Did the Mathematics intervention repair the weak link—or did it simply help the student route around it? Both can improve performance. They are not the same outcome. A student may score better because the underlying prerequisite became stronger. Or the student may have learned a compensating strategy, received more prompting, used a calculator, avoided a difficult method, memorised a question family, or relied on a representation that makes the weak skill unnecessary.
Sometimes compensation is sensible. A calculator can legitimately reduce arithmetic load. A diagram can legitimately make a relationship visible. A more efficient method can legitimately replace an awkward one. The diagnostic problem begins when a workaround is reported as if the original dependency itself had been repaired.
This article owns that evidence question: mathematics repair versus compensation. The existing Mathematics Fracture and Repair Map explains why recurring mistakes can originate from earlier dependencies. The Foundation Problem or Method Problem? page helps identify where the break sits. This guide asks what comes after intervention: did the broken capability itself become stronger, or did the system merely learn to avoid loading it?
Evidence-informed intervention guidance from the Institute of Education Sciences repeatedly emphasises systematic instruction, mathematical language, representations, word-problem structure and progress monitoring. The older Response to Intervention practice guide explicitly includes progress monitoring as a recommendation. The 2021 guide for students struggling with Mathematics emphasises systematic instruction and well-chosen representations. These sources apply most directly to specific school-age populations and intervention settings, but they support a broader principle useful here: an intervention should be judged by what changes in the learner, not merely by the fact that supported performance rose.
1. Repair and compensation can both be useful
Repair means the weak capability itself becomes more reliable. If algebraic fraction manipulation was the weak link, repair means the learner can now manipulate algebraic fractions more accurately, independently and across fresh contexts. If reading a graph was the weak link, repair means graph interpretation itself improves.
Compensation means another route reduces the need to use the weak capability. A student with weak mental arithmetic may use a calculator accurately. A learner who struggles to hold a complex word problem in working memory may use a diagram or table. A student who cannot remember a formula may use a permitted formula sheet. These can be intelligent supports.
The question is not “Is compensation bad?” It is “Which outcome did we actually produce, and is that outcome sufficient for the next layer of Mathematics?”
2. Why the distinction matters in cumulative Mathematics
Mathematics is dependency-rich. A weak lower capability can be avoided for a while but become unavoidable later. A student may route around weak factorisation with a calculator or pattern memory in one topic, then struggle when factorisation is required inside algebraic fractions, graph analysis or Additional Mathematics. A workaround can therefore look successful until a later topic removes it.
Conversely, some compensations are permanently legitimate. Graphing calculators are part of certain courses. Diagrams are authentic mathematical tools. A student does not need to abandon a good representation merely to prove independence. The important distinction is whether the support replaces a capability that later Mathematics will still require.
3. The five post-intervention possibilities
- True repair. The weak capability itself improves and survives fresh, delayed and independent use.
- Supported repair. The capability is improving but still needs a cue, representation or scaffold.
- Compensation. Performance improves because another route reduces exposure to the weak capability.
- Avoidance. Tasks are selected or simplified so the weak capability is rarely tested.
- Masking. External help produces correct performance while the learner’s own capability changes little.
These states can coexist. A student may have partial repair plus useful compensation. The teaching decision depends on what future Mathematics will demand.
4. The dependency isolation test
After an intervention, isolate the original weak link in a short clean task. If the problem was fraction arithmetic, test fraction arithmetic directly. If the problem was algebraic manipulation, test it without the later topic wrapped around it. If the problem was method selection, use a small mixed set where the learner must choose among known methods.
If isolated performance improved, that is direct evidence of repair. If only the larger supported task improved while the isolated dependency remains weak, compensation or masking is more likely.
5. The removal test
Temporarily remove the suspected workaround where doing so is safe and educationally appropriate. Close the worked example. Delay the hint. Ask for the first step before calculator use. Remove the labelled diagram and ask the student to create one. Use a fresh question without the memorised template.
If performance collapses immediately, the workaround was carrying substantial load. That does not make it wrong; it tells us the original dependency is not yet independent.
6. The changed-route test
Give a new problem where the same underlying prerequisite is required but the familiar workaround does not apply in exactly the same way. A memorised quadratic template may fail when factorisation appears inside a rational expression. A calculator-based graph routine may fail when the question asks for exact reasoning. A rehearsed word-problem model may fail when information order changes.
If the learner can still use the underlying capability, repair is more plausible. If performance drops only when the workaround loses its usual form, the support remains the main carrier.
7. The delayed test
A workaround can create immediate smoothness without durable repair. Retest the isolated dependency after time has passed and without the original scaffold visible. If the capability survives, the evidence strengthens.
If it disappears, return to direct retrieval and conceptual work rather than simply adding more supported examples.
