Bukit Timah Tuition OS — Article 5

A child can sit through the same explanation three times and still be unable to start the question alone. Another can understand the idea immediately but produce a messy first attempt, then repeat the same mistake for weeks. A third can finally solve the problem correctly with a tutor beside them, yet lose the method as soon as the wording changes. These are not reasons to “teach the whole topic again”. They are reasons to repair the exact part of the learning chain that is still unstable.

Bukit Timah Tuition OS — Article 5 keeps the original eduKateSG “Takeoff Repair” idea and rebuilds it for parents, students and readers. The metaphor is simple: after Article 3 has located where a skill breaks and Article 4 has shown what happens when performance collapses under load, Article 5 asks the practical question—how do we rebuild one weak skill so the student can actually start, complete, explain and repeat it independently?

The answer is not “explain more” by default. Good repair separates six jobs: identify the exact missing piece, rebuild the smallest useful model, show or study one clear example, guide one successful attempt, fade the support, then test whether the student can reproduce the skill under a small change. Only after that should we add wider variation, delay, mixing and time pressure.

This article uses “takeoff” as a metaphor for the first reliable independent use of a repaired skill. It is not a scientific or clinical category. The educational principles around it are familiar: clear models, worked examples, manageable cognitive load, active explanation, deliberate practice, retrieval, variation, feedback and gradual transfer of control from adult to learner.

What counts as a repaired skill?

A repaired skill is not simply a question that has been corrected. It is not a method that the student can copy while the example remains visible. It is not an answer that becomes correct after the tutor gives the crucial first step. Those are stages of repair, but they are not yet proof of ownership.

For a narrow academic skill, a useful minimum standard is that the student can:

recognise what the question is asking;
choose or retrieve an appropriate method;
execute the method accurately enough;
explain the core reason or sequence in their own words;
complete a fresh related problem without the original support;
survive at least one small variation.

That is the parent-readable version of the original “one clean success cycle”. It is deliberately modest. It does not mean the student has mastered the entire topic, retained the skill for six months or become examination-proof. It means the first repair has genuinely taken hold and can now be strengthened.

Repair begins with a narrow target, not a broad subject label

“Weak at Mathematics” is too broad to repair. “Weak at algebra” is still broad. “Loses negative signs when expanding brackets” is narrow enough to work on. “Cannot tell whether to factorise or use the quadratic formula” is a different target. “Can solve equations but cannot translate word problems into equations” is another.

The same rule applies outside Mathematics. “Weak in English” could mean sentence control, evidence selection, inference, task fulfilment, planning, vocabulary precision, reading pace or paragraph development. “Weak in Science” could mean recall, causal explanation, data interpretation, formula selection, unit control or applying known concepts to unfamiliar contexts.

A useful repair target usually contains an observable verb: identify, translate, select, explain, compare, calculate, plan, retrieve, classify, check, revise. Observable targets let parents and tutors see whether the intervention changed anything.

The first repair decision: concept, prerequisite, method, retrieval or representation?

Before teaching again, decide what is actually missing. Article 3 provides the diagnostic logic. Article 5 now turns that diagnosis into a repair route.

If the concept is missing

Rebuild meaning before speed. Use a concrete example, diagram, analogy where helpful, counterexample and explanation. Ask what the idea represents and why the method works. Avoid burying the student under ten procedural steps before the core relationship is visible.

If a prerequisite is missing

Repair the older skill that is creating friction. A student may understand calculus but lose control because algebraic fractions are weak. A child may understand ratio but calculate too slowly because multiplication facts are fragile. A writer may understand paragraph structure but lose clarity because sentence boundaries are unstable.

If method selection is missing

Do not teach the method again immediately. Teach how to recognise when it applies. Compare similar-looking questions requiring different methods and different-looking questions requiring the same method. Practise classification before execution.

If retrieval is missing

The student may understand perfectly when notes are open but fail to recall the formula, structure or first step later. Use closed-book retrieval with feedback and spaced return rather than another long explanation.

If representation or translation is missing

The student may know the idea but fail to convert text to diagram, diagram to equation, evidence to inference, data to explanation, or prompt to writing plan. Insert an intermediate representation and practise the conversion explicitly.

Correct diagnosis makes the repair smaller. That is one of the main advantages of the Tuition OS approach: instead of “do the chapter again”, the student works on the mechanism that is actually costing performance.

The Minimum Working Model: the smallest explanation that lets the student act

A common teaching mistake is to respond to confusion with more and more explanation. Sometimes that is useful. Sometimes it increases load. The Minimum Working Model asks a stricter question: what is the smallest accurate model that lets the learner make the next correct decision?

A useful model usually answers three things:

What am I trying to find or produce?
What structure or relationship is present?
Why does this method fit that structure?

For completing the square, the model may be: “I am rewriting the quadratic into a form that exposes its turning point.” For an inference question: “I am using what the text shows to reach a conclusion it implies but does not state directly.” For a Science explanation: “I need a chain from the changed condition to the mechanism to the observed result.”

The model can grow later. At takeoff stage, usefulness matters more than encyclopaedic completeness.

Worked examples: show decisions, not just steps

A worked example is strongest when it makes the hidden decisions visible. Students often see a polished solution and copy its surface without understanding why one route was chosen.

When modelling a repair, narrate the decisions that matter: “I notice the question asks for the turning point, so this form will help.” “The phrase ‘despite’ signals contrast, so I expect the surrounding sentences to pull in different directions.” “The graph shows a rise while the controlled variable remains fixed, so the explanation must connect the changed variable to the mechanism.”

Also show one likely error and why it is tempting. This is more useful than presenting perfection as effortless. Students need to know where the method tends to break and what cue will catch the break.

The What Works Clearinghouse practice guide on organising instruction and study includes recommendations to interleave worked examples with problem-solving exercises and to connect abstract and concrete representations. The practical implication here is straightforward: do not leave the student watching examples for too long. A model should lead quickly into an active attempt.

The completion problem: the bridge between watching and doing

For a genuinely new or unstable procedure, the jump from full worked example to completely independent problem can be too large. A completion problem keeps some support while requiring the student to generate the critical step.

In Mathematics, the first line might be supplied while the student chooses and executes the next steps. In English, a paragraph may include the evidence while the student writes the inference and explanation. In Science, the condition and mechanism may be supplied while the student completes the causal chain.

The completion stage is temporary. Its purpose is to reduce unnecessary load while the sequence becomes organised. Once the student can supply the missing decisions, remove more support.

The first independent attempt: where repair becomes real

The first fully independent attempt should be close enough to the worked example that the student can focus on the repaired mechanism, not on several new complications. This is not “making it too easy”. It is controlled testing.

During this attempt, the adult should resist completing the thinking. If a prompt is needed, record the level of prompt. A student who succeeds after “What type of problem is this?” is closer to independence than a student who needs the first three steps demonstrated.

If the student succeeds, ask them to explain the route. If they fail, identify the first point of breakdown. Do not automatically restart the entire explanation. Repair the exact break and attempt again.

Explain-back: ownership without performance theatre

Explain-back is useful because correct answers can hide fragile knowledge. The student should be able to say what the task required, why the method fit, what the main steps were, where an error could occur and how the answer can be checked.

But explain-back should not become a speech contest. A student does not need elegant adult language. The purpose is to expose the internal model. A short, clumsy but accurate explanation can be excellent evidence.

As expertise grows, explanations often become more compressed. That is fine. The question is whether the student can still unpack the reasoning when needed.

The first variation: test whether the repair has a spine

After one clean independent success, change one meaningful feature. Change the numbers, reverse the direction, alter the wording, change the representation, or insert one common trap. Do not change everything at once.

If the skill collapses immediately, the first success was still closely tied to the example. That is not a disaster. Compare the two problems and ask what stayed the same. The student is now learning the structure rather than only the template.

