PRIMARY 4 MATHEMATICS LEARNING GUIDE · BATCH 7 · GUIDE 26
An error becomes useful when it tells us what to repair. Writing “careless” beside a page may describe frustration, but it does not identify whether the learner misread a unit, misunderstood a relationship, chose the wrong representation, executed a correct method inaccurately or interpreted a remainder incorrectly.
This error atlas organises recurring Primary 4 Mathematics mistakes by the first weak link that can plausibly generate them. It is not a list of labels to attach permanently to children. Every diagnosis should remain provisional until a changed task confirms it.
Series route: return to the Primary 4 Mathematics Learning Hub. Start with Diagnostic Assessment, Capability Profile and Repair Routing when the first weak area is not yet known.
Curriculum boundary: examples use Primary 4 content families referenced to the MOE Primary Mathematics Syllabus, updated October 2025. The error taxonomy and repair routines are independent teaching tools.
Navigate: first weak link · number errors · fraction/decimal errors · geometry/data errors · word problems · repair design · error laboratory.
1. Correct the first divergence, not every later symptom
A learner reads 4.08 as larger than 4.8, then subtracts the smaller from the larger and obtains a neat positive difference. The subtraction may be perfectly executed relative to the learner’s mistaken comparison.
If the first error was decimal place value, reteaching subtraction is not the first repair.
Find the earliest step at which the learner’s representation no longer matches the problem. Later steps often inherit that mismatch.
This is why full working matters. A final answer alone can hide whether the error began in reading, modelling, operation choice or calculation.
The repair should target the earliest unstable dependency that can explain the visible chain.
2. Use five broad error families
| Family | What went wrong | Typical evidence |
|---|---|---|
| Reading | Information or representation was misread | Graph scale, unit, condition or keyword interpreted incorrectly |
| Concept | Relationship itself is misunderstood | Additive and multiplicative comparison confused |
| Representation | Model does not match known relationships | Equal bars drawn without equality condition |
| Execution | Method is suitable but carried out inaccurately | Long multiplication place-value slip |
| Interpretation | Calculated value is not returned to the context correctly | Remainder or unit mishandled in final answer |
Some errors involve more than one family. Record the first supported one and confirm it with another task.
3. Error patterns are evidence, not character judgements
A repeated mistake can indicate a stable misconception. It does not justify a broad personal label.
“Adds denominators in three different fraction questions” is observable evidence. “Bad at fractions” is a generalisation that hides the exact relationship needing repair.
“Reads every graph interval as one” is actionable. “Does not pay attention” is not sufficiently specific.
Use language that points toward the next instructional move.
Precision protects both teaching quality and the learner’s sense of what can be changed.
4. Place-value error: digit size replaces digit position
Example: 40,706 is read as “forty thousand seven hundred six”, but the learner says the 7 has value 7 rather than 700.
First repair: place the number in a place-value chart and ask what each digit counts. Then rotate the 7 into a different position.
Do not begin with: ten more pages of addition. The error concerns representation of number, not addition.
Retest: ask for the value of 7 in 47,026 and then in 4,072.
Transfer is shown when the learner tracks place rather than memorising one number.
5. Rounding error: next digit used as a ritual without midpoint meaning
A learner rounds 47,362 to the nearest hundred as 47,300 because “the tens digit is 6 but the hundreds digit stays”. The remembered procedure has broken.
Repair: place 47,362 between 47,300 and 47,400 and identify midpoint 47,350.
Once the learner sees that 47,362 lies above the midpoint, 47,400 follows naturally.
Retest: use a number just below the midpoint, exactly at the midpoint under the school convention, and just above it.
The aim is to reconstruct the rule from distance, not to add another verbal chant.
6. Multiplication error: tens multiplier treated as ones
In 246 × 23, a learner writes the second partial product as 246 × 2 rather than 246 × 20.
Repair: write 23 as 20 + 3. Calculate 246 × 20 and 246 × 3 separately before returning to the compact written algorithm.