8. The explanation test
Ask the learner what changed. “Why can you do this now?” is not a perfect measure, but a useful answer can reveal whether the student has a new understanding, a new strategy or a new external support. “I draw the two quantities first so I can see the relationship” is a legitimate strategy. “I know the tutor will tell me which formula” reveals dependence.
Explanation should be combined with performance evidence, not used alone.
9. Repair should reduce fragility upstream and downstream
A repaired prerequisite should help more than the one question that triggered the intervention. Stronger fraction structure should help ratios, algebra and percentages. Stronger algebraic manipulation should help equations, graphs and calculus. Stronger graph interpretation should help functions, rates and statistics.
Look for this spread. If improvement appears only in the exact intervention task, the repair may be too local.
10. Thirty intervention outcomes: repair, compensation, avoidance or masking?
1. Weak fraction arithmetic replaced by calculator dependence
Observed improvement. The student now completes algebra questions accurately because every fraction operation is sent to a calculator. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Test fraction operations directly without the later algebra wrapped around them, then compare with calculator-supported work. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If isolated fraction accuracy remains weak, the intervention improved task performance through compensation rather than repairing the arithmetic dependency. Decide whether that dependency will still matter in future non-calculator or symbolic work. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
2. Weak factorisation hidden by quadratic formula use
Observed improvement. The learner solves quadratic equations reliably by formula but still cannot factor simple expressions needed elsewhere. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Test factorisation outside the quadratic-equation context and inside algebraic fractions. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. Quadratic performance improved, but the weak link remains. The formula is a legitimate alternative for equations; it does not replace factorisation as a separate capability. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
3. Weak algebra hidden by graphing calculator output
Observed improvement. The student finds roots and intersections numerically but cannot manipulate the equations or justify exact results. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use a task where exact algebra is required or where the calculator is not the most informative first step. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If performance collapses, technology is compensating for weak algebra. Repair may still be necessary because later topics require symbolic control. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
4. Weak number sense hidden by calculator checking
Observed improvement. The learner gets answers correct but cannot estimate magnitude or detect impossible outputs. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Ask for a rough estimate and sign prediction before calculator use. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If estimation remains poor, numerical judgement was not repaired. Calculator accuracy masks an important checking capability. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
5. Weak word-problem translation hidden by a memorised template
Observed improvement. The student succeeds on a familiar problem family after learning a fixed model layout. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Change nouns, sentence order and which quantity is unknown while preserving the structure. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the template no longer works and the learner cannot reconstruct the equations, the intervention built a narrow route around translation rather than repairing it. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
6. Weak graph interpretation hidden by teacher annotations
Observed improvement. The tutor labels intercepts, asymptotes and turning points before discussion, and the student then answers accurately. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Give an unannotated graph and ask the student to identify relevant features independently. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. A collapse suggests the annotation was doing the perceptual work. Repair requires the learner to detect the features themselves. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
7. Weak method selection hidden by chapter worksheets
Observed improvement. The student performs strongly when every page is titled with the method. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Mix nearby topics without headings. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If errors return, execution improved but selection did not. The intervention routed around the decision by supplying it through page organisation. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
8. Weak retrieval hidden by open notes
Observed improvement. The learner is accurate with notes beside the page but cannot begin without them. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use a short closed-note fresh task after a delay. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the method disappears, reference materials are compensating for retrieval. Whether that is acceptable depends on the assessment context and long-term goal. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
9. Weak notation hidden by tutor translation
Observed improvement. The tutor routinely rewrites dense notation into friendlier language before the student starts. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Present ordinary syllabus notation and ask the learner to explain it before solving. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If interpretation fails, the content may be understood while notation literacy remains unrepaired. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
10. Weak geometry reasoning hidden by measuring diagrams
Observed improvement. The student measures angles or lengths from the drawing instead of using properties. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use an intentionally not-to-scale diagram and forbid measurement. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If reasoning fails, measurement was a workaround. Repair requires theorem and property use independent of appearance. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
11. Weak trigonometric definitions hidden by orientation
Observed improvement. The learner succeeds only on standard right-triangle drawings. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Rotate the triangle and remove familiar labels. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If side identification fails, the intervention improved a prototype rather than the relational definitions of opposite, adjacent and hypotenuse. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
12. Weak proportional reasoning hidden by unitary recipes
Observed improvement. The student can execute a memorised ‘find one unit first’ sequence but cannot recognise proportional structure in a graph or table. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use another representation and ask what remains constant. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the student cannot connect the forms, the recipe compensates for weak structural understanding. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
13. Weak function concept hidden by substitution routine
Observed improvement. The learner can compute f(3) but treats f as if it were a variable or coefficient. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use mapping, graph and verbal-rule tasks. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the meaning remains unstable, substitution fluency was local compensation rather than full repair of function concept. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
14. Weak inverse-function meaning hidden by algebraic steps
Observed improvement. The student can swap x and y and rearrange but cannot explain what the inverse does. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use a mapping or composition check. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the reverse relationship is unclear, the algebraic algorithm is carrying performance without conceptual repair. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