When the first variation succeeds, the skill has more evidence behind it. It is ready for the next phase—Article 6’s “climb”, where repeated and varied practice builds stability.

Scaffolding: give enough support to learn, then remove it

Scaffolding includes worked examples, diagrams, checklists, prompts, sentence frames, step maps, formula cues and partially completed solutions. These tools can be extremely useful because they reduce unnecessary load while a new structure is being learned.

The danger is leaving the scaffold in place so long that success depends on it. A child can appear highly competent while the worksheet, tutor or parent is carrying the first decision.

Plan the fade. Full model → completion problem → open prompt → self-prompt → independent problem. The exact sequence varies, but the direction should be toward student control.

The prompt ladder

Open prompt: “What is the question asking?”
Focused prompt: “Which part of the question tells you what kind of method may fit?”
Method cue: “Which of these two methods matches the structure?”
Partial step: adult supplies the first representation or first move.
Full model: adult demonstrates the route again.

Use the least support that allows productive progress. Then re-test later with less. The goal is neither “never help” nor “always rescue”. It is calibrated transfer of control.

Error-specific practice: repair the mistake, not the chapter

Once the first clean cycle exists, repeated errors can be trained directly. A student who repeatedly loses negative signs does not need a whole new algebra course. A writer who omits required content points needs task-fulfilment practice. A Science student who knows facts but fails causal chains needs explanation practice.

Use small sets built around the error family. Keep other demands relatively stable so the learner can focus on the repair. Then reinsert the skill into mixed work. Micro-practice is only useful if it transfers back to the real task.

Anti-error cues: build a prevention mechanism

A correction tells the student what was wrong. An anti-error cue tells the student how to catch the same failure next time.

Examples: “negative outside bracket → mark every term before expanding”; “inference → evidence first, conclusion second”; “Science numerical answer → quantity + unit”; “situational writing → tick every required content point”; “ratio before/after → mark what stays constant”.

The cue should be short enough to use under pressure. If it requires half a page of instructions, it will not survive an examination.

When a backup method helps—and when it adds unnecessary load

The original Article 5 recommended “method redundancy”: having an alternative route when one method stalls. This can be valuable in some topics. Simultaneous equations can be solved by elimination or substitution. Quadratics may allow factorisation or the formula. A graph can sometimes provide a check on an algebraic result.

But backup methods should not be introduced automatically at the first sign of weakness. A novice who is already confused may experience more overload if two competing methods are taught simultaneously. Stabilise one clear route first unless comparison itself is needed to understand the concept.

Add redundancy when the primary method is stable enough that the second method increases flexibility rather than confusion.

Retrieval: can the repaired skill start again tomorrow?

A repair session can create a strong immediate performance because the method is still active. The next test is delayed return. Close the notes and revisit the skill the next day or after several days.

If the student needs only a small cue, fade it and try again later. If the entire method has disappeared, the repair needs more retrieval and perhaps better organisation. If the student remembers the idea but not the first step, train the start cue specifically.

Spacing and retrieval are useful because durable learning depends on more than immediate familiarity. A repaired skill should eventually reappear without the original teaching context.

Timing: do not compress a repair before it can stand

The first repair should prioritise correctness, structure and independence. Once those are reasonably secure, realistic timing can be added. Timing too early can teach the student to associate the skill with urgency and error.

Use short timed micro-sets rather than full papers at first. Measure where time is spent. If the student understands but hesitates at method selection, train selection. If execution is slow, train the relevant procedure. If checking consumes too much time, build a targeted check.

The final standard is not “fast”. It is “fast enough for the real task without losing the repaired control”.

A clean correction cycle

Identify the exact error.
Explain or reconstruct why it happened.
Review the correct principle or method.
Close the model or notes.
Complete a fresh related item independently.
Return after a delay.

Copying the correct answer beside the wrong one is not a full correction. The fresh independent attempt is the moment where the student proves that the correction changed behaviour.

Why “one clean success” should come before large practice sets

If the method is still wrong, ten repetitions can automate the wrong method. If the student still misreads the task, thirty questions create thirty opportunities to practise the misreading. Volume is useful only after the core route is good enough to reinforce.

This is why Article 5 places a clean cycle early. The student first needs evidence that the repaired method can work. Then Article 6 can build “bind thickness” through repetition, variation and anti-error practice.

Worked repair 1: Primary 2 subtraction with regrouping

Problem: the child can subtract when no regrouping is needed but becomes confused when a ten must be exchanged.

Minimum model: tens and ones are different place-value units; one ten can be exchanged for ten ones without changing the total value. Use blocks or a drawing. Show one example physically, then connect the exchange to the written notation.

Completion problem: set up the columns and show the exchange, but let the child complete the subtraction. Independent problem: a similar calculation with different numbers. Variation: a word problem that requires the same operation.

The repair is successful when the child can explain the exchange, perform it independently and recognise when subtraction with regrouping is needed.

Worked repair 2: Primary 4 fraction comparison

Problem: the child memorises denominator procedures but gives implausible answers about fraction size.

Minimum model: a fraction is a location or amount relative to a whole. Use 0, 1/2 and 1 as benchmarks. Model 3/4 and 5/8 on a number line. Ask which is closer to one.

Independent problem: compare two new fractions first by estimate, then verify. Variation: use a shaded diagram or measurement context. The student should increasingly connect symbol, image and magnitude.

Worked repair 3: Primary 6 ratio before-and-after problems

Problem: the student handles direct ratios but collapses when one quantity changes.

Minimum model: identify what stays constant before building the ratio model. The invariant quantity is the anchor. Draw the before and after states separately and mark the unchanged quantity.

Worked example: narrate why the invariant matters. Completion problem: student identifies the invariant and fills the model. Independent problem: new numbers. Variation: different quantity changes.

The anti-error cue is short: “Before/after ratio → what stayed the same?”

Worked repair 4: Secondary 1 negative signs in algebra

Problem: the student understands expansion but repeatedly loses negative signs.

Minimum model: the multiplier outside the bracket applies to every term inside. Mark the multiplier visually. Model one correct expansion and one tempting wrong expansion. Ask the student to explain the difference.

Independent problem: one negative bracket. Variation: two brackets, then substitution with a negative value. Anti-error cue: “negative multiplier → touch every term.”

Once the sign routine is stable, return to mixed algebra so the cue appears only when needed.

Worked repair 5: Secondary 2 linear graphs

Problem: the student can read y = mx + c but cannot find an equation from a graph.

Minimum model: gradient describes change in y per change in x; the intercept is the value of y when x = 0. Show the same line as equation, table and graph. Move between the representations deliberately.

Completion problem: intercept is marked, student calculates gradient. Independent problem: neither is marked. Variation: interpret gradient in a real context.

Worked repair 6: Secondary 3 quadratic method selection

Problem: the student knows factorisation, completing the square and the formula but chooses inefficiently.

Minimum model: method choice depends on structure and purpose. Compare three quadratics and ask which route is most natural. No solving yet. Build cues: obvious integer factors, need for turning-point form, non-factorisable roots.

Independent cycle: classify five equations before solving two. Variation: include a question whose requested form changes the best method. The repaired skill is decision-making, not formula recall.

Worked repair 7: Additional Mathematics functions

Problem: f(3) is fine; f(x + 1) and f(g(x)) create confusion.

Minimum model: a function takes an input and produces an output. The entire expression in the brackets is the input. Use a function-box diagram. For composition, draw two boxes in sequence.

Completion problem: boxes are drawn, student substitutes. Independent problem: notation only. Variation: distinguish f(x + 1) from f(x) + 1 and explain why.

Worked repair 8: English comprehension inference

Problem: the student either copies the text or invents conclusions not supported by evidence.

Minimum model: inference = evidence + justified conclusion. Use two columns: “what the text shows” and “what I can reasonably conclude”. Model one answer that is too literal and one that over-infers.

Independent problem: short passage with clear evidence. Variation: different genre or subtler attitude. Anti-error cue: “What proves my inference?”

