A marked answer can tell us that something went wrong.
A well-chosen next question can tell us what went wrong.
That difference is the heart of mathematical diagnosis.
Suppose a Primary 5 learner gets a percentage problem wrong.
There are several possible explanations:
- the learner does not understand percentage as a fraction out of 100;
- the learner understands percentage but chose the wrong reference quantity;
- the learner chose the correct method but made an arithmetic error;
- the learner interpreted the wording incorrectly;
- the learner can solve familiar percentage questions but not a changed representation.
Giving twenty more percentage questions before distinguishing these possibilities is inefficient.
A diagnostic decision tree uses each answer to decide what question should come next.
Quick read: the next question should separate competing explanations
- Inspect the learner’s working, not only the final answer.
- Identify two or three plausible causes.
- Choose a small follow-up question that gives different outcomes under those causes.
- Observe both the answer and the reasoning.
- Route to the smallest repair that explains the evidence.
- Retest with a changed surface to check transfer.
The best next question is often not harder.
It is more discriminating.
What a decision tree is doing
A decision tree is a sequence of conditional questions.
In Mathematics teaching, the logic can be simple:
If the learner fails this simple representation test, inspect concept or representation. If the learner passes it but fails the full problem, inspect method selection, procedure or transfer.
The purpose is not to label the learner permanently.
The purpose is to narrow uncertainty.
Begin with the earliest visible break
Consider the working:
Red:blue = 2:3, total = 40.
Student writes:
2/3 × 40.
The arithmetic has not even begun, but the representation is already wrong.
The first useful follow-up is not another long ratio word problem.
Ask:
Red:blue = 2:3. What fraction of the total is red?
If the learner says 2/3, the ratio-to-fraction misconception is confirmed.
If the learner says 2/5, then the original error may have come from reading or haste rather than concept.
Decision tree 1: wrong final answer with clear working
When the learner shows working, ask:
- Is the first representation valid?
- If yes, is the selected method appropriate?
- If yes, are the procedural steps valid?
- If yes, is the arithmetic correct?
- If yes, are units and requested quantity correct?
- If yes, did checking fail to catch a final transcription or rounding error?
This prevents the common diagnostic mistake of starting at the last wrong line instead of the first one.
Decision tree 2: blank working
A blank answer is much more ambiguous than it looks.
Possible causes include:
- the learner does not understand the concept;
- the learner cannot translate the wording;
- the learner cannot choose a starting representation;
- the learner knows the method but could not retrieve it under pressure;
- the learner strategically skipped the question;
- the learner ran out of time.
The next question should therefore reduce the task.
Ask the same relationship in a simpler form.
If the original was a multi-step percentage question, ask one direct percentage-of-a-quantity question.
If that succeeds, restore the original representation but remove the time pressure.
If that succeeds, the problem may be retrieval, language or examination execution rather than concept.
Decision tree 3: arithmetic or concept?
Suppose a learner solves:
8 units = 64
1 unit = 7
The unitary structure is correct. The division is not.
Do not reteach ratio immediately.
Ask a short arithmetic discriminator:
56 ÷ 8 = ?
72 ÷ 8 = ?
If those fail, arithmetic fluency is implicated.
If those succeed easily, the original error may have been a one-off slip.
Do not use a broad reteaching intervention when a narrow discriminating test can show that the concept is already stable.
Decision tree 4: fraction misconception or procedure error?
A learner writes:
1/3 + 1/4 = 2/7.
Possible explanation A:
the learner treats numerator and denominator as separate whole numbers.
Possible explanation B:
the learner knows common denominators but forgot the procedure.
Ask:
Which is larger: 1/3 or 1/4? Explain without calculating a decimal.
If the learner reasons correctly about unit fraction size, the conceptual base may be stronger than the addition procedure suggests.
Then ask:
What fraction is equal to 1/3 with denominator 12?
What fraction is equal to 1/4 with denominator 12?
The path of answers separates equivalence knowledge from addition procedure.
Decision tree 5: decimal place value or calculation?
A learner says 0.52 > 0.8.
Do not begin with decimal subtraction.
Ask:
- Which is larger: 52 hundredths or 80 hundredths?
- Place 0.52 and 0.8 on a number line between 0 and 1.
- Write 0.8 in hundredths.
If the learner can answer all three, the original comparison error may be unstable rather than deeply conceptual.
If the learner continues comparing digit strings, return to place-value representation.
Decision tree 6: area or perimeter confusion?
A learner adds all side lengths when asked for area.
Ask two concrete questions:
- What would we measure if we wanted to put a fence around the shape?
- What would we measure if we wanted to cover the inside with tiles?
If the learner can distinguish the physical jobs but selects the wrong formula in symbols, the issue may be vocabulary or retrieval.
If the learner cannot distinguish the jobs, rebuild the concept using boundary and covering representations.
Decision tree 7: word problem language or mathematics?
A learner fails a question containing “18 more than”, “remaining”, “each”, or “of the remainder”.
Rewrite the same relationship with simpler language while preserving the numbers.
If the learner now succeeds, the mathematical method may be stable and the bottleneck is interpretation.
If the learner still fails, simplify further into a diagram or model.
The decision tree becomes:
original wording → simplified wording → model → direct arithmetic relationship.
The first level at which performance stabilises locates the translation boundary.
Decision tree 8: method selection or method execution?
Suppose a learner can solve ratio questions in a worksheet labelled “Ratio” but fails a mixed paper.
Ask one unlabeled question that could plausibly be solved using fraction, ratio or percentage.
Before calculation, ask:
Which relationship do you think this is, and what evidence in the question tells you?
If the learner selects correctly but calculates poorly, execution needs repair.
If the learner cannot select, more mixed discrimination is needed.
Decision tree 9: geometry fact or diagram reading?
A learner misses an angle question.
Ask the underlying fact without the full diagram:
What is the sum of the interior angles of a triangle?
If the learner knows 180°, the problem may be identifying which triangle or angle relationship in the composite diagram is relevant.
Then isolate the relevant triangle visually.
If the learner now solves it, the gap was representation rather than fact recall.
Decision tree 10: speed formula or unit control?
A learner writes distance = speed × time correctly but gets the answer wrong.
Before reteaching speed, inspect units.
If speed is in km/h and time is 30 minutes, ask:
How many hours is 30 minutes?
If the learner says 0.5 h, the unit relationship is available.
If the learner multiplies by 30 directly, the issue is unit alignment.
A general decision tree for Primary Mathematics
| Observation | Next discriminating question | Likely route |
|---|---|---|
| Wrong representation | Can the relationship be shown with a simpler model? | Concept / representation |
| Correct model, wrong method | What quantity would your chosen operation produce? | Method selection |
| Correct method, wrong arithmetic | Can the same arithmetic be done in isolation? | Fluency / slip |
| Blank answer | Can the same relationship be solved in a simpler representation? | Concept / retrieval / language / time |
| Works only with chapter labels | Can the learner select among near-neighbour methods? | Transfer |
| Works untimed, fails timed | Can the learner complete a short timed set accurately? | Retrieval / pacing / exam execution |
The rule of minimum intervention
Use the smallest intervention that explains and repairs the observed failure.
If one arithmetic fact is unstable, do not reteach the entire topic.
If the learner has a deep fraction misconception, do not hide it under faster procedures.
If the learner understands the concept but fails under time pressure, do not give another conceptual lecture.
Precision saves time.
The rule of one-variable change
A good diagnostic question changes one important feature at a time.
If you change the numbers, wording, topic, representation and difficulty simultaneously, you will not know which change caused success or failure.
Useful controlled changes include:
- same relationship, simpler numbers;
- same numbers, simpler wording;
- same relationship, different representation;
- same method, different context;
- same concept, delayed retrieval;
- same skill, timed versus untimed.
Why a correct follow-up answer is not the end
Suppose the learner corrects a fraction misconception after a model is shown.
That is evidence of immediate learning.
It is not yet evidence of durable transfer.
Retest later:
- with different numbers;
- without the model supplied;
- inside a mixed set;
- after a delay;
- under examination wording.
A repair is stronger when it survives all five.
A diagnostic sequence from Primary 1 to Primary 6
Primary 1–2
- Can quantities be counted and compared reliably?
- Are part-whole relationships stable?
- Does the learner understand addition and subtraction meanings?
- Is place value understood rather than recited?
- Can simple word problems be represented?
Primary 3–4
- Are multiplication and division structures stable?
- Can fractions be treated as quantities?
- Can decimals be aligned by place value?
- Can multi-step problems be decomposed?
- Can units and geometry definitions be preserved?
Primary 5–6
- Can fraction, ratio and percentage representations be distinguished?
- Can reference quantities be tracked through change?
- Can rate and speed units be controlled?
- Can method selection survive mixed questions?
- Can the learner explain and check reasoning independently?
How this connects to Singapore Primary Mathematics
The current MOE Primary Mathematics syllabus organises content across Number and Algebra, Measurement and Geometry, and Statistics while emphasising mathematical processes such as reasoning, communication, representation, application and modelling.
A diagnostic decision tree does not replace the syllabus. It helps a teacher, tutor or parent determine which part of the learner’s mathematical process is preventing the syllabus knowledge from being used successfully.
For the 2026 PSLE, SEAB’s published assessment objectives include straightforward computation and procedures, interpretation and application in varied contexts, and mathematical reasoning with strategy selection. Those different demands make mechanism-level diagnosis especially useful.
What parents can do with one wrong question
Do not immediately ask the child to redo the whole worksheet.
- Ask what the child thought the question was asking.
- Ask what each number in the working represents.
- Ask for a simpler example of the same relationship.
- Change one feature and see whether the method survives.
- Stop once the likely cause is clear.
The purpose is information, not pressure.
What tutors should record
- the original error;
- the competing hypotheses;
- the discriminating question used;
- the learner’s response;
- the selected repair;
- the transfer test;
- whether the repair held later.
Over time, this turns isolated wrong answers into a useful learner map.
The deeper lesson: diagnosis is an information problem
A wrong answer creates uncertainty.
Several causes are possible.
The next question should reduce that uncertainty as efficiently as possible.
That is why a tiny question can sometimes be more useful than another full paper.
Good Mathematics diagnosis does not ask for more evidence indiscriminately. It asks for the next piece of evidence that will change the teaching decision.
Official sources and further reading
- Ministry of Education Singapore — Primary Mathematics Syllabus 2021, updated October 2025
- Singapore Examinations and Assessment Board — PSLE Mathematics 0008, for examination from 2026
- eduKateSG — Mathematics Diagnostic Conditions Master Index
- eduKateSG — How to Diagnose Learning Gaps From a Marked PSLE Mathematics Paper