Start here: Constraint-Change Dependency Trace is a practical way to turn a vague “check your work” instruction into a visible sequence of decisions. In the next ten minutes, take one attempted question, mark the exact feature named in this guide, and explain one decision aloud before doing more practice. The goal is controlled G2 SEC performance, not a promise about marks or a prediction of a particular paper.
For 2027 school candidates, the official SEAB G2 syllabus list identifies English Language K200, Mathematics K210 and the common Science syllabus used for K223, K224 and K225. MOE’s announced SEC transition details provide the wider system context. These sources describe announced assessment and syllabus arrangements; your school’s current instructions remain the authority for local administration, schedules and permitted materials.
This chapter adds a narrow, practice-rich job to Volume 0268, the canonical guide to matching checks with failure modes. It also connects to Volume 0272 on asking for the smallest useful help, the relevant eduKateSG learning hub, Examination Craft and the PSLE Learner’s Guide series. Use the route below to diagnose the mechanism, practise it in controlled steps, and then transfer it to an unseen task.
Choose a task route
Route 1 — Diagnose: Chapters 1–3 make the target visible. Route 2 — Build: Chapters 4–7 teach the mechanism through worked contrasts. Route 3 — Check: Chapters 8–11 connect representation, calculation or wording to evidence. Route 4 — Repair: Chapters 12–13 classify misconceptions and make the smallest useful correction. Route 5 — Transfer: Chapters 14–15 move from guided practice to independent use.
Chapter contents
1. Name the changed constraint
3. Separate fixed facts from derived facts
4. Find the first affected decision
5. Trace downstream consequences
6. Protect independent branches
7. Use a reason tag on every revision
8. Handle changed units safely
10. Handle changed assumptions
11. Compare minimal and full rebuilds
13. Diagnose recurring dependency errors
15. Build an independent weekly routine
1. Name the changed constraint
Chapter 1 of 15 · Diagnose route
Rewrite the changed condition in one short sentence and underline the quantity, audience, evidence rule or scientific condition that is different. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A practice instruction changes from “give two reasons” to “give one reason and one limitation”. The number of paragraphs may shrink, but the need for supported reasoning remains. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Starting the whole response again hides which decisions were actually affected. Preserve the unchanged command words and trace from the altered phrase only. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Take three familiar questions. Change one condition in each, then label the exact phrase that activates a different decision. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
2. Build a dependency arrow
Chapter 2 of 15 · Diagnose route
Draw an arrow only when a later claim, value or sentence genuinely uses the changed information. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
In Mathematics, a radius changes from 4 cm to 6 cm. Every value calculated from r changes; a separately given height does not. Mark r → base area → volume, while height stays fixed. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Chronological order is not the same as dependency. A line written later may be independent, while an early derived value can feed many later lines. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
For each answer line, ask: “Could this line remain true if only the changed condition were replaced?” Explain yes or no. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
3. Separate fixed facts from derived facts
Chapter 3 of 15 · Diagnose route
Make two columns: facts copied from the task and results produced from those facts. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A Science setup keeps the same material and temperature but doubles wire length. Material and temperature remain fixed; resistance prediction changes because length is an input to that relationship. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Students often erase fixed facts because they appear near the changed one. Protect them explicitly so the repair stays small and auditable. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Create a fixed/derived table for one English, one Mathematics and one Science example. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
4. Find the first affected decision
Chapter 4 of 15 · Build route
Move through the solution until the first line whose justification uses the changed condition. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
An English audience changes from classmates to the principal. The event facts stay, but the first affected choice may be greeting, tone or degree of explanation. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
The first affected decision is not always the first sentence. It is the earliest point at which the new constraint alters a justified choice. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Compare two versions of an answer and circle the first point where the reason for the choice changes. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
5. Trace downstream consequences
Chapter 5 of 15 · Build route
From the first affected decision, follow every later step that imports its output. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
If a Mathematics scale factor changes, lengths, areas and volumes do not all change by the same multiplier. The trace branches into k, k² and k³. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
A dependency trace prevents the “change one number everywhere” habit. Each branch needs its own relationship. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Draw a three-branch trace for a changed scale factor and annotate each arrow with the operation it triggers. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
6. Protect independent branches
Chapter 6 of 15 · Build route
Mark branches that use different evidence or different given quantities. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
In a Science explanation, changing light intensity may affect rate evidence but need not alter the identification of the apparatus. Keep the apparatus statement if its basis is unchanged. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Over-editing creates fresh errors. Preservation is a skill: valid work should survive unless a real dependency connects it to the changed condition. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Highlight two lines that must change and two that should remain untouched in a four-line response. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
7. Use a reason tag on every revision
Chapter 7 of 15 · Build route
Attach a brief reason such as audience, unit, relationship, evidence or boundary condition to each changed line. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A revised English closing is not better merely because it sounds formal. Tag it “audience and purpose” and explain how the new reader changes the requested action. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Reason tags expose cosmetic edits. If you cannot name the dependency, the change may be unnecessary. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Revise one answer with tracked changes and add a five-word reason after every edit. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
8. Handle changed units safely
Chapter 8 of 15 · Check route
Treat a unit change as a representation constraint, then inspect conversions and final labels. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A length changes from 2 m to 200 cm. The physical length is unchanged, but direct substitution may fail if the formula uses metres. The numerical path changes while the situation does not. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Unit changes can alter numbers without altering meaning. Separate quantity, numerical value and unit before deciding what to recalculate. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Convert the same measurement three ways and identify which calculations remain invariant. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
9. Handle changed evidence
Chapter 9 of 15 · Check route