Worked repair 9: English situational writing task fulfilment

Problem: fluent language, missing content points or wrong tone.

Minimum model: before writing, identify role, reader, purpose, required content and tone. Build a compact task scan. Model one prompt and show how the relationship between writer and reader affects language choices.

Independent cycle: scan a fresh task and produce a short plan. Variation: same purpose but different audience. Only after task control is stable should full writing be the main practice unit.

Worked repair 10: English composition planning

Problem: student has ideas but spends too long deciding what to write.

Minimum model: conflict → turning point → ending, supported by two or three scene beats. Demonstrate a compact plan that is sufficient but not exhaustive.

Independent cycle: create a plan for a new prompt. Variation: another prompt with a fixed three-minute planning window once the routine is understood. The repaired skill is planning throughput, not literary brilliance.

Worked repair 11: Science causal explanation

Problem: student knows keywords but cannot connect them.

Minimum model: changed condition → mechanism → intermediate effect → observable result. Use arrows first. Then convert the chain into a sentence.

Independent cycle: familiar scenario. Variation: same principle in an unfamiliar context. The student should regenerate the chain rather than memorise one model-answer sentence.

Worked repair 12: Science formula selection

Problem: student knows formulas but chooses the wrong one in mixed work.

Minimum model: scenario → variables → relationship. Present several short situations and ask only which quantities are given, what must be found, and which formula connects them.

Independent cycle: one full calculation after classification. Variation: include irrelevant information. This repairs recognition rather than memorisation.

Worked repair 13: JC H2 Mathematics algebra bottleneck

Problem: advanced concept is understood, but algebraic manipulation creates errors and consumes time.

Minimum repair: isolate the exact manipulation—fractions, indices, logarithms, expansion, rearrangement. Model one efficient route and one common failure. Use short fluency practice.

Return to context: immediately place the repaired operation back inside the H2 problem. The repair counts only when it improves the advanced task.

Worked repair 14: JC General Paper paragraph reasoning

Problem: student has examples but paragraphs read like lists.

Minimum model: claim → reason → mechanism → example → implication. Use one example only and force the causal connection to become explicit.

Independent cycle: build a new paragraph skeleton on another topic. Variation: add qualification or counterargument. Content reading remains useful, but the repair target is argument architecture.

A seven-stage repair sequence parents can remember

Name the exact skill.
Rebuild the smallest useful model.
Study one clear worked example.
Complete one guided or completion problem.
Complete one fresh independent problem.
Explain the route and complete one small variation.
Return later and prove the repair again.

This is the public-facing “takeoff” sequence. It is compact enough to use, but rich enough to prevent the most common repair mistake: confusing exposure with ownership.

How Article 5 connects to Articles 2, 3, 4 and 6

Bukit Timah Tuition OS — Article 2 helps parents read the learning signals. Article 3 isolates where a skill breaks. Article 4 explains how performance collapses under added demand. Article 5 repairs the first weak link. Article 6 then builds stability through repeated, varied and anti-error practice.

The sequence matters. Diagnosis before repair. Repair before heavy repetition. Stable direct performance before strong time pressure. Maintenance after success so the repair does not drift away.

Evidence-aware foundations

The “takeoff” vocabulary is an eduKateSG metaphor rather than a formal scientific construct. The instructional logic draws on established teaching and learning principles: worked examples can reduce unnecessary load during early learning; students benefit from active problem solving rather than passive exposure alone; retrieval and spacing strengthen durable access; varying examples can support discrimination and transfer; feedback should be actionable; and scaffolding should increasingly hand control to the learner.

Readers who want formal background can begin with the U.S. Institute of Education Sciences / What Works Clearinghouse practice guide Organizing Instruction and Study to Improve Student Learning, and the Education Endowment Foundation’s resources on Metacognition and Self-Regulated Learning, Feedback and Cognitive Science Approaches in the Classroom.

Deep Repair Casebook: How One Weak Skill Is Rebuilt Across Primary, Secondary and JC

The strongest way to understand Article 5 is through cases. Each case begins with a narrow breakdown, not a broad judgement. Then the repair moves through the same logic: find the missing model or control, reduce unnecessary load, create one clean success, fade support, introduce a small variation and schedule a return.

The cases deliberately cover different subjects and ages because “repair” is not one teaching trick. Sometimes repair means rebuilding meaning. Sometimes it means restoring a prerequisite. Sometimes the student already knows the content and only needs a better selection rule, representation, checking cue or retrieval schedule.

Case 15: Primary 1 number bonds—repairing counting dependence

Observed problem: the child answers correctly but counts every unit one by one. This is not a failure of accuracy. It is a fluency and relationship problem that may create later load.

Minimum working model: numbers can be broken and recombined in useful ways. Ten is a powerful anchor. Show 8 as “10 minus 2”, 6 + 4 as a ten-pair, and 7 + 8 as 7 + 3 + 5. Use objects if useful, then pictures, then mental relationships.

Worked example: model 8 + 7 by moving 2 from 7 to make 10, leaving 5. Narrate the decision, not merely the answer. Completion problem: show 9 + 6 with a ten-frame and ask the child to decide how much moves. Independent problem: 8 + 5 with no frame. Variation: use the same relationship inside a simple story.

The repair target is not speed for its own sake. It is a more compressed number representation that eventually frees attention for multi-step work.

Case 16: Primary 2 multiplication—repairing repeated-addition-only thinking

Observed problem: the child understands multiplication only as counting every group and becomes slow as numbers grow.

Minimum working model: multiplication represents equal groups and can also be viewed as an array. Show 4 × 6 as four groups of six, six groups of four and a 4-by-6 rectangle. Connect the representations rather than presenting separate facts.

Repair sequence: build one array, write the two related multiplication statements, connect to a known fact, then remove the array. Variation: ask a missing-factor question. This begins the bridge toward division without requiring a second isolated memorisation system.

Case 17: Primary 3 division—repairing fact-family disconnection

Observed problem: 6 × 7 = 42 is fluent but 42 ÷ 6 creates hesitation.

Minimum model: the same relationship can be read in several directions. Use a fact family rather than introducing division as an unrelated set of facts. “If six groups of seven make 42, then 42 can be split into six groups of seven.”

Practise four statements from one relationship. Then hide one quantity. After one clean cycle, return several days later with a mixed multiplication/division set. The repair is stable when the learner spontaneously uses the known relationship rather than rebuilding from counting.

Case 18: Primary 3 English—repairing weak sentence boundaries

Observed problem: ideas are strong but sentences run together. The child writes long strings joined by “and”, making meaning difficult to follow.

Minimum model: a sentence carries one complete thought around a subject and what is said about it. Read a short run-on aloud and ask where the voice naturally finishes one idea. Mark the boundary. Show how punctuation helps the reader see the structure.

Completion problem: give a short paragraph with boundaries marked but punctuation missing. Independent problem: child writes three short connected sentences from a picture. Variation: combine two sentences deliberately with a conjunction. The repair grows from control before complexity.

Case 19: Primary 4 Mathematics—repairing unit confusion in measurement

Observed problem: calculations are often correct but the student chooses centimetres, square centimetres and cubic centimetres inconsistently.

Minimum model: unit type follows what is being measured: length, area or volume. Use a line, a square and a cube. Ask what one unit “looks like” in each case.

Then connect the visual to notation. Anti-error cue: “line, surface or space?” Independent cycle: classify units before calculating. Variation: give a correct numerical answer with the wrong unit and ask the student to repair it.

Case 20: Primary 4 English—repairing literal-only comprehension

Observed problem: the student can retrieve stated facts but struggles when the answer is implied.

Minimum model: some questions ask “what does the text say?” and others ask “what does the text show?” Use a tiny example: “Maya pulled her coat tighter and rubbed her hands.” What is stated? What can be inferred?