15. Weak calculus meaning hidden by derivative rules
Observed improvement. The student differentiates accurately but cannot interpret sign, slope or rate. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use graph and context questions without asking for a derivative formula first. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If meaning fails, procedure improved while conceptual dependency remains. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
16. Weak integration meaning hidden by formula matching
Observed improvement. The learner integrates standard forms but cannot identify accumulation or signed area in context. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use a rate graph or area interpretation problem. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the learner cannot connect the integral to accumulated quantity, formula fluency is compensating for missing meaning. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
17. Weak probability modelling hidden by calculator commands
Observed improvement. The student obtains binomial or normal probabilities correctly after being told the model. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Remove the model label and ask which distribution applies and why. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If selection fails, calculation is supported while modelling remains weak. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
18. Weak hypothesis-test reasoning hidden by sentence templates
Observed improvement. The learner fills in standard conclusion phrases accurately but cannot explain what the p-value or null hypothesis means. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Change context and ask for a plain-language interpretation before the template. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If reasoning collapses, the writing frame is compensating for inferential understanding. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
19. Weak regression interpretation hidden by calculator output
Observed improvement. The student reports coefficients but cannot say what they mean or when prediction is unsafe. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Ask for interpretation, interpolation versus extrapolation, and variable roles without the GC visible. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If those fail, computation improved while model interpretation remains unrepaired. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
20. Weak retrieval hidden by spaced tutor prompts
Observed improvement. The tutor gives small reminders before the student visibly gets stuck. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Delay prompting and record the first independent move. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If performance drops, the intervention may be masking retrieval weakness through well-timed external cues. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
21. Weak checking hidden by teacher confirmation
Observed improvement. The learner proceeds only after asking whether each step is correct. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Withhold confirmation until the end of a short problem and ask the student to mark their own uncertainty points. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If errors multiply or progress stops, external validation is compensating for weak self-monitoring. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
22. Weak planning hidden by worked-example imitation
Observed improvement. The student can solve after seeing an almost identical example. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Remove the example and ask for a representation or plan before calculation. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the plan disappears, imitation was routing around independent problem formulation. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
23. Weak language processing hidden by tutor paraphrase
Observed improvement. The student succeeds after every word problem is rewritten in simpler language. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Give the original wording plus permission to annotate or diagram it independently. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If the learner cannot extract the structure, paraphrasing was compensation. Teach mathematical reading strategies rather than relying permanently on translation by another person. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
24. Weak memory hidden by formula sheet
Observed improvement. The learner performs well when every needed formula is supplied. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Test which formulas must be recalled in the actual assessment and ask for derivation or meaning where relevant. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. A formula sheet can be a legitimate support, but it should not be mistaken for repaired retrieval if future conditions require memory. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
25. Weak arithmetic hidden by algebraic software or apps
Observed improvement. The learner uses solver tools to complete equations without understanding transformations. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Ask for a simple hand-worked case and an explanation of each equivalence step. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If that fails, software output is masking the dependency rather than repairing it. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
26. Weak current topic hidden by easier question selection
Observed improvement. After intervention, the student is given only the most accessible versions of the topic and appears improved. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Restore representative difficulty and variation gradually. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If errors return immediately, avoidance was part of the intervention. The task distribution changed more than the learner. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
27. Weak examination control hidden by untimed tuition
Observed improvement. The student solves everything in lessons but leaves school papers unfinished. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Use timed sections after knowledge is secure. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If performance drops mainly under time, the intervention repaired content but not examination control. That is a different remaining weak link. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
28. Weak confidence hidden by constant reassurance
Observed improvement. The learner works accurately only when the tutor repeatedly signals that they are on the right path. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Reduce reassurance while preserving a supportive environment and ask the student to use evidence-based self-checks. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If progress stalls, confidence regulation remains externally supported. Repair means the learner can continue through ordinary uncertainty. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
29. Weak transfer hidden by repeated task family
Observed improvement. The student succeeds across many examples that share the same format. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Change representation, wording or context while preserving the mathematical invariant. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. If performance collapses, the intervention improved local familiarity rather than broad repair. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
30. Weak prerequisite truly repaired
Observed improvement. The learner previously failed algebraic fractions, then after intervention succeeds on isolated fresh algebraic-fraction tasks, uses the skill inside equations, retains it after delay and needs fewer prompts. The student may look better on the headline task, but the mechanism of improvement is still open.
Dependency test. Continue to sample the dependency occasionally while reducing dedicated repair time. The test should isolate the original weak link or remove the suspected workaround without introducing unrelated difficulty.