When a source, graph or observation changes, trace only claims supported by that evidence. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A graph point is corrected. The trend statement may remain, but a calculated gradient using that point may change. A conclusion depending on the gradient must then be checked. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Do not let one corrected datum destroy an unrelated pattern. State which claim used which evidence. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Make an evidence-to-claim map for a short data set with one amended reading. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
10. Handle changed assumptions
Chapter 10 of 15 · Check route
Write assumptions beside the step that uses them instead of leaving them invisible. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A word problem originally assumes constant speed; the revised version includes a stop. Distance = speed × total elapsed time is no longer justified without separating moving time. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Hidden assumptions create large downstream errors because they look like facts. Making them visible turns a vague restart into a precise repair. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
List two assumptions in a familiar method and describe what would break if each were removed. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
11. Compare minimal and full rebuilds
Chapter 11 of 15 · Check route
Do a minimal repair first, then compare it with a clean solution only after the trace is complete. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
If both versions agree on preserved branches and differ only where dependency arrows predict, the trace is probably sound. If they diverge elsewhere, inspect the first unexplained difference. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
A full rebuild is useful as verification, not as the first reaction. Otherwise you lose evidence about which decision caused the mismatch. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Repair one changed task minimally, then solve it cleanly and reconcile line by line. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
12. Use stop rules under time
Chapter 12 of 15 · Repair route
Stop tracing when every final claim is supported by either an unchanged fact or a recalculated branch. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
A practical stop rule is: all changed inputs have outgoing arrows, every affected output has been rebuilt, and no revised line lacks a reason tag. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
More checking is not automatically safer. Unbounded checking consumes time and can introduce speculative changes. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Practise a ninety-second trace, then state which checks earned their time. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
13. Diagnose recurring dependency errors
Chapter 13 of 15 · Repair route
Classify failures as missed branch, false branch, wrong relationship, hidden assumption or unit mismatch. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
If a learner changes every number after one input changes, classify it as a false branch. If a later conclusion still uses an old derived value, classify it as a missed branch. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
A stable category supports a targeted drill. “Careless” does not tell you what to practise next. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Collect five repaired answers and tally the dependency-error categories. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
14. Transfer across subjects
Chapter 14 of 15 · Transfer route
Keep the same trace language while changing the object being traced. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
English traces purpose to wording, Mathematics traces quantities to derived values, and Science traces conditions to mechanisms and evidence. The surface features differ; dependency logic remains. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Transfer succeeds when the learner can name both the invariant process and the subject-specific relationship. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
Explain the same trace routine using one example from each G2 subject. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
15. Build an independent weekly routine
Chapter 15 of 15 · Transfer route
Use small changed-condition pairs, retrieval, delayed rechecking and one mixed set. This is the chapter’s decision rule. Say it before looking at a worked answer. A usable rule names what to notice, what remains fixed, and what action follows. If the rule cannot guide the next line, shorten it until a learner can apply it without guessing.
Mechanism
The constraint-change dependency trace separates observation from action. First identify the relevant feature; next connect it to the governing relationship; then produce an answer that can be checked against that relationship. A strong learner can state the boundary of the rule, contrast it with a nearby case and explain why the move fits. This leaves a trace: when an answer fails, inspect the feature, connection or execution instead of calling the whole attempt “careless”.
Worked example
On day one, compare paired tasks. On day three, reconstruct the trace without notes. On day seven, solve a mixed task and justify which lines were preserved. Work in four visible moves: mark the cue; write the relationship in words or symbols; carry out the smallest justified operation; and read the result back into the task. Then change one feature while holding the rest constant and predict before recomputing or rewriting. The contrast reveals which element controls the decision and catches an old routine being followed automatically.
Misconception diagnosis and repair
Parents and tutors should ask “What depends on that?” before showing a method. This keeps the learner responsible for the decision while still providing useful structure. Preserve the learner’s first attempt and underline the earliest decision that cannot be justified. Add one corrective prompt rather than the finished answer. After the repair, close the model and reconstruct from a blank space. Immediate copying tests sight, not retrieval; a delayed second attempt shows whether the decision can be reproduced independently. Record the error category and corrective cue, not just a tick or cross.
Practice progression
End the week with one unseen condition change and a short reflection on the first affected decision. Begin with two clean contrasts, continue with four mixed items, and finish with one item containing irrelevant information. Use three levels: identify the feature without solving; complete the method with labels; then transfer to a differently worded or drawn task. A complete explanation should show the controlling feature, relevant relationship, resulting action and brief verification. Vary context, representation and command separately so a transfer failure can be diagnosed precisely.
Official format and scope note
Check subject-specific claims against the 2027 K200 English syllabus, 2027 K210 Mathematics syllabus and 2027 K223/K224/K225 common Science syllabus. A cross-subject routine does not change the official format of any paper. It helps learners manage decisions inside the format described by the official documents and their schools.
Parent and tutor guidance
Use the learner’s first decision as evidence. Ask them to point to the cue, state the relationship and justify the next move. If they are stuck, offer the smallest prompt that restores the decision, then remove that prompt on the next item. Avoid turning a diagnostic into a performance of adult speed. The aim is independent reconstruction.
A useful weekly record has four columns: task type, earliest unstable decision, corrective cue and delayed-retest result. This is more informative than recording only scores. Praise accurate noticing and successful self-correction, while keeping the standard clear. If school instructions differ from a general example here, follow the school’s current directions.
For a benchmark of depth, deliberate explanation and learner-first practice design, see eduKateSG’s Secondary 1 Mathematics tutorial reference. The present article uses its own original examples and a text-only Gutenberg structure so every paragraph and heading remains editable.