Completion problem: evidence is highlighted and the student writes the conclusion. Independent problem: student identifies both evidence and inference. Variation: use a different emotion or relationship. This repairs the evidence-to-conclusion bridge before longer passage work.

Case 21: Primary 5 Mathematics—repairing decimal place-value errors

Observed problem: the student treats decimals as whole numbers with a dot and claims 0.37 is larger than 0.5 because 37 is larger than 5.

Minimum model: decimal positions represent tenths, hundredths and smaller parts of one whole. Use a hundred grid or number line. Rewrite 0.5 as 0.50 and compare equal place-value columns.

Independent cycle: compare several decimals with explanation. Variation: place them on a number line or convert simple decimals to fractions. The repair must connect symbol to magnitude, not merely introduce a comparison trick.

Case 22: Primary 5 Science—repairing “keyword dumping”

Observed problem: answers contain correct scientific words but do not explain the mechanism.

Minimum model: a good explanation contains a relationship. Use arrows: cause → change → effect. Ask the student to place each keyword on the chain and add the verb connecting it to the next idea.

Independent cycle: build a chain for a familiar scenario. Variation: use the same concept in a new situation. The repair is complete enough for takeoff when the student can regenerate the relationship rather than recite a stored sentence.

Case 23: Primary 6 Mathematics—repairing percentage reverse problems

Observed problem: forward percentage questions are easy; original-value questions fail.

Minimum model: identify which quantity represents 100%. In a 20% discount where the sale price is $80, the $80 represents 80%, not 100%. Draw a percentage bar before calculating.

Worked example: show 100% → 80% relationship. Completion problem: student labels the percentage bar. Independent problem: different discount. Variation: increase instead of discount. The structure, not the shopping story, is the target.

Case 24: Primary 6 English—repairing situational writing tone

Observed problem: all messages sound the same regardless of audience.

Minimum model: tone follows relationship and purpose. Compare the same request sent to a close friend and to a school principal. Highlight greeting, modal verbs, degree of directness and closing.

Completion problem: choose between two phrasings for a given audience. Independent problem: write three sentences for a new task. Variation: same purpose, different reader. The student learns to control tone as a decision rather than memorise one formal template.

Case 25: Secondary 1 Mathematics—repairing algebra translation

Observed problem: equations can be solved once written, but verbal statements are translated incorrectly.

Minimum model: identify the quantities and relationship before choosing operations. Compare “three less than twice x” with “twice three less than x” and discuss order. Use a bar or verbal paraphrase before writing algebra.

Repair set: translate only—no solving. This isolates the skill. Then place translation back into complete word problems. The repair has transferred when the student forms the correct equation without an adult naming the operation.

Case 26: Secondary 1 English—repairing weak evidence selection

Observed problem: the student understands the passage orally but quotes too much or selects evidence that does not directly support the answer.

Minimum model: evidence must do a job. State the answer idea first, then ask which exact phrase proves it. Use a “claim → evidence” two-column structure.

Completion problem: claim supplied, student chooses evidence. Independent problem: both claim and evidence generated. Variation: inference rather than literal retrieval. This builds precision before full comprehension sets.

Case 27: Secondary 2 Mathematics—repairing simultaneous equations method choice

Observed problem: the student knows elimination and substitution but uses whichever was taught most recently.

Minimum model: method choice should reduce work. Compare equations where one variable is already isolated with equations whose coefficients align easily. Ask which method creates fewer operations.

Repair set: classify five pairs of equations without solving. Then solve two using the chosen method. Variation: ask the student to solve one problem by both methods and compare efficiency. Redundancy now increases flexibility because one route is already stable.

Case 28: Secondary 2 English—repairing language-effect explanations

Observed problem: the student identifies a quotation and writes “this makes the reader interested” regardless of the language.

Minimum model: word choice → meaning/connotation → specific impression or effect in context. Take one word and unpack its literal meaning, associations and role in the sentence.

Completion problem: quotation supplied, student explains meaning. Independent problem: student selects and explains. Variation: use imagery, verb choice or sound rather than the same language feature. The repair is analytical reasoning, not memorised effect vocabulary.

Case 29: Secondary 3 Mathematics—repairing quadratic formula substitution

Observed problem: formula is memorised, but coefficients—especially negative b—are substituted incorrectly.

Minimum model: identify a, b and c from standard form before touching the formula. Write them explicitly with signs. Model why b can be negative and why “−b” may therefore become positive.

Anti-error cue: “write a, b, c first”. Independent cycle: one equation with negative b. Variation: equation not initially in standard form. This repairs the substitution control rather than the formula itself.

Case 30: Secondary 3 Additional Mathematics—repairing completing-the-square meaning

Observed problem: the student follows a memorised sequence but cannot explain why the constant is added and subtracted or how the final form reveals the turning point.

Minimum model: we create a perfect-square expression without changing the original value. Use a simple numeric or area interpretation if helpful, then connect the final form to the graph.

Model one example slowly. Then give a completion problem where the student supplies the half-coefficient and square. Independent problem follows. Variation: ask for the turning point from the completed-square form without further expansion.

Case 31: Secondary 3 Science—repairing unit conversion inside formulas

Observed problem: formula selection is correct but numerical answers are wrong because units are inconsistent.

Minimum model: the formula assumes compatible units. Before substitution, write quantity, value and unit in a small list. Convert before calculating.

Anti-error cue: “units before numbers”. Practise classification and conversion separately, then return to full calculations. Variation: include one question where no conversion is needed so the student must decide rather than convert automatically.

Case 32: Secondary 4 Mathematics—repairing move-on decisions

Observed problem: the student knows enough to pass comfortably but loses large amounts of time on one difficult problem.

This is a performance-control skill rather than a content gap. Minimum model: examination time is a limited resource; difficult questions must not consume the marks available elsewhere. Teach a move-on rule based on time, progress and accessible marks.

Practise in short timed sets containing one deliberately sticky item. The clean success is not necessarily solving that item; it may be recognising the trap, taking available marks, moving on and finishing the set.

Case 33: Secondary 4 English—repairing summary compression

Observed problem: relevant points are found, but answers are too long because the student copies examples and explanation.

Minimum model: identify the core proposition and remove illustration, repetition and decorative detail. Model one sentence and show what can be deleted without losing the point.

Completion problem: core point highlighted, student compresses. Independent problem: select and compress. Variation: combine two related points without duplication. This repairs economy, not just paraphrasing vocabulary.

Case 34: Secondary 4 Science—repairing data-to-explanation linkage

Observed problem: graph is described correctly and concept is known, but the answer fails to connect them.

Minimum model: observation first, mechanism second, explicit bridge between them. “As X increases, Y decreases. This is because…” Model the separation between what the graph shows and why the science explains it.

Independent cycle: a familiar data set. Variation: reverse trend or two data series. The student should not substitute memorised theory for the actual evidence.

Case 35: JC H2 Mathematics—repairing partial fractions setup

Observed problem: the student can solve coefficients once the decomposition is written but chooses the wrong form for repeated or quadratic factors.

Minimum model: decomposition form follows denominator structure. Build a classification table: distinct linear factors, repeated linear factors, irreducible quadratic factors. Practise writing the form without solving coefficients.

Then complete one full problem. Variation: mix denominator types. The repaired micro-skill is setup selection, which reduces errors before any algebra begins.

Case 36: JC H2 Mathematics—repairing integration method selection

Observed problem: the student knows substitution, parts and algebraic simplification but hesitates in mixed work.

Minimum model: choose based on structure. Is there a composite expression with its derivative? A product where one factor simplifies when differentiated? An algebraic form that should be rewritten first? Use classify-only items.

Independent success means the student can justify the chosen route before calculating. Variation includes near-miss questions that look similar but need different approaches.

Case 37: JC General Paper—repairing thin evaluation

Observed problem: the student makes clear claims and gives examples but rarely qualifies conditions or weighs limits.