The shifted row is not a formatting tradition. It records a tens-place value.
Retest: compare 246 × 23 with 246 × 203. Ask what each partial product represents.
If the learner cannot name the row, the compact method is still too opaque.
7. Division error: placeholder zero disappears
A learner writes 1,248 ÷ 6 = 28.
Repair: estimate first: 1,200 ÷ 6 is about 200, so 28 is immediately too small. Then track hundreds, tens and ones explicitly. The quotient has 2 hundreds, 0 tens and 8 ones: 208.
Retest: use 1,212 ÷ 6 or another example requiring a zero in the quotient.
Check: multiply 208 × 6 = 1,248.
Estimation and inverse multiplication together protect the place-value structure.
8. Remainder error: arithmetic result is not interpreted
161 people need vehicles that each hold eight. A learner calculates 161 ÷ 8 = 20 remainder 1 and reports 20 vehicles.
Repair: ask where the remaining person goes. Twenty vehicles provide only 160 places.
The division is correct; the final interpretation is not. The minimum is 21 vehicles.
Retest contrast: 161 stickers packed in complete sets of eight asks for 20 full packs and one sticker left. Same arithmetic, different world return.
Do not reteach long division when the quotient and remainder were already accurate.
9. Factor/multiple error: relationship direction reverses
A learner says 24 is a factor of 6 because 24 is in “the six times table” reasoning somehow associated with multiplication.
Repair: test exact division. A factor must divide the number exactly. Six divides 24; therefore 6 is a factor of 24, and 24 is a multiple of 6.
Use the sentence pair: “6 fits into 24” and “24 can be built from groups of 6.”
Retest: ask whether 8 is a factor of 40 and whether 40 is a multiple of 8.
Direction language prevents the two labels from floating independently.
10. Fraction error: numerators and denominators added independently
A learner writes 1/3 + 1/4 = 2/7.
Repair: return to unit size. Thirds and quarters are not equal-sized parts. Convert to twelfths: 4/12 + 3/12 = 7/12.
Use a visual partition if the symbolic conversion is not yet meaningful.
Retest: 1/2 + 1/3, then 2/5 + 1/3.
The target concept is “equalise part size before counting parts together”, not simply “find LCM” as an isolated command.
11. Fraction error: reference whole changes unnoticed
A learner solves: “Use 1/3 of 72, then 1/4 of the remainder” by calculating both fractions of 72.
Repair: label states. First use = 24; remainder = 48. The second quarter refers to 48, not the original 72.
Write “1/4 of what?” beside the second fraction.
Retest contrast: compare “1/3 and 1/4 of the original amount” with “1/3, then 1/4 of what remains”.
The fraction notation can be identical while the reference whole changes.
12. Mixed-number error: regrouping is performed without equivalence
In 4 1/3 − 1 5/6, a learner changes 4 1/3 into 3 7/3 or another arbitrary form.
Repair: first express thirds as sixths: 4 2/6. Then exchange one whole for 6/6, giving 3 8/6.
Every representation must preserve the same total value.
Retest: 5 1/4 − 2 3/4.
Ask the learner to verify the regrouped mixed number before subtracting.
13. Decimal error: digits compared as whole-number strings
A learner says 3.18 is greater than 3.8 because 18 is greater than 8.
Repair: write 3.8 as 3.80. Compare tenths first: 3.80 has eight tenths; 3.18 has one tenth.
A place-value chart or number line can make the magnitude visible.
Retest: order 4.07, 4.7 and 4.17.
The error is not fixed by saying “add a zero” unless the learner understands why trailing zeros preserve decimal value.
14. Rounding error: repeated rounding changes the original value
A learner rounds 4.449 to two decimal places as 4.45, then rounds that to one decimal place as 4.5.
Direct rounding from 4.449 to one decimal place gives 4.4, because the hundredths digit is 4.
Repair: always identify the original value and the requested final place before rounding.