Interpretation. This is strong evidence of true repair: the capability itself improved, spread to downstream work and became less dependent on the intervention context. The goal is not to discredit a useful strategy; it is to describe accurately whether the underlying capability changed.
11. The post-intervention evidence protocol
- Restate the original weak link. What capability was the intervention actually meant to change?
- Test it directly. Use a clean fresh task that isolates the dependency.
- Test it downstream. Use the capability inside the later Mathematics that previously broke.
- Remove or vary the support. Check whether the learner still succeeds when the intervention’s usual scaffold is reduced.
- Delay the retest. Immediate smoothness can be produced by recent support; delayed success is stronger evidence.
- Change the context. Use a different representation, wording or problem family to check portability.
- Track prompting. Repair should generally reduce how much external regulation is required.
This protocol turns “the intervention worked” into a set of observable claims. A family can then decide whether to consolidate, fade support, keep a compensation strategy, or return to the weak dependency.
12. True repair should change more than one task
A repaired lower-floor capability usually improves multiple downstream tasks. Stronger fraction sense can help ratio, percentage and algebra. Stronger algebraic manipulation can improve equations, functions and calculus. Stronger mathematical reading can improve several word-problem families.
Look for this spread. If only the practised intervention task improves, the intervention may be overfitted to the training form.
13. Compensation is sometimes the correct destination
Not every weakness must be repaired to the same degree. In an assessment where a graphing calculator is legitimately available, efficient calculator use may be an appropriate permanent support for certain computations. In real mathematics, diagrams, tables and software are normal tools. A student with a writing or memory constraint may appropriately use supports permitted by the learning environment.
The educational question is whether the compensation preserves access to the Mathematics without creating a hidden failure later. If future work requires the weak capability directly, repair remains important. If the support is authentic and permitted, compensation can be part of a strong system.
14. Avoidance should not masquerade as differentiation
Good differentiation adjusts challenge so a learner can engage productively. Avoidance occurs when the weak capability quietly disappears from the programme because it is uncomfortable or time-consuming to teach. The student may look better simply because fewer tasks require the weak link.
Audit the task distribution. Is the student still encountering representative versions of the Mathematics? Are difficult but essential dependencies being reduced temporarily for teaching—or removed indefinitely?
15. Masking often appears in highly supportive tuition
A skilled tutor can make a student look excellent by asking the right question at the right moment. That is part of teaching. The danger is interpreting guided performance as independent repair. Record the smallest prompt needed and deliberately fade it.
If the student can increasingly generate the same question internally—“what relationship is this?”, “what should I check?”, “which representation helps?”—the external scaffold is becoming internal control.
16. The repair-versus-route-around matrix
Use two axes: dependency strength and task performance. High dependency strength plus high performance suggests true repair or robust learning. Low dependency strength plus high performance suggests compensation, masking or avoidance. High dependency strength plus low task performance suggests another bottleneck downstream. Low dependency strength plus low performance means the original repair is still incomplete.
This matrix prevents a single score from carrying more meaning than it can support.
17. The subtraction test
Ask what happens if the intervention’s special ingredient is removed. If the learner used a diagram, can they create it rather than receive it? If the tutor supplied a checklist, can the student reproduce the checklist from memory? If a calculator was used, can the learner estimate and interpret the output? If a formula sheet was available, can the student still explain the relationship?
The subtraction test should not remove legitimate accommodations just to prove toughness. It is a diagnostic exercise to identify what the learner owns and what the environment supplies.
18. The substitution test
Replace the support with another support that serves the same mathematical function. If a student relies on one specific bar model, can they use a table or equation for the same relationship? If a memorised verbal prompt is removed, can a diagram trigger the concept instead?
Successful substitution suggests the learner understands the function of the support rather than being dependent on one exact tool.
19. The stress test
Once the dependency appears repaired in calm conditions, add modest load: mixed topics, unfamiliar wording, time or a longer multi-step question. Observe whether the old fracture returns. A repaired capability should be more resistant to realistic stress than it was before intervention.
Do not jump directly to maximal difficulty. Stress should reveal robustness, not overwhelm the learner with unrelated demands.
20. The regression test
After several weeks of less intensive support, sample the old weak link again. If performance remains stable, the repair is holding. If the old error returns, consider whether maintenance practice, retrieval or conceptual connections were insufficient.
Regression does not mean the intervention was worthless. It identifies the durability problem that now needs attention.
21. The transfer test
Use the repaired skill in a new context where it remains mathematically relevant. Factorisation can appear in algebraic fractions, graph roots or calculus preparation. Proportional reasoning can appear in rates, scale, science or finance. Graph interpretation can appear in functions, motion or statistics.
A repaired capability should increasingly travel with the learner rather than remain attached to the intervention worksheet.