Minimum model: evaluation asks “how far, under what conditions, for whom, compared with what?” Model one claim and add a boundary condition. Show how evaluation changes the precision of the argument rather than merely adding “however”.

Independent cycle: student adds one meaningful qualification to a fresh claim. Variation: compare two contexts where the judgement changes. This repairs evaluative reasoning rather than collecting more examples.

Case 38: vocabulary—repairing recognition without production

Observed problem: the student says “I know that word” when reading but cannot define or use it accurately.

Minimum model: ownership includes meaning, grammatical role, common partners and appropriate register. Take one word, compare a correct and awkward sentence, then have the student produce a simple accurate sentence.

Variation: new context. Delayed return: use the word several days later without the original list. Repair succeeds when the word becomes available for purposeful use rather than only recognition.

Case 39: examination checking—repairing a ritual that catches nothing

Observed problem: the student says they “checked” but the same sign, unit or task-fulfilment errors remain.

Minimum model: checking must target known risks. Build a two- or three-item routine from actual errors. Demonstrate it on a completed solution containing one planted error.

Independent cycle: student applies the check to their own fresh work. Variation: timed use. The routine survives only if it is short enough to use and specific enough to find real mistakes.

Case 40: study planning—repairing vague revision

Observed problem: the student writes “revise Math” or “study Science” and then spends time rereading whatever is easiest.

Minimum model: a study block needs a target, method and proof. “Retrieve five formulas, practise three mixed method-selection questions, then check one delayed ratio repair” is actionable.

The clean success is a student who can choose a small learning job and verify whether it changed. This is a metacognitive repair, not a subject-content repair.

Repair Playbook 1: when the model is missing

Use fewer problems initially. Rebuild the idea using one or two representations. Ask the student to compare an example and a non-example. Ask “what changes, what stays the same, and why?” Then create a direct independent problem.

Do not judge concept repair by whether the student can repeat the teacher’s sentence. Judge it by whether the model helps them make a correct decision in a fresh situation.

Repair Playbook 2: when the prerequisite is missing

Identify the smallest lower-level operation causing repeated friction. Practise it briefly in isolation until accuracy and fluency improve. Then reconnect it immediately to the higher-level task.

A prerequisite repair that never returns to context can become another disconnected worksheet skill. The proof is downstream improvement.

Repair Playbook 3: when the first step is missing

Many students understand once started but cannot initiate. Build an auto-start question: “What is asked? What type is this? What first representation or method fits?” Practise only the start across several items.

The exercise can stop after the first step. This sounds incomplete, but it isolates initiation. Once starts are reliable, restore full solutions.

Repair Playbook 4: when the sequence collapses

Use a short numbered step map. Ask the student to reconstruct it from memory before solving. Highlight anchor steps whose order matters. Then fade the map one line at a time.

Eventually the learner should recognise the sequence as a coherent procedure rather than recite a list. Explain-back helps reveal whether the steps have meaning.

Repair Playbook 5: when selection is missing

Separate classification from execution. Build contrast pairs. Ask why one method applies and another does not. Remove topic headings. Use mixed sets only after each candidate method is individually stable.

Selection repair often produces speed gains without any direct timing because hesitation at the junction disappears.

Repair Playbook 6: when recall is missing

Close the notes. Retrieve a small amount. Give feedback. Return later. Keep retrieval low-stakes and brief enough that it can happen repeatedly across time.

If a small cue restores everything, train the first cue and fade it. If the structure remains absent, the problem may be organisation or concept rather than memory alone.

Repair Playbook 7: when translation is missing

Insert a bridge representation. Text → bar model. Text → equation skeleton. Passage → claim/evidence table. Graph → observation statement. Writing prompt → audience/purpose/content map.

Practise the translation without always completing the final task. Once mapping is reliable, reintegrate calculation or writing.

Repair Playbook 8: when accuracy is missing

Classify the error before adding practice. Is it knowledge, sign, arithmetic, copied data, unit, grammar, spelling, evidence, command word or checking? Build a small set around the repeated family.

Add an anti-error cue and test whether it changes the next fresh attempt. A repair is stronger when the learner can prevent the error, not merely recognise it afterward.

Repair Playbook 9: when speed is missing

Do not time the whole task until you know where the time goes. Measure retrieval, decision, calculation, reading, planning or checking separately. Practise the bottleneck at a safe pace, then compress gradually.

Keep an accuracy floor. If faster practice creates a large accuracy loss, return to stable execution before compressing again.

Repair Playbook 10: when confidence is lower than competence

Use evidence, not reassurance alone. Ask the student to predict success, complete a short task and compare prediction with outcome. Build a record of independent successes under increasingly realistic conditions.

The academic aim is calibration: the student learns what they can rely on and what still needs work. Significant distress requires appropriate support beyond academic coaching.

Repair Playbook 11: when support dependence is high

Track prompt level. Move from full model to partial model, focused question, open question, self-prompt and independence. Do not remove all support at once if that simply recreates failure.

The success measure is not only marks. It is fewer and less specific prompts for the same quality of performance.

Repair Playbook 12: when the child can do it once but not later

This is a maintenance problem until evidence shows otherwise. Schedule short returns at widening intervals. Use mixed retrieval so the student must also recognise the skill.

Once retrieval stabilises, stop daily drilling. Durable learning does not require permanent maximum frequency.

A one-week takeoff repair plan

Day 1: diagnose and model

Name the micro-skill. Confirm the failure with one or two direct tasks. Build the minimum working model and study one clear example. End with a completion problem if needed.

Day 2: first independent cycle

Retrieve the model briefly, then complete a fresh direct problem independently. Record prompts if any. Explain the route and create an anti-error cue.

Day 3: small variation

Change one feature. Compare with the previous problem. Ask what remained the same. If variation fails, teach the structural connection and try another controlled example.

Day 4: rest or light retrieval

Use one short closed-book return rather than a large set. If retrieval is good, stop. More practice is not automatically better.

Day 5: mixed application

Place the repaired skill beside plausible alternatives. The student must recognise when it applies. This begins the transition from isolated repair to normal subject use.

Day 6: realistic output

Use the skill inside a normal homework, writing, Science or Mathematics task. Do not add heavy time pressure unless speed is already the known repair target.

Day 7: proof and release

Complete a fresh direct item, a variation and an explain-back. If stable, move the skill into Article 6-style reinforcement and maintenance. If not, identify the exact stage that still needs support.

A 30-day repair-and-release cycle

Week 1: first-principles repair and one clean independent success. Week 2: repeated core practice, anti-error work and modest variation. Week 3: mixed application, wider context shift and delayed retrieval. Week 4: realistic performance proof and maintenance decision.

The cycle is intentionally finite. The objective is not to keep a child permanently “in remediation”. A repaired skill should graduate into ordinary use. If it keeps requiring intensive support, revisit the diagnosis or consider whether a broader issue needs attention.

The scaffold-fading protocol

Adult demonstrates and explains.
Student completes missing steps.
Student solves with a visible step map.
Student solves with one open prompt available.
Student self-prompts from a small card or cue.
Student solves independently.
Student solves a variation independently.
Student retrieves the skill after delay.

Not every learner needs every step. Some move from model to independence rapidly. Others need more completion problems. The important feature is that support has an exit route.

The “too much help” test

Ask whether the student can reproduce the same performance when the adult stops talking. If not, identify which part of the adult’s help was carrying the task. Was the adult selecting the method, translating the wording, recalling the formula, checking every line or managing the sequence?

Convert that external support into a temporary self-prompt, then fade it. Good tuition should reduce the amount of tutor needed for the same task over time.

The “not enough help” test

Independence is not created by withholding all support while a novice repeatedly fails. If the student cannot identify a plausible first step and errors do not become more informative, add enough structure for productive learning: a representation, a worked example, a partial step or a focused question.

Then fade. The aim is successful transfer, not proving that the student can struggle unaided.

