group tutorial reference.
1. Separate symptom from cause
Mechanism. Separate symptom from cause. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. A missing unit is the symptom; the earlier cause is that the learner did not name the requested quantity before calculating. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Every red mark is copied into one undifferentiated correction list. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Ask what first decision made the later error possible. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
2. Score impact without drama
Mechanism. Score impact without drama. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. An error that invalidates a six-step method receives more attention than a harmless wording preference. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Any mistake on a high-mark question is labelled catastrophic. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Estimate which credit-bearing work the error actually affected. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
3. Measure recurrence across attempts
Mechanism. Measure recurrence across attempts. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. The same sign reversal appears in three mixed sets, so it is more urgent than a one-off slip. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. A single recent mistake is assumed to be the learner’s main weakness. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Use several attempts and contexts before naming a pattern. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
4. Trace dependency forward
Mechanism. Trace dependency forward. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. A wrong table heading corrupts every later graph reading based on that column. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Only the final wrong value is corrected. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Draw arrows from the first unstable decision to the steps that reuse it. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
5. Prefer upstream repairs
Mechanism. Prefer upstream repairs. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. Teaching the learner to identify the required output prevents several unit, formula and conclusion errors. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Each downstream answer is corrected separately. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Fix the smallest earlier decision with the widest useful reach. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
6. Keep controllability visible
Mechanism. Keep controllability visible. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. The learner can control a command-word check but cannot control an unfamiliar surface context. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Feedback focuses on task difficulty rather than an actionable choice. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Phrase the repair as a visible learner action. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
7. Distinguish knowledge from execution
Mechanism. Distinguish knowledge from execution. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. The learner can explain the rule untimed but fails when copying values under time pressure. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. More concept notes are assigned automatically. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Test recall, selection, execution and checking as separate stages. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
8. Use cost-aware triage
Mechanism. Use cost-aware triage. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. A frequent low-cost slip and a rare high-cost misconception can both deserve attention for different reasons. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Only the number of lost marks determines practice order. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Balance impact, recurrence, dependency and repair effort. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
9. Design the smallest decisive drill
Mechanism. Design the smallest decisive drill. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. Three changed items test whether the corrected decision transfers without a worked model. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. The learner completes twenty near-identical questions. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Use enough variation to test the cause, not to exhaust the learner. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
10. Require delayed transfer
Mechanism. Require delayed transfer. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. The repair is retested two days later inside a mixed set. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Immediate success beside the model is treated as mastery. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Remove the cue, delay the test and mix near-neighbour methods. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
11. Update the score after evidence
Mechanism. Update the score after evidence. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. A previously frequent error falls after two independent mixed checks. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. The priority list becomes a permanent label. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Lower or remove a target when current performance justifies it. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
12. Protect strengths while repairing
Mechanism. Protect strengths while repairing. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. The learner keeps one short retrieval item from a secure skill while focusing on the priority cause. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. All practice becomes remediation and confidence narrows. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Maintain a small dose of successful independent work. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
13. Use neutral error language
Mechanism. Use neutral error language. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. “Formula choice failed before substitution” describes a process that can change. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. The learner is called careless or weak. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Name the observable decision, condition and next test. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
14. Parents and tutors should review patterns
Mechanism. Parents and tutors should review patterns. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. An adult asks which error recurred and which later steps depended on it. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. Every red mark produces a new lecture or worksheet. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Hold a short weekly review and choose one or two priorities. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
15. Transfer the score across subjects
Mechanism. Transfer the score across subjects. Make the controlling decision visible before compressing it into a fast routine. A visible decision can be inspected and corrected; an unspoken habit can produce a plausible response for the wrong reason.
Original worked example. A reading omission, algebra setup error and science conclusion gap can all be ranked by impact, recurrence and dependency. The wording and values are created for teaching, not copied from an examination. Identify the cue, apply the relationship and state the conclusion at the precision the evidence supports.
Misconception diagnosis. The learner uses one subject’s correction format mechanically everywhere. Treat this as evidence about the first failed decision, not as a verdict on the learner. Mark that earliest unstable step so correct later work does not need to be rebuilt unnecessarily.
Repair. Keep the triage questions while adapting the repair to the subject. Close the model and reproduce the decision from memory. Change one surface feature, then explain what remained invariant and why the method still fits.
Practice progression. Start untimed with the decision labelled. Remove the label, vary numbers or wording, add a strong near-neighbour distractor and finish with a short realistic time boundary. Record whether the learner chose the move independently.
Explained answer check. A secure answer preserves the governing relationship, relevant evidence, units or language strength, and the exact task demand. Compare assumptions with the prompt instead of choosing the most familiar method.
Transfer. Give a fresh item after a delay without naming this chapter. Success means the learner detects the same structure, completes the task and can justify the choice. If transfer fails, return to the earliest decision rather than adding identical repetition.
A seven-day practice plan
Day 1: learn the distinction and complete two guided examples.
Day 2: diagnose three deliberately flawed responses.
Day 3: practise a varied set without chapter labels.
Day 4: explain one answer aloud and shorten the explanation without losing its reason.
Day 5: use a timed mini-set and record the first failed decision.
Day 6: repair that decision, then attempt a changed transfer item.
Day 7: complete a delayed mixed check and write one next-step target.
Final learner checklist
□ I can state the exact task before I start.
□ I can identify the decision that controls the later steps.
□ I can show enough working or notes to make that decision visible.
□ I can diagnose the first wrong step instead of only comparing final answers.
□ I can complete a changed example without copying the model.
□ I follow current paper and school instructions rather than treating this guide as an examiner rule.