22. The explanation test for compensation
Ask the student what the support is doing. “I draw a table so I can compare corresponding values,” “I use the calculator to reduce arithmetic load, then I check the sign and size,” or “I write a diagram so I do not lose relationships” shows strategic compensation.
If the student cannot explain the support and simply follows a ritual, the workaround may itself be fragile.
23. The future-dependency test
Look ahead. Will the weak link be required more heavily next term or next level? Weak algebra may become a major cost in Additional Mathematics. Weak function notation may matter greatly in H2 Mathematics. Weak fraction reasoning may continue to affect ratio and algebra. A compensation acceptable today may be unsafe if the next stage removes it.
This is why intervention planning should include the future dependency chain, not only the current assessment.
24. The opportunity-cost test
Repair consumes time. So does maintaining a workaround. Compare the costs. If a student can repair a high-spread algebra weakness in a focused block, that may be more efficient than carrying elaborate compensations through many topics. If a low-spread weakness can be handled reliably by an authentic tool, permanent compensation may be reasonable.
The goal is a sustainable learning system, not ideological purity about doing everything one way.
25. Repair can change confidence for the right reason
Confidence after true repair often becomes more specific. The learner knows why they can begin, what to check and how to recover. Confidence after masking can be more fragile: it depends on the tutor, worksheet or visible example remaining present.
Ask what evidence the student’s confidence is based on. Independent fresh-task success is a stronger foundation than repeated reassurance.
26. A repair record for tutors
Keep a short record with five fields: original weak link, intervention used, direct dependency result, downstream result, and support level required. Add one delayed retest. This is enough to distinguish many repairs from workarounds without excessive administration.
If the intervention is changed, note the reason. Teaching memory becomes more useful when it records mechanisms, not just topics completed.
27. Parent-facing language
A useful update might say: “Her equation scores improved after we introduced a graphing routine, but isolated algebraic manipulation is still weak. The graph is a useful support, but it has not repaired the algebra yet. We are keeping the graph strategy while directly rebuilding the algebra because later A-Math will need it.”
This is more honest and useful than simply saying “the intervention worked.”
28. When the intervention should be stopped
Stop or reduce an intervention when the target capability is stable across fresh, delayed and downstream work and support can fade without relapse. Continuing intensive repair after the weak link is secure can waste time and encourage dependence.
Move the learner back into ordinary maintenance and monitor occasionally.
29. When the intervention should be changed
Change the intervention when the isolated dependency shows little movement despite good implementation, when the student is learning the scaffold rather than the concept, when task performance rises but future-dependent capability remains flat, or when the intervention creates excessive cognitive load.
A change should be based on evidence about mechanism, not boredom alone.
30. When the intervention should be intensified
Intensify when the weak link is high-spread, continues to block multiple topics, and shows some response to teaching but insufficient stability. Intensification can mean more explicit instruction, more frequent retrieval, smaller steps, stronger representation links or closer feedback—not simply more worksheets.
The intervention should remain targeted enough that progress can be read.
31. How representation can be both repair and compensation
A representation can repair understanding by making a hidden relationship visible. It can also compensate by carrying information the student cannot yet hold mentally. These functions are not mutually exclusive.
The test is what happens later. Can the learner connect the representation to symbols, create it independently and eventually reason without the teacher supplying it? If yes, the representation may have helped build repair. If no, it remains primarily a support.
32. How technology can be both repair and compensation
Technology can provide feedback, visualisation and repeated practice that strengthens understanding. It can also bypass arithmetic, symbolic manipulation or graph reasoning. Judge the result at the capability level.
A graphing tool that helps a student see how parameters change a curve may support conceptual repair. A solver that produces answers without the student understanding the equation may only route around the weakness.
33. How memorisation can be both repair and compensation
Memorising essential facts and formulas can reduce cognitive load and support fluent mathematics. Memorising a sequence of moves can also conceal missing meaning. The distinction appears when conditions change.
If the learner can explain, select and adapt the memorised knowledge in fresh contexts, memory supports repair. If the sequence fails as soon as wording changes, it is narrow compensation.
34. How scaffolding should fade
A scaffold earns its place by making a capability possible before it is independent. Fading should be planned: full worked example → partial example → prompt → self-prompt → independent task. The pace can vary by learner.
If the scaffold never fades, ask whether it is actually an accommodation that should remain, or whether repair has stalled.
35. A three-student small-group audit
In a small group, compare how much support each learner needs for the same dependency. One may require a diagram, another a verbal prompt, another nothing. Then give slightly varied fresh tasks independently before group discussion.
This prevents peer answers from masking individual repair while still preserving the benefits of collaborative explanation.