How parents can help without reteaching the whole lesson

Parents can support Article 5 repairs by asking for the target, maintaining a short practice schedule, protecting a quiet environment, checking that the child attempts before seeing the answer, and reminding them of their own anti-error cue.

Parents do not need to reproduce the tutor’s full explanation. If home teaching creates method conflict, send the problem back to the teacher or tutor with a concrete observation. “She understands the example but cannot choose the first step independently” is much more useful than adding another competing method.

How tutors should communicate a repair

A strong message is short: “The ratio concept is fine. The breakdown is identifying the invariant in before-and-after questions. We repaired that with a two-state model today. At home, one short variation on Thursday is enough. We will recheck without the model next lesson.”

This tells the family what is working, what was repaired, what support was used, what home practice is needed and when independence will be tested.

How students should describe their own repair

Older students can use one sentence: “I used to fail when ____. The rule/model I now use is ____. My error cue is ____. I proved it by ____. I will recheck it on ____.”

This forces the learner to turn a vague weakness into a mechanism, strategy and proof.

Repair versus relearning the entire chapter

Whole-chapter relearning is appropriate when knowledge is broadly absent. It is inefficient when most of the chapter is stable and one micro-skill repeatedly breaks performance.

Use marked papers and short probes to estimate the scope. If errors cluster tightly, patch locally. If the student cannot explain foundational ideas across the topic, broader reteaching may be necessary.

Repair versus more worksheets

Worksheets are tools, not repairs by themselves. The same worksheet can help one student and waste another’s time. A retrieval gap needs recall. A selection gap needs mixed classification. A concept gap needs meaning. A speed gap needs targeted fluency. A transfer gap needs controlled variation.

Before printing or assigning another set, finish the sentence: “This set is designed to improve ____.” If the blank cannot be filled precisely, the practice may be poorly targeted.

Repair versus full-paper practice

A full paper is a stress-test. It can reveal that a repair has transferred, but it is often an inefficient place to perform the initial repair. If one micro-skill fails repeatedly, take it out of the full paper, patch it, prove it, then return it to the paper.

This alternation—broad test, narrow repair, broad proof—is one of the most efficient patterns in examination preparation.

Repair versus memorising a model answer

Model answers can show structure and quality. They become a poor repair when the student copies surface wording without learning the generative process.

Reverse-engineer the model: what decision did the writer make, what evidence was selected, how was the causal chain built, what made the tone appropriate? Then hide the model and apply the principle to a fresh task.

Repair versus “carelessness”

A repeated execution error can often be repaired. A sign error can gain a cue. A unit omission can gain a final-unit check. A task-content omission can gain a pre-writing scan. Calling these “carelessness” without a prevention mechanism does not change behaviour.

If errors remain broad and unpredictable, Article 2’s volatility analysis and Article 4’s load analysis may be more useful than another narrow patch.

Repair versus significant broader difficulty

Article 5 is an academic repair framework. If difficulties are substantial, persistent across settings, unusually broad or accompanied by concerns about reading, writing, attention, hearing, vision, sleep, mood, distress or daily functioning, coordinate with school staff and seek appropriate qualified professional support where indicated.

The useful contribution of a repair framework is specific observation: which task failed, what support changed the outcome, how often the pattern occurs and whether it transfers across conditions.

The repair evidence ladder

Recognition: student understands the model when shown.
Completion: student fills missing steps in a supported example.
Independent direct: student completes a fresh direct item alone.
Explain-back: student can describe the important decision or structure.
Small variation: student survives a controlled change.
Delayed return: skill remains retrievable later.
Mixed selection: student recognises when to use the skill among alternatives.
Realistic performance: skill works at the pace and independence demanded by the real task.

Article 5 is mainly concerned with the first five rungs. Later articles in the sequence strengthen and maintain the upper rungs.

The repair stop rule

Stop intensive repair when the student can start independently, execute accurately enough, explain the core structure, handle a small variation and retrieve the skill after an appropriate delay. Move the skill into normal practice and maintenance.

Do not keep a student in “repair mode” because the adult enjoys seeing perfect easy repetitions. Repair exists to return the learner to real work.

Advanced Repair Handbook: Designing Practice That Actually Changes the Weak Skill

Once a weak skill has been diagnosed, the quality of the repair depends on practice design. A pile of questions is not a design. A useful repair set changes the exact mechanism that failed, gives feedback quickly enough for the learner to adjust, and then tests whether the improved behaviour survives after support is removed.

This handbook section turns Article 5 into a practical operating manual. It answers the questions families and tutors face after the first explanation: How easy should the first practice be? How much help is too much? When should examples be mixed? When should the student be timed? How many repetitions are enough? What should happen after a correction? When does a repaired skill graduate into normal revision?

Practice Design Rule 1: isolate before you integrate

If the failure is narrow, the first repair should usually be narrow. A student who repeatedly misidentifies the gradient from a graph does not need a full coordinate-geometry paper to practise that skill. A student who omits the audience in situational writing does not need to write a full letter every time.

Isolate the decision or operation that needs change. Build enough successful repetitions for the student to understand and control it. Then integrate it back into the larger task. Isolation makes the signal clear; integration proves usefulness.

Practice Design Rule 2: make the first success representative, not trivial

The first independent problem should remove irrelevant difficulty without removing the target difficulty. If the target is negative-sign control, keep the signs but avoid adding a complicated word problem. If the target is inference, keep the evidence-to-conclusion relationship but use a short passage. If the target is causal explanation, keep the mechanism but reduce the number of interacting concepts.

A problem that is too easy can produce false confidence because it no longer tests the repaired skill. A problem that adds several unrelated demands can hide whether the repair worked. Representative simplicity is the goal.

Practice Design Rule 3: one meaningful change at a time

Variation is essential, but uncontrolled variation is noisy. After a successful direct example, change one feature: numbers, wording, representation, direction, context or a common trap. Ask what remained structurally the same.

Once the student can tolerate several single-feature changes, combine them. This ladder lets the learner build transfer rather than simply experience surprise.

Practice Design Rule 4: mix only after individual methods can stand

Mixed practice forces the learner to choose. That is valuable when method selection is the goal. It is less useful when none of the candidate methods is stable enough to execute. A novice can become overwhelmed by having to choose among several half-learned routes.

Build each method to a workable direct level, then introduce contrast. For example, stabilise factorisation and the quadratic formula separately before asking the student to choose between them. Stabilise literal retrieval and inference separately before mixing comprehension question types.

Practice Design Rule 5: correction must end in generation

A student can understand a correction perfectly while it is being explained and still reproduce the same error tomorrow. Every important correction should therefore end with the learner generating a fresh response.

Close the model answer. Remove the tutor’s pen. Use a parallel item. Ask the student to perform the repaired step, explain the cue and check the result. The change becomes real only when the learner generates the behaviour.

Practice Design Rule 6: feedback should point to the next action

“Wrong” is information but not instruction. “Careless” is a judgement but often not a repair. Useful feedback says what happened and what action will change the next attempt.

“Your equation is correct; the sign changed when you expanded the negative bracket. Mark the multiplier before touching the terms.” “Your inference is plausible, but the evidence supports a narrower conclusion. Quote the evidence first.” “Your Science explanation has the right keywords, but the causal verb is missing. State what the temperature change causes the particles to do.”

The EEF’s feedback evidence work emphasises actionable information about improvement. At home, the same principle keeps feedback attached to the task rather than the child’s identity.

Practice Design Rule 7: retrieval should be short enough to repeat

A retrieval check does not need to be a test paper. One formula, one paragraph plan, one causal chain, one vocabulary definition or one direct calculation can reveal whether the repair remains available.

Short returns reduce the temptation to postpone review because “there is no time”. They also let the schedule adapt: skills retrieved easily can wait longer; fragile skills can return sooner.

