A student finishes a question, looks at the answer once, and says:
“Checked.”
What exactly was checked?
The arithmetic?
The unit?
The percentage base?
The interpretation of the question?
The calculator entry?
Checking is reliable only when the learner knows what kind of error each check is designed to catch.
A good PSLE Mathematics checking routine is therefore not “read everything again”.
It is a layered system of targeted checks applied at the moments when different errors are most likely to occur.
The quick answer: check locally, then globally
A reliable routine has three levels.
- Local check: was this calculation executed correctly?
- Quantity check: does this number represent the right thing, with the right unit and base?
- Global check: does the whole solution answer the actual question and make sense in context?
Many students perform only the first level.
That is why a perfectly recalculated wrong model can survive.
Check 1: magnitude before exact arithmetic
Before calculating 38% of 1,980, estimate.
40% of 2,000 is about 800.
The exact result should be somewhat below 800.
If the final result is 75.24, the error can be rejected immediately.
This pre-calculation estimate becomes a reference point for the later answer.
Estimation is strongest when it happens before the exact answer is known, because it creates an independent expectation rather than merely agreeing with the result afterwards.
Check 2: inverse operation
If:
2,016÷4 = 504,
check:
504×4 = 2,016.
If:
3.75×5 = 18.75,
check:
18.75÷5 = 3.75.
Inverse checks are valuable because they use a different operation.
Repeating the same algorithm in the same way can repeat the same mistake.
Check 3: units
Area should end in square units.
Volume should end in cubic units.
Speed should carry distance per time.
Money should be expressed as a monetary quantity.
A result of 72 for a speed question is incomplete until it becomes, for example, 72 km/h.
Units also catch structural errors.
If a calculation intended to produce area ends with cm, the wrong relationship may have been used.
Check 4: percentage base
Percentage questions often fail because the arithmetic is correct on the wrong base quantity.
Example:
500 items.
20% removed.
Then 25% of the remainder removed.
First remainder:
400.
Second percentage:
25% of 400 = 100.
A quick base check asks:
“What quantity is 100% at this step?”
This catches one of the most persistent upper-primary errors.
Check 5: ratio reference
If red:blue = 2:3, then red is 2/5 of the total, not 2/3.
A ratio check asks:
- What are the two quantities being compared?
- Is the question part-to-part or part-to-whole?
- Does the simplified final ratio match the stated relationship?
If the final quantities are 24 and 36:
24:36 = 2:3.
The ratio relationship has been restored successfully.
Check 6: geometry labels
Before accepting a circle answer, ask:
- Was the given length radius or diameter?
- Was the question asking for circumference or area?
- Was the unit linear or square?
Before accepting a triangle area answer:
- Is the chosen height perpendicular to the base?
- Was the factor 1/2 included?
Before accepting a composite area answer:
- Were pieces overlapping?
- Was anything omitted?
- Does the result fit inside an obvious enclosing bound?
Check 7: boundary and region
Area and perimeter can produce numerically plausible answers from the same diagram.
Trace the requested object:
- outside edge → perimeter;
- inside coverage → area;
- filled 3D space → volume.
This is a model check rather than an arithmetic check.
Check 8: average definition
For average speed:
average speed = total distance ÷ total time.
Do not automatically average two speed values.
For ordinary mean:
mean = total value ÷ number of values.
A definition check is often stronger than recomputing the final division.
Check 9: reasonableness bounds
If a shape lies inside a 100 cm² rectangle, its area cannot exceed 100 cm².
If 40% of a positive quantity is taken, the result should be less than the whole.
If a divisor is less than one and positive, the quotient may be larger than the dividend.
If a 5 km journey is completed in 30 minutes, a result of 0.1 km/h is implausible.
Bounds do not need the exact answer.
They define where the answer is allowed to exist.
Check 10: calculator transcription
Paper 2 allows approved calculators.
This creates a new error class:
- wrong number entered;
- wrong bracket grouping;
- correct output copied incorrectly;
- display value reused after the quantity changed;
- mode or setting confusion.