Retest: compare direct and repeated rounding on 2.349.
Intermediate approximations should not silently replace the original number when the question asks for direct rounding.
15. Area/perimeter error: formula chosen from the picture, not the question
A learner multiplies length by width when asked for perimeter.
Repair: name the measured attribute before the formula. Area covers a region and uses square units. Perimeter traces the boundary and uses length units.
Ask the learner to physically trace the border with a finger and shade the area.
Retest: give the same rectangle twice, asking for area once and perimeter once.
The goal is not to memorise “perimeter means plus” but to preserve what is being measured.
16. Composite-perimeter error: internal edges are counted
A learner adds the perimeters of two rectangles joined along an edge.
Repair: trace only the exterior boundary. The shared internal edge does not belong to the outside journey.
Use a coloured finger-trace or mark the exterior edges before adding.
Retest: use a different composite figure whose internal seam has another length.
Do not rely on a memorised subtraction of “twice the shared edge” until the learner can explain why it works.
17. Angle error: wrong protractor scale chosen
A visibly acute angle is reported as 132°.
Repair: predict acute/right/obtuse before measuring. Then identify the zero aligned with the starting ray and follow that scale.
The visual classification is a plausibility check, not a substitute for measurement.
Retest: present two angles with the same protractor position but opposite starting rays.
A learner should explain why the inner or outer scale applies.
18. Symmetry error: shape copied by appearance rather than equal distance
A learner completes a reflected figure by drawing what “looks balanced”.
Repair: mark each vertex’s perpendicular distance from the symmetry line and place the reflected vertex the same distance on the other side.
Retest: change the line from vertical to horizontal.
The invariant is equal perpendicular distance, not visual resemblance alone.
This gives a precise construction rule that survives different orientations.
19. Net error: face count replaces foldability
A learner accepts any connected arrangement of six squares as a cube net.
Repair: track folding, adjacency and overlap. Six faces are necessary but not sufficient.
Predict the opposite face before physically folding a paper model.
Retest: compare one valid and one invalid six-square arrangement.
The learner should identify the structural conflict, not simply memorise pictures of valid nets.
20. Graph error: lines counted instead of intervals
Labels 20 and 40 have four equal gaps between them. A learner divides by five because five grid lines are visible.
Repair: count spaces between the labelled values. Difference = 20; four equal intervals means five units per interval.
Retest: use labels 10 and 30 with four gaps, then another scale with three gaps.
Representation reading comes before extracting plotted values.
Do not assign more subtraction practice when the subtraction 40 − 20 was already correct.
21. Data error: evidence boundary is exceeded
A graph shows Class A collected more books than Class B. A learner concludes Class A “worked harder”.
Repair: separate measured quantity from inferred cause or character. The graph supports the count comparison; it does not directly measure effort.
Retest: ask for one statement definitely supported and one statement requiring additional evidence.
This is mathematical communication and evidence control, not merely graph reading.
Precision matters when a numerical display is used to justify a verbal conclusion.
22. Keyword error: operation selected from one word
A learner sees “received” and automatically adds, even when the question asks for the amount before the receipt.
Repair: label start, action and finish. If an unknown start plus 149 gives 580, the start is found by subtracting 149.
Words describe events; they do not always identify the direction of calculation.
Retest: use the same event with different unknowns: start unknown, change unknown, finish unknown.
Unknown rotation is stronger than another list of keyword-operation pairs.
23. Bar-model error: equality is invented
A learner divides a total of 90 into two equal bars even though the question only says two amounts total 90.
Repair: ask where the equality condition comes from. Without it, many pairs are possible.
Models are claims. Equal-looking units need evidence.
Retest: compare “two amounts total 90” with “two equal amounts total 90”.
The second has a unique answer of 45 each; the first does not.
24. Multi-step error: operation sequence begins before the needed quantity exists
A problem asks for equal shares after some items are removed. A learner divides the original total before subtracting the removed amount.
Repair: ask “