36. Evidence-informed references
Useful sources include the What Works Clearinghouse guide Assisting Students Struggling with Mathematics: Response to Intervention for Elementary and Middle Schools, which includes progress monitoring and systematic intervention recommendations; the newer Assisting Students Struggling with Mathematics: Intervention in the Elementary Grades, which emphasises systematic instruction, mathematical language and well-chosen representations; and the Algebra Knowledge practice guide, which highlights algebraic reasoning and structure. These guides apply to defined populations and should inform principles rather than be treated as direct prescriptions for every Secondary, Additional Mathematics or JC learner.
For the wider evidence question, use Did the Teaching Change Cause the Improvement?. For recurring fractures, use The Mathematics Fracture and Repair Map.
37. Frequently asked questions
Is a workaround always bad in Mathematics?
No. Calculators, diagrams, tables, formula sheets and alternative methods can be legitimate mathematical tools. The question is whether the workaround is appropriate for the assessment and whether the underlying weak capability will still be needed later.
How do I know if a weak link was repaired?
Test the capability directly, then test it inside downstream Mathematics, after some delay, with reduced prompting and in at least one fresh context.
What if performance improved but the isolated skill did not?
That suggests compensation, masking or avoidance. Decide whether the support can remain legitimately or whether the dependency still needs direct repair.
Can a student have both repair and compensation?
Yes. A representation may strengthen understanding while also reducing memory load. A calculator may support performance while number sense improves separately. Mixed outcomes are common.
Should support always be removed?
No. Remove or vary support diagnostically when appropriate. Legitimate accommodations and authentic mathematical tools may remain. The purpose is to know what the learner owns and what the environment supplies.
When should an intervention end?
When the target capability is stable enough across fresh, delayed and downstream work that intensive support no longer produces useful additional return.
38. The shortest useful answer
An intervention repaired the weak link when the capability itself becomes stronger: it works in isolation, appears in downstream Mathematics, survives fresh and delayed tasks, and requires less external prompting. If performance improves mainly because another route avoids the weak capability, the intervention produced compensation.
Compensation can be excellent. The mistake is not using it. The mistake is believing the weak link has disappeared when future Mathematics may still depend on it.
39. Twenty-four advanced cases: what exactly did the intervention change?
1. The student’s score rises because the tutor pre-selects the questions
Observed state. The intervention feels successful because the student is now completing a carefully curated set that avoids the most diagnostic versions of the weak skill. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Restore a representative spread gradually, including some questions that require the original dependency in less supported forms. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If performance falls only when the omitted forms return, the programme has been managing exposure rather than repairing the dependency. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
2. The learner can now start because a checklist is always visible
Observed state. A structured checklist reduces blank-page paralysis and produces better work. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Ask the student to recreate the checklist from memory, then shorten it to internal questions and finally remove it on a fresh task. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If the learner can generate the sequence independently, the scaffold has been internalised. If not, the checklist remains a useful but external control system. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
3. A new mnemonic fixes recall but not meaning
Observed state. A mnemonic allows the student to remember an order of operations, theorem condition or statistical test sequence. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Use a changed problem where one step should be omitted or modified and ask why. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If the learner follows the mnemonic rigidly when conditions change, memory has compensated for understanding. If they can adapt it, the mnemonic may be supporting genuine repair. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
4. A diagram solves every word problem but the student cannot build one alone
Observed state. Teacher-created diagrams produce excellent performance. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Give the student a fresh problem and a blank page, asking only what should be represented. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If the learner can now construct a useful diagram, the intervention built representation skill. If not, the diagram remains a supplied workaround. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
5. A stronger peer carries the group
Observed state. The student performs better in small-group work because another learner identifies methods first. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Require silent first attempts and separate representations before discussion. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If independent entry remains weak, peer cues were masking method-selection difficulty. Group learning can continue after individual evidence is collected. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
6. The student has learned to avoid the weak method entirely
Observed state. An alternative method produces correct answers and is often more efficient. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Check whether the avoided method is itself a future prerequisite or merely one optional route. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If future Mathematics does not require it, the alternative may be a legitimate replacement. If the avoided skill supports later topics, direct repair remains necessary. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
7. The intervention improves one chapter but not the shared prerequisite
Observed state. A student now succeeds in one equations unit but the same algebraic weakness appears in functions and graphs. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Test the shared algebra skill directly and across two downstream topics. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Local improvement suggests chapter-specific compensation or memorisation. True prerequisite repair should spread more broadly. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
8. The student reads better because the tutor highlights key words
Observed state. Highlighted phrases help the learner translate word problems accurately. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Remove highlighting and ask the student to mark quantities, relationships and conditions themselves. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. The support becomes repair when the learner can perform the highlighting function independently rather than needing the tutor to pre-process the text. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
9. The learner checks correctly because every task ends with a reminder
Observed state. A ‘check your answer’ box reduces careless errors. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Remove the printed cue on some fresh items and observe whether checking still occurs. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If checking persists, the external cue has become a habit. If it disappears completely, the intervention changed the environment more than the learner. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
10. The student succeeds only with one representation
Observed state. A bar model, table or graph is now used successfully every time. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Ask the learner to solve one matched problem using another legitimate representation or to explain when the preferred representation would be unhelpful. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If understanding survives representation change, repair is deeper. If not, the representation may be compensating for a concept that remains narrow. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
11. The student performs only with extra time
Observed state. Extended time produces accurate mathematics while standard timing remains incomplete. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Determine whether the learning goal is conceptual mastery or examination performance, and test each separately. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Extra time can legitimately reveal knowledge that standard timing hides. It does not, by itself, repair fluency or paper-control weakness if those are required outcomes. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
12. The learner stops making sign errors because the tutor circles every negative sign
Observed state. Visual marking reduces errors dramatically. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Ask the student to create their own sign-marking system, then gradually reduce it as accuracy stabilises. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. The intervention repairs self-monitoring only when the student can initiate the control without external marking. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
13. The student passes because only calculator-permitted tasks are practised
Observed state. Performance rises in a course where some non-calculator reasoning still matters. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Include representative non-calculator or exact-reasoning tasks where required by the curriculum. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If weakness reappears there, calculator use was a route around the dependency rather than a full repair. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
14. The learner’s confidence improves but uncertainty tolerance remains low
Observed state. Reassurance produces better attempts, yet the student still freezes when no one confirms the first step. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Use bounded independent starts and ask the learner to mark confidence rather than request immediate confirmation. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Repair means the student can proceed through ordinary uncertainty using evidence, not that all reassurance disappears. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
15. The intervention fixes homework but not tests
Observed state. Supported practice becomes accurate while timed mixed papers still expose the old weakness. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Compare the same dependency in calm, mixed and timed conditions. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. The concept may be repaired while performance access remains fragile. This is a downstream control problem, not evidence that the repair failed entirely. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
16. The intervention fixes tests but not delayed recall
Observed state. Short-term exam performance rises after intensive rehearsal, then the skill fades. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Retest after several weeks with no special review immediately beforehand. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. A durable repair should remain reasonably accessible. If it vanishes, spacing and retrieval need strengthening. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
17. A formula derivation is understood but never retrieved
Observed state. The student can follow why a formula works but cannot use it later without notes. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Use low-stakes delayed retrieval and then reconnect the derivation when memory fails. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Conceptual repair and retrieval are different layers. The intervention may have improved meaning without yet creating durable access. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
18. The learner knows the prerequisite but still fails downstream
Observed state. Direct tests show strong algebra, yet the later topic remains weak. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Stop repairing the prerequisite and diagnose the downstream concept, representation or method selection. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. This is an important false-positive control: not every later failure proves the lower floor is still broken. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
19. The learner uses a shortcut that works only on special cases
Observed state. A trick produces rapid success on common examples. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Include a case where the shortcut fails and ask for the general method. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If the learner cannot distinguish the shortcut’s boundary, the intervention has created brittle compensation rather than robust repair. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
20. The learner uses an authentic expert shortcut
Observed state. A concise method replaces a longer school procedure but is mathematically valid and general. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Test the shortcut across representative conditions and ask for justification. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. If it remains valid, efficient and explainable, this is not a problematic workaround. It is strategic mathematical development. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
21. The intervention changes motivation more than skill
Observed state. The student attempts more questions and persists longer, but isolated accuracy changes little. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Track attempt rate, time-on-task and direct dependency performance separately. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Motivational improvement is valuable and may enable later repair. It should be reported as motivation and persistence rather than prematurely as skill repair. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
22. The intervention changes communication more than understanding
Observed state. Written working becomes clearer and marks rise because the method is now visible. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Use a fresh oral or mental probe to see whether the underlying concept was already present. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Clear communication is a real capability gain. The score increase may come from making existing understanding visible rather than repairing the concept itself. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
23. The intervention changes checking more than core knowledge
Observed state. The student now catches and fixes many own errors before submission. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Compare first-attempt error rate with final submitted accuracy. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. Self-correction is genuine learning even if initial slips persist. Name the repaired layer accurately: monitoring improved, while execution may still need work. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
24. The repair survives when the intervention disappears
Observed state. Weeks after intensive support ends, the student still performs the dependency directly, uses it downstream, detects old error traps and rarely needs prompts. Better performance is real, but the mechanism still needs to be named correctly.
Audit. Sample the skill occasionally but return most time to current curriculum. The audit should preserve dignity and legitimate support while isolating the capability that was originally weak.
Interpretation. This is the desired endpoint: the intervention can shrink because the learner now carries the capability. The next teaching decision should follow the layer that actually changed rather than the headline mark alone.
40. A four-week fade-and-recheck schedule
Week 1: use the intervention fully and record direct dependency performance. Week 2: reduce one scaffold while keeping the task comparable. Week 3: place the dependency inside a fresh downstream problem. Week 4: retest after delay with ordinary support conditions. This schedule does not fit every intervention, but it illustrates the principle: repair should be tested as support fades and context widens.
If performance remains strong, the intervention can often move into maintenance. If it falls at a specific stage, the point of failure tells you whether the next target is retrieval, transfer, independence or downstream application.
41. The support inventory
List every support currently present: tutor prompts, peer cues, notes, formula sheets, calculator, diagrams, worked examples, extra time, colour coding, checklists, simplified language and task selection. Mark each as one of three types: authentic tool, temporary scaffold or diagnostic accommodation.
This inventory prevents accidental masking. A student may be using several supports simultaneously, making it impossible to know which capability improved until the supports are understood.
42. The repair scorecard without a score
Direct capability: stronger or unchanged? Fresh-task use: survives or collapses? Delayed retrieval: available or fragile? Downstream spread: improves related topics or stays local? Prompt dependence: falling or stable? Support portability: can the learner generate the support independently?
These descriptors are often more useful than a single intervention percentage because they tell the tutor exactly what to do next.
43. Repair should make the learning system cheaper
A good intervention usually reduces future instructional cost. The tutor needs fewer reminders. The learner can start sooner. Corrections become smaller. More attention is available for current topics. Downstream errors caused by the same dependency become less frequent.
If the intervention requires permanent intensive management merely to maintain ordinary performance, ask whether the capability was repaired or whether the system is continuously compensating for it.
44. Final principle
The strongest post-intervention question is not “Did the mark go up?” It is “What can the learner now do that they could not do before, and under what conditions can they still do it?” That question keeps repair, compensation, avoidance and masking distinct.
A workaround can be intelligent. A scaffold can be necessary. An accommodation can be appropriate. But cumulative Mathematics depends on knowing which foundations are genuinely stronger and which are still being carried by the environment. Diagnose that difference before the next layer places more load on the same weak link.
45. The final weak-link audit before an intervention is declared complete
Before closing an intervention, run four short checks. First, isolate the original dependency and test it directly on fresh items. Second, place it back inside the downstream Mathematics that originally exposed the problem. Third, reduce the special support that was added during intervention and observe whether performance remains stable. Fourth, retest after enough time has passed that the student is retrieving rather than continuing a recently rehearsed routine.
The four checks answer different questions. Direct isolation asks whether the weak capability itself changed. Downstream use asks whether the repair matters where the curriculum needs it. Support reduction asks whether the learner owns more of the work. Delay asks whether the repair is durable. A student does not need perfection on every check, but the pattern should be clearly stronger than before intervention.
If the direct skill is strong but downstream work is still weak
Stop blaming the repaired prerequisite automatically. Diagnose the next layer. The later topic may have its own concept, representation, method-selection or examination-control problem. Continuing to remediate an already-secure floor can waste time and reduce confidence.
If downstream work is strong but the direct skill is still weak
The system is compensating successfully. Decide whether that is enough. If the weak skill will be required independently later, continue direct repair while preserving the useful workaround. If the compensation is authentic, permitted and future-safe, maintaining it may be entirely reasonable.
If both direct and downstream performance are strong but prompts remain heavy
The mathematical knowledge may be substantially repaired while independence is unfinished. Shift the intervention toward prompt fading, self-questioning and delayed feedback rather than reteaching the content from the beginning.
If everything works only immediately after intervention
The next target is durability. Use spaced retrieval, cumulative review and fresh re-entry rather than increasing the intensity of same-day practice. A repair that exists only while it is warm is not yet a reliable foundation.
46. Closing answer
A Mathematics intervention should be judged at the capability it was meant to change. Better marks, smoother worksheets and higher confidence are important outcomes, but they do not by themselves prove that the original weak dependency was repaired. Test the dependency directly, test it downstream, vary the support, add delay and watch whether prompting falls.
If the weak link itself is stronger, the intervention repaired. If another route carries the student successfully, the intervention compensated. If the weak link disappears only because tasks stopped requiring it, the system avoided it. If correctness depends mainly on live external help, the support is masking it. Naming the state accurately is what allows the next educational decision to be intelligent.
47. The minimum proof of repair
If only one final check is possible, use a fresh task that isolates the original dependency, remove the intervention’s special cue where appropriate, and ask the learner to explain the first decision before completing the work. Then sample the same dependency later inside a downstream topic. This small pair of checks is not a perfect experiment, but it separates many genuine repairs from local supported performance.
The practical standard remains simple: the capability should be more available, more accurate, less prompt-dependent and more portable than it was before the intervention. When those changes are visible, the evidence supports repair. When performance improves mainly because the environment carries the missing capability, report compensation honestly and decide whether that support is suitable for the learner’s next mathematical stage.