If a child can do the sums but not the problem sums, the arithmetic may not be the problem. The learner can add, subtract, multiply, divide, work with fractions or perform algebraic procedures when the operation is already known, but becomes stuck when the same mathematics is hidden inside language and relationships.
Parents often describe this as “weak problem solving” or “doesn’t understand word problems”. That can be true, but it is still too broad. A problem sum adds several jobs before calculation: read the language, identify quantities, infer relationships, decide what is unknown, choose a representation, select a method and only then execute the arithmetic.
At eduKate Sengkang, this symptom is diagnosed from story → quantities → relationship → representation → method → calculation → verification. The first broken link matters. If we teach only keywords such as “more means add”, the child may become faster on familiar worksheets while remaining unable to model new problems.
Before choosing the repair: locate the learner state
The visible symptom is only the entrance. Ask what this capability can currently survive: Blocked — the learner cannot reliably start or choose a method; Fragile — it works today but does not reliably return; Stable — it holds on familiar work; Transfer-ready — it survives changed wording, representation or context; Examination-ready — it survives time pressure, mixed questions and paper-level interference.
These are states of a capability, not labels for the child. Use the Learning Atlas V2.0 to move from symptom → state → first weak link → action → evidence → retest.
Quick answer: what should parents test first?
- Give one problem sum without asking the child to calculate.
- Ask: What quantities are here?
- Ask: What is the question asking us to find?
- Ask: How are the quantities related?
- Ask the child to draw, tabulate or write an equation for the relationship.
- Only then calculate.
- Change the story while keeping the mathematical relationship the same.
If the relationship can be represented but arithmetic then fails, the issue is calculation. If arithmetic is strong but the relationship cannot be represented, the problem is translation or modelling.
Why keywords fail
Words such as “more”, “left”, “altogether” and “difference” can be useful clues, but they do not uniquely determine an operation. “Ali has 5 more than Ben” describes a relationship; whether you add or subtract depends on which quantity is unknown. Keyword matching teaches a surface shortcut where modelling is needed.
The seven-link problem-sum chain
| Link | Question | Typical failure |
|---|---|---|
| 1. Language | What does each sentence mean? | Important comparison or condition is misread. |
| 2. Quantities | What values or unknowns exist? | Numbers are used without roles. |
| 3. Relationship | How are the quantities connected? | Child hunts for operations instead of structure. |
| 4. Representation | Can the relation be shown? | Diagram/model/table/equation does not match the story. |
| 5. Method | What mathematical operation or sequence fits? | Known methods are selected from keywords. |
| 6. Execution | Can the arithmetic/algebra be carried out? | Local calculation errors. |
| 7. Verification | Does the answer fit the story? | Impossible or mismatched answer is accepted. |
A simple diagnostic example
Suppose one quantity is 24 and another is three times as large. The learner may know multiplication facts perfectly and still fail if “three times as large” is not represented as a multiplicative relationship. Reverse the unknown — give the larger quantity and ask for the smaller — and keyword strategies often break. Relationship-based understanding survives the reversal.
Primary 1–2: protect relationship language
Young learners need concrete and visual representations tied to language: part and whole, comparison, equal groups, repeated addition, sharing and grouping. The goal is not to rush into model drawing as a ritual. The drawing should express the relationship the child can explain.
Primary 3–4: multi-step problems expose hidden decisions
As problem sums become multi-step, a correct first operation can still lead to failure if the learner does not know what the intermediate result represents. Ask the child to label the meaning of each result before moving on. This reduces “blind calculation chains”.
Primary 5–6 and PSLE: ratio, percentage, speed and changing unknowns
Upper Primary problem solving increases the number of possible representations and relationships. The same surface topic can require different models. A learner who memorises templates may appear strong until the unknown changes position or two familiar structures combine.
Use fresh problems where the numbers and story change but the relationship remains, then contrast them with similar-looking problems where the relationship changes. Existing deeper routes include Primary 5–6 and PSLE word problems and Primary 3–5 word-problem translation and models.
The representation test
Ask the learner to solve the same relationship in three forms: a short story, a bar model or diagram, and an equation. The learner does not need to prefer every form equally. The important evidence is whether meaning survives translation.
| If the learner can… | But cannot… | Likely next job |
|---|---|---|
| Calculate once the equation is given | Build the equation from the story | Translation / relationship modelling |
| Draw a model | Explain what each part means | Representation may be procedural rather than meaningful |
| Explain the story verbally | Turn it into a model/equation | Representation bridge |
| Choose the method | Execute accurately | Calculation / algebra repair |
| Solve the original | Solve a reversed unknown | Transfer / structural understanding |
What not to do
- Do not teach “circle all numbers and choose an operation”.
- Do not assume more problem sums will repair a representation gap.
- Do not leave the model beside every fresh problem indefinitely.
- Do not praise a correct answer if the child cannot explain the relationship.
- Do not turn every difficult problem into a memorised template.
- Do not skip verification: the answer should return to the story.
A six-question repair set
- One direct familiar problem.
- The same relationship with different numbers.
- The same relationship with a different context.
- The same relationship with the unknown reversed.
- A similar-looking problem requiring a different relationship.
- A mixed problem combining the target with one known second structure.
This sequence reveals whether the learner has only procedural familiarity or genuine structural control.
When the child freezes before drawing anything
If the learner cannot identify quantities or the question demand, the issue may be reading access or the Blocked learner state. If they can model the problem today but cannot do so after a few days, use the Fragile learner route. If familiar models are reliable but changed contexts fail, use Stable and Transfer-ready.
How parents can help without doing the problem
- Ask “What do we know?” and “What do we need?” before suggesting an operation.
- Ask the child to name each quantity’s role.
- Ask for a drawing or equation only after the child explains the relationship.
- If the child is stuck, give one structural cue, not the full first step.
- After solving, ask whether the answer is plausible in the story.
- Return later with a fresh context instead of repeating the identical question.
How a three-student tutorial helps
Three learners can solve structurally related problems using different representations, then compare. One may draw a model, another write an equation, and another verbalise the relationship. The tutor can test whether each learner understands the invariant rather than copying a preferred template.
Frequently asked questions
Should my child memorise model-drawing types?
Models are useful when they express relationships. Memorised types can help organise early learning, but the learner should eventually choose and adapt representations rather than wait for an exact template.
Is this a reading problem or a Maths problem?
It can be either, or both. The chain helps distinguish them. If the learner cannot explain the sentence meaning, reading access may be the first weak link. If language is clear but relationship modelling fails, the mathematical representation is the target.
Should we practise harder questions?
Only after the current weak link is clear. A harder question that adds several new demands can hide what needs repair.
Why can the child solve it after I show the first step?
That suggests the later arithmetic may be intact and the block sits in entry, representation or method selection. Use less and less cueing until the learner generates the first step independently.
Where this page sits in Atlas V2.0
This is a Mathematics symptom owner under When Learning Slips. It hands topic depth into the Mathematics estate and specialist Mathematics where needed, while the Atlas keeps the learner-state route visible. See Learning Atlas V2.0.
Last updated: 26 September 2026.
Technical owner | where this symptom goes next
This is a Mathematics transfer and representation problem. Continue to the Mathematics learning system to separate arithmetic fluency from problem representation, model choice, multi-step reasoning and method selection. When mathematical representation itself needs specialist depth, hand off to BukitTimahTutor’s World Mathematics Atlas.
Complete the learning loop: use the six-question repair set on this page, then move to Learning Practice & Review for a fresh attempt after feedback. Change the numbers, surface story or representation before accepting success. Use How We Know Learning Has Really Held to test delay and transfer rather than treating the immediately repeated problem as proof.
Return to the Learning Atlas V2.0 whenever the evidence changes the state or the first weak link.
Symptom return routes | Atlas V2.0
- Start Here — return to the symptom library if this is not the only thing happening.
- Learning Runtime — classify whether the capability is Blocked, Fragile, Stable, Transfer-Ready or Examination-Ready.
- Parents’ Guide — decide what the family should start, change, reduce or stop.
- The Tutor System — use when a tutor needs to act on the diagnosed weak link.
- Mathematics Hub — return to the Sengkang Mathematics route when the weak link is mathematical.
- BukitTimahTutor — deeper specialist Mathematics depth when required.
- Learning Atlas V2.0 — return to the whole route map and retest the learner after the intervention.
Word-Problem Difficulty | Separate Language, Relationship and Calculation
When a child can calculate accurately but still fails problem sums, do not treat “word problems” as one skill. Separate three jobs: Language — can the learner understand the task and its conditions? Relationship — can the learner identify how the quantities are connected and choose a useful representation? Calculation — once the mathematical structure is clear, can the learner execute accurately and verify the result?
This separation matters because the same wrong answer can come from three completely different weak links. A child who misreads “12 more than” needs a different repair from a child who understands the sentence but cannot model the comparison, and both need a different repair from a child who builds the correct equation but calculates incorrectly.
1 | Language: can the learner decode the task without giving away the Mathematics?
Language work begins before arithmetic. The learner must identify who or what is being compared, the quantities involved, the order of events, conditions such as “each”, “remaining”, “altogether”, “per”, “twice”, “more than”, “less than”, “after”, “before” and “at least”, and the exact quantity the question asks for.
Do not diagnose language from keywords alone. “More” does not always mean add; “left” does not always mean subtract. The important question is whether the learner can paraphrase the situation without changing the relationship. Ask: “Tell me what is happening, without solving it yet.”
If procedural or condition language is the bottleneck, route deeper language work to SETC: From Instructions to Action. Its job is language → execution: sequence, conditions, roles, constraints and task completion. Then return to the Mathematics problem so the learner still has to build the relationship.
2 | Relationship: can the learner see the mathematical structure?
Once the wording is understood, the learner must decide what relationship exists. Is this part–whole, change, comparison, equal groups, sharing, rate, ratio, percentage, repeated identity, before-and-after, or a multi-stage combination? The learner then chooses a representation: bar model, table, number line, diagram, equation or another valid structure.
This is often the true weak link when a child “can do sums”. Arithmetic worksheets tell the operation in advance. Problem sums require the learner to select the operation or model from the relationship. Topic labels and familiar worksheet patterns can hide that selection difficulty.
For specialist depth, use BukitTimahTutor: Singapore Math Word Problems | From Language to Mathematical Structure. BTT owns the deeper representation/model/structure work. Sengkang keeps the learner-state diagnosis and the return to a fresh problem.
3 | Calculation: can the learner execute once the structure is supplied?
Only after language and relationship are controlled should calculation become the main suspect. Give the learner the correct representation or equation without doing the arithmetic. If the learner still makes number-fact, fraction, unit, sign, place-value or procedural errors, the calculation layer needs repair.
If calculation is accurate once the structure is supplied, do not prescribe pages of arithmetic practice as the main solution. The evidence says the bottleneck sits earlier. If the calculation fails, route to the relevant Mathematics dependency and then rebuild the complete problem afterward.
A four-test diagnostic that parents and tutors can actually use
| Test | What changes | What it helps separate |
|---|---|---|
| A · Original fresh problem | Normal wording; no help. | Baseline: where does the route first fail? |
| B · Same mathematical structure, clearer language | Simplify procedural wording without stating the operation or relationship. | If performance improves, language load may be contributing. |
| C · Relationship/representation supplied | Give the correct model, diagram or equation; learner calculates. | If performance becomes accurate, representation/method selection—not arithmetic—is likely the larger bottleneck. |
| D · Fresh changed problem | New surface/context; support recorded and reduced where appropriate. | Tests whether the repair transfers rather than whether the child remembers the earlier question. |
Do not run all four as a high-pressure test. They are comparison conditions. Record what help was given. A correct answer after an adult supplies the relationship is useful evidence, but it is not independent problem-solving evidence.
The support ledger
| Support used | What the result can show | What it cannot yet show |
|---|---|---|
| Instruction reread only | Whether attention to wording restores the route. | Independent reading if adult rephrased meaning. |
| Vocabulary/condition clarified | Mathematical reasoning after language barrier is removed. | Independent interpretation of that language. |
| Representation prompt: “Could a diagram help?” | Whether a generic strategic cue unlocks selection. | Fully independent representation choice. |
| Correct model/equation supplied | Calculation capability under supplied structure. | Independent relationship modelling. |
| Worked solution shown | Understanding/imitating after modelling. | Fresh independent problem solving. |
Worked example: comparison language
Question: “Mei has 18 stickers. She has 7 more stickers than Arun. How many stickers does Arun have?” A keyword rule can fail because the word “more” appears even though the unknown requires subtraction. The language step is to identify the comparison: Mei = Arun + 7. The relationship step is to decide that Arun is the smaller quantity. The calculation step is 18 − 7.
If the learner says “25” immediately because “more means add”, the first repair is relationship interpretation, not subtraction practice. Contrast it with: “Arun has 7 more than Mei” and ask the learner to state who is larger before touching the numbers.
Worked example: multi-step calculation hiding a representation problem
A learner accurately calculates fractions and percentages separately but fails a Primary 5 problem where a fraction is used first and a percentage applies to the remainder. If the learner calculates each step correctly once the sequence is drawn, arithmetic is not the primary problem. The profile should identify changing reference quantity / representation, then retest on a fresh story where the whole changes again.
Worked example: language genuinely blocks the Mathematics
A learner understands equal groups when shown pictures but misreads “each”, “per” and “for every” in word problems. Here language is not incidental. Run a small language contrast: “4 bags with 6 apples each”, “6 apples in each of 4 bags”, “4 groups of 6”. If the same structure becomes available after wording clarification, connect SETC language work to BTT/Sengkang representation work rather than treating the child as globally weak in problem solving.
Why keyword teaching is a fragile shortcut
Keyword rules can help a beginner notice language, but they become dangerous when the keyword replaces relationship reasoning. Real questions use the same word in different structures, omit familiar keywords, contain irrelevant language or require more than one operation. A Transfer-Ready learner should be able to justify the relationship even when the usual cue is missing.
The fresh-problem acceptance test
The repair is not complete because the learner can now solve the corrected original problem. Give a fresh question with the same underlying relationship but changed names, numbers, wording or representation. Record whether support was needed. Then revisit after a delay or inside mixed work.
Use Learning Practice & Review to choose the next task and How We Know Learning Has Really Held to test delay and transfer. If you need a precise learner-specific record, use A Capability Profile Without Labelling the Child.
Parent decision: what should we practise next?
- If simpler wording restores the route but normal wording does not, strengthen task language while keeping the Mathematics target visible.
- If the learner understands the story but cannot represent the relationship, practise contrast cases and representation/model selection.
- If the learner builds the right representation but calculates inaccurately, repair the calculation dependency.
- If all three work in isolation but mixed papers fail, route to transfer/examination integration rather than reteaching the topic.
- If heavy adult prompting is required, record it and fade systematically before calling the capability independent.
The key is to repair the earliest unreliable link. Language → relationship → calculation is a diagnostic sequence, not a rigid teaching order. Once the weak link is repaired, the child returns to a complete fresh problem so the parts reconnect.