Practice Design Rule 8: time pressure is a later layer, not proof of seriousness

A stopwatch can make practice look rigorous while hiding weak instruction. Add time when the student has a route worth compressing. If the method is still conceptually confused, timing measures confusion quickly.

Begin with micro-timing: how long to identify the method? How long to complete a familiar calculation? How long to produce a three-point plan? Then add section timing. Full-paper timing belongs after the repaired components can operate together.

Practice Design Rule 9: independence is a variable you can observe

Instead of saying “the child understands”, note how much support was needed. Full demonstration? Partial example? Focused question? Open prompt? Self-cue? No help?

A skill can improve even when the score stays temporarily similar if the amount of support falls. This is especially important during scaffold fading. Independence is part of the outcome, not an optional extra.

Practice Design Rule 10: stop before practice becomes noise

Once the target behaviour is stable enough for the current stage, more identical repetitions may add little. Switch to variation, delay or mixed application. If the student is tired and error quality deteriorates without learning, stop and return later.

Practice quality is not measured by exhaustion. A useful session should create or verify a change.

When to reteach from first principles

Reteach when the student lacks a coherent model, cannot explain the relationship, cannot begin a direct item, or gives contradictory reasoning even in simple conditions. Use clearer representations and simpler examples rather than repeating the same explanation louder.

Ask whether the problem is actually prerequisite knowledge. If the current concept depends on an older gap, repairing the prerequisite may be more efficient than another full explanation of the current chapter.

When to practise rather than reteach

Practise when the model is basically sound but execution is hesitant, recall is fragile, selection is slow, transfer is weak or the sequence needs fluency. The student does not need to hear the same concept lecture every time a procedure is slow.

A short explanation may refresh the model, but most of the session should target the actual performance gap.

When to compare examples

Comparison is especially useful for selection and transfer. Put two examples side by side and ask why the same method works in both. Or put two similar-looking examples side by side and ask why they require different methods.

The contrast directs attention toward structural features that ordinary repetition can leave invisible.

When to use a worked example

Use one when the procedure or representation is new, when errors show that the student has no stable route, or when cognitive load is so high that independent attempts produce only chaos. Keep the example clean and narrate decisions.

Then transition quickly to completion and independent problems. The example is a bridge, not a place to live.

When to use discovery

Exploration can be powerful when the learner has enough prior knowledge to notice useful patterns and when the task is bounded. It is less productive when the student lacks the prerequisites and is expected to rediscover an entire method through repeated failure.

A good repair can include guided discovery: compare cases, predict, test, explain. The adult still controls the problem space enough that the learner can see the structure.

When to use memorisation

Some knowledge deserves efficient memory: multiplication facts, core formulas, vocabulary, definitions, notation and common structures. Memorisation becomes weak when it substitutes for meaning or when students cannot recognise when the memorised item applies.

Pair memory with structure and application. Know the formula, understand the variables, recognise the situation and use the formula. Know the vocabulary, understand meaning and register, then produce it in context.

When to use a checklist

Checklists are useful for repeated multi-step tasks and predictable error families: situational-writing requirements, experimental-data checks, algebra sign control, examination submission routines. Keep them short.

Fade them once the sequence is internalised. A checklist that is never removed can become another support the student misses in an examination.

When to use alternative methods

Alternative methods can deepen understanding, provide checks and create flexibility. Introduce them after one route is stable enough to serve as an anchor, unless the comparison itself is essential to building the concept.

Teach the learner when the alternative is useful, not merely that it exists. A backup method should reduce risk or work in certain situations, not increase the number of choices without a selection rule.

The error-led worksheet

A powerful repair worksheet can be built from the student’s own errors rather than a publisher’s chapter sequence. Choose three to five recurring errors. Create one clean example, one direct repair item and one variation for each.

For Mathematics: sign handling, unit conversion, method selection. For English: evidence selection, sentence boundaries, task fulfilment. For Science: causal verbs, formula selection, graph reading.

The worksheet is small by design. Its value comes from relevance. After it works, retire it rather than repeating the same sheet indefinitely.

The contrast worksheet

Build pairs of questions that expose the selection rule. Factorisation versus quadratic formula. Literal retrieval versus inference. Observation versus explanation. Percentage increase versus reverse percentage. Area versus perimeter.

Ask the student to classify and explain the difference before solving. This type of practice is especially valuable when the student “knows everything separately” but chooses poorly in mixed work.

The representation worksheet

Use the same relationship in several forms. Ratio as bar model, table, equation and story. Linear relationship as graph, table, formula and verbal rate. Science mechanism as diagram, arrows and paragraph. English argument as claim map and prose paragraph.

Ask what stays the same. Representation flexibility reduces context-shift failure because the learner is no longer tied to one surface form.

The retrieval worksheet

Keep it short and closed-book. Mix knowledge from different days. Ask for formulas, definitions, first steps, tiny examples and explanations. Give immediate feedback, then schedule another return.

The purpose is not to produce a high score for its own sake. It is to make important knowledge available without the original page in front of the learner.

The transfer worksheet

Begin with one familiar example. Then alter one feature in each subsequent item: wording, direction, representation, context, irrelevant information, mixed method. Ask the student to annotate what changed and what stayed invariant.

This makes transfer visible rather than treating every unfamiliar question as an arbitrary trick.

The independence worksheet

Design the same short task at decreasing support levels: worked example visible, step map visible, cue card visible, no support. Record performance and prompt need.

The worksheet measures whether support is being internalised. It is particularly useful for students who appear strong during tuition but cannot reproduce the method alone.

The correction notebook that does not become a museum

Traditional correction books can accumulate dozens of old mistakes that students never revisit. A repair notebook should stay active and small. Record only recurring or high-cost errors.

Each entry contains: trigger, mistake, correct principle, prevention cue, fresh proof, next return date. Retire the entry when the error stops appearing under realistic conditions.

The goal is not a thick correction book. It is a shrinking list of active failure modes.

The repair card

Skill: ____________________
Breaks when: ____________________
My model/rule: ____________________
My first-step cue: ____________________
My common error: ____________________
My anti-error cue: ____________________
Independent proof: ____________________
Next return: ____________________

Older students can maintain their own cards. Younger students may need an adult to simplify the language. The card should make the repair easier to retrieve, not create administrative work.

Primary-level repair principles

Keep models concrete, language simple and practice brief. Use objects, drawings and oral explanation where useful. Protect the distinction between learning and performance: young children do not need every new skill timed.

Move from concrete to visual to symbolic when the subject supports it. Use frequent small successes, but do not hide all challenge. Variation should be gentle enough that the child can recognise the connection.

Parent support should focus on routine, encouragement, reading the task together when appropriate and communicating recurring patterns to school or tutor. Home does not need to become a second classroom every evening.

Lower Secondary repair principles

Abstraction increases and students must begin to manage more of the learning process themselves. Repairs should therefore include method-selection language and self-prompts, not only correct procedures.

This is a good stage to teach error classification, mixed practice, delayed retrieval and simple self-monitoring. The student should increasingly be able to say what broke and what practice is needed.

Upper Secondary repair principles

Time becomes more valuable and prerequisites matter more. Small algebra, reading or writing-control weaknesses can contaminate many topics. Repair should prioritise high-leverage bottlenecks.

Move from direct repair to mixed and timed proof relatively quickly once the skill is stable. Examination conditions are approaching, so independence and throughput matter. But do not use full papers as the only repair method.

JC repair principles

JC students often need micro-repairs that feel surprisingly basic: algebra, formula retrieval, paragraph architecture, reading pace, data interpretation. There is no shame in repairing the operation that limits higher-level performance.

Because content moves quickly, repairs should be efficient and immediately reconnected to the current syllabus. The test is downstream effect: does the ten-minute patch make the advanced task easier, faster or more accurate?

Mathematics repair principles

Separate representation, selection and execution. Many “Math problems” are actually one of those three. A student may represent correctly but select poorly, or select correctly but execute algebra badly.

Use diagrams and equations to connect meaning. Use classification sets for selection. Use short fluency sets for execution. Use mixed and varied questions for transfer. Use targeted checks for recurring error families.

English repair principles

Repair the writing or reading operation rather than telling the student to “improve English”. For comprehension: question type, evidence, inference, paraphrase, language effect, pacing. For writing: task parsing, planning, paragraph development, sentence control, vocabulary precision, editing.

Full essays and full papers are integration tasks. Use smaller units when repairing one component, then return to the full task to prove transfer.

Science repair principles

Separate recall, representation, calculation, data reading and causal explanation. A correct keyword is not a complete explanation. A correct formula is not method selection. A correct graph reading is not scientific interpretation.

Use arrows, variable maps, unit lists and observation/explanation separation as temporary scaffolds. Fade them once the student can generate the structure independently.

Vocabulary repair principles

Move beyond copied definitions. A repaired vocabulary item should be understood, recognised in context, produced accurately and retrieved after delay. Teach collocation and register where relevant.

Short repeated encounters across reading, speaking and writing are often more useful than one long copying session.

Exam-strategy repair principles

Not every repair is subject content. Students may need to repair first-step orientation, move-on decisions, time checkpoints, checking routines or recovery after a difficult question.

These skills should be practised in realistic sections after the academic knowledge is stable enough. Strategy cannot compensate indefinitely for missing content, but good content can be wasted by poor control.

How to choose the highest-leverage repair

When several weaknesses exist, ask which one appears inside the others. A sign-control problem may affect algebra, functions and calculus. Slow reading may affect several English components. Weak causal explanation may appear across multiple Science topics.

Also consider cost: how many marks, how much time and how many topics does the weakness affect? High-leverage repairs deserve priority even when they are not the most dramatic-looking errors.

How to know whether a repair is too broad

If the plan contains several chapters, several strategies and several error types, it is probably a revision programme rather than a micro-skill repair. That may be necessary, but it is a different job.

Article 5 works best when the target can be described in one sentence and tested in a few minutes.

How to know whether a repair is too narrow

If the student can perform the repaired item only when it is nearly identical to the example, the target may have been reduced to a template rather than a transferable skill. Add a small variation and ask what structure connects the two.

A useful micro-skill is narrow enough to teach clearly but broad enough to recur in meaningful work.

The repair hypothesis

Before starting, write a simple hypothesis: “If we teach the student to identify the invariant before modelling before-and-after ratio, accuracy on those problems should improve.” Or: “If method selection is the bottleneck, classify-only quadratic practice should reduce hesitation in mixed sets.”

Then test the intervention. If performance changes as predicted, the diagnosis gains support. If not, revisit the evidence. Good repair is iterative rather than dogmatic.

The repair log

A concise log can contain: date, target skill, baseline, repair used, prompt level, independent proof, variation result, next return. That is enough to prevent the same lesson from being repeated without knowing whether anything changed.

For families, one line is often enough. For tutors, the same structure can guide lesson planning and communication.

The release decision

A repaired skill should not occupy intensive lesson time forever. Release it when independent direct work is reliable, a small variation is manageable and delayed retrieval is acceptable. Then test it through normal mixed work.

If the skill later drifts, Article 10 and Article 11 in this series address drift and maintenance. Repair and maintenance are different jobs.

The difference between “fixed today” and “durable”

Article 5 produces a credible first repair. Durability requires later repetition, variation and spaced retrieval. Parents should therefore avoid two extremes: declaring permanent mastery after one good lesson, or refusing to acknowledge progress until the skill is flawless in every possible condition.

Learning develops in layers. First the skill becomes possible. Then reliable. Then flexible. Then durable. Then efficient under pressure.

The parent operating manual

Ask what exact skill is being repaired.
Ask what is already working so the child does not relearn everything.
Keep home practice short and aligned with the repair.
Let the child attempt before helping.
Use the child’s own cue rather than inventing a second method.
Do not convert every correction into a long lecture.
Return after a delay to see whether the repair lasted.
Communicate repeated patterns to teacher or tutor.
Stop intensive repair once the skill is ready for ordinary practice.

The tutor operating manual

Diagnose the smallest meaningful failure.
Choose the minimum working model.
Model decisions and common errors.
Use completion problems if the independence jump is too large.
Record the prompt level needed.
Require a fresh independent attempt.
Add one controlled variation.
Schedule delayed retrieval.
Communicate the repair in ordinary language.
Release the skill into reinforcement when takeoff is achieved.

The student operating manual

Know what exact part you are fixing.
Understand the reason or structure, not only the answer.
Try before looking back at the example.
Notice the first place you become unsure.
Use your anti-error cue.
Explain the route in your own words.
Try one changed version.
Return later without notes.
Tell the adult what still feels fragile.

Final repair checklist

Is the target narrow and observable?
Do we know whether the gap is concept, prerequisite, selection, retrieval, translation, sequence, accuracy or speed?
Is the minimum model clear enough to guide action?
Has the student seen one clear example?
Has support been reduced rather than maintained indefinitely?
Has the student completed a fresh independent item?
Can the student explain the core decision or structure?
Has one small variation been tested?
Is there a delayed return?
Do we know when to stop repair and move to reinforcement?

The deepest rule in Article 5

A weak skill is not repaired because an adult has explained it beautifully. It is repaired when the learner can use a workable model to make the right decision, execute the skill independently, explain enough of the structure to show control, and repeat the performance when the surface changes.

That is why the first clean independent success matters. It is the point where the repair stops belonging to the explanation and starts belonging to the student.

From there, repetition, variation, retrieval and examination training can strengthen the skill. But without that first owned cycle, more practice can simply make the wrong thing more familiar.

Frequently asked questions

Why doesn’t explaining the topic again always work?

Because the missing part may not be explanation. The student may understand the concept but lack retrieval, method selection, translation, sequencing, fluency or independent execution. Re-teaching the stable part leaves the bottleneck untouched.

How many correct questions prove that a skill is repaired?

There is no universal number. One clean independent success plus a small variation is a useful first proof, not permanent mastery. Later evidence should include delayed retrieval, mixed context and realistic performance conditions appropriate to the skill.

Should the first independent problem be difficult?

Not unnecessarily. Its job is to test the repaired mechanism. Complexity can increase after the route is stable.

Should parents correct immediately?

It depends on the stage. During initial repair, allowing a completely wrong route to run for a long time may rehearse confusion. But adults should avoid taking over. Use the least prompt needed, record it, and re-test later with less support.

What if the student can explain but still cannot do?

Then representation, procedure, fluency or output may be the missing link. Explanation is evidence of one layer, not proof of the entire skill.

What if the student can do but cannot explain?

The procedure may be memorised or the student may simply find verbal explanation difficult. Ask for a diagram, step labels or comparison rather than assuming the concept is absent. The goal is to reveal enough of the internal model to judge stability.

When should timing be added?

After the route is sufficiently correct and independent. Timing should test or compress a working method, not substitute for one.

When should we stop repairing and move on?

When direct performance is independent, the student can handle a small variation, delayed retrieval is acceptable and the skill is sufficiently reliable for its current role. Then reduce dedicated repair and build wider stability through normal practice.

The parent summary

When a child is weak in one area, resist the urge to attack the whole subject. Name the smallest skill that is breaking. Decide whether the missing piece is concept, prerequisite, method selection, retrieval, translation, sequence or fluency. Build the smallest useful model. Study one clear example. Let the student complete part of the next one. Then remove the support and ask for a fresh independent attempt.

If that attempt succeeds, ask the student to explain the route and try one small variation. Return after a delay. Only then add heavier mixing, timing and full-paper pressure. That sequence makes practice more efficient because each stage has a defined job.

The most useful change in language is this: instead of “My child still doesn’t get the topic,” ask “Which exact part of the skill still requires support?” That question turns frustration into a repair route.

Return to Bukit Timah Tuition OS