After an important calculator result:
- read the display;
- write the number;
- look back at the display;
- compare digit by digit;
- attach the unit or quantity label.
This small routine catches errors that mathematical understanding alone cannot prevent.
Check 11: answer the question actually asked
A problem may require:
- difference, not total;
- remainder, not amount removed;
- time, not speed;
- larger quantity, not one ratio unit;
- area remaining, not area removed.
Underline or mentally restate the final requested quantity before boxing the answer.
This catches “stopped one step early” errors.
Check 12: all sub-parts completed
Structured questions can contain multiple parts.
A student can solve part (a), use it in part (b), then forget to record one of the requested final answers clearly.
During final review, scan question numbers and sub-part labels before redoing calculations.
Micro-checks versus final checks
Micro-checks happen during solving.
Examples:
- confirm a unit conversion before using it;
- verify one ratio unit;
- check a calculator entry;
- confirm that the percentage base has changed.
Final checks happen near submission.
Examples:
- unanswered parts;
- flagged questions;
- units;
- transcription;
- high-risk geometry labels;
- answers outside expected bounds.
Micro-checks stop errors from spreading. Final checks catch what survived.
A five-second local check
After a routine question:
- Is the magnitude sensible?
- Is the unit right?
- Did I answer the requested quantity?
This should become automatic enough not to derail pacing.
A thirty-second high-risk check
After a multi-step question:
- Identify the base or starting state.
- Trace each intermediate quantity.
- Check the final unit.
- Use an inverse or alternate method on one critical step.
- Compare with an estimate or bound.
Not every long question needs a full second solution.
Target the highest-risk transformations.
A final-paper checking order
When only a few minutes remain, prioritise:
- blank questions and missing sub-parts;
- questions previously flagged;
- transcription and calculator-copy errors;
- units and rounding;
- percentage-base and ratio-reference errors;
- geometry radius/diameter and area/perimeter errors;
- one or two answers whose magnitude looks suspicious.
This is more efficient than rereading every question from the beginning with equal attention.
Use error logs to personalise the checking routine
A generic checklist is a starting point.
A strong learner-specific routine is built from repeated evidence.
If a student repeatedly:
- uses diameter as radius;
- forgets units;
- applies percentages to original totals;
- copies calculator outputs incorrectly;
- stops at one ratio unit;
those should become explicit final-review priorities.
Checking is strongest when it is informed by the learner’s own error history.
Common misconception 1: checking means redo every calculation
Repeating the same method can repeat the same error.
Repair: use inverse operations, estimates, units and bounds.
Common misconception 2: checking is only for careless students
Complex reasoning creates multiple failure points even for strong students.
Checking is system reliability, not a judgement about intelligence.
Common misconception 3: exact calculator output needs no check
The output can be exact for the wrong input.
Common misconception 4: final review should start at Question 1
High-risk and incomplete questions deserve priority when time is limited.
Common misconception 5: one checklist fits every learner forever
The routine should change as recurring error patterns change.
A checking-readiness ladder
- Can the learner estimate before calculation?
- Can the learner use inverse operations?
- Can the learner check units?
- Can the learner identify percentage bases?
- Can the learner verify ratio references?
- Can the learner check geometry labels?
- Can the learner detect implausible bounds?
- Can the learner verify calculator transcription?
- Can the learner distinguish local arithmetic checks from global model checks?
- Can the learner perform targeted final review under time pressure?
How this fits the current PSLE format
The current 2026 PSLE Mathematics examination uses two papers with different calculator conditions. Paper 1 is calculator-free; Paper 2 allows an approved calculator. Both include item types where accurate execution and clear working matter, and Paper 2 structured/long-answer questions require methods to be shown clearly.
This checking routine is not an official SEAB checklist. It is a practical error-control framework aligned to the mathematical and examination demands of the current format.
The deeper lesson: checking is a second model of the answer
A calculation produces an answer through one route.
A good check asks whether another independent constraint agrees:
- inverse operation;
- unit;
- bound;
- estimate;
- original ratio;
- geometric property.
Checking becomes powerful when the answer has to survive more than one way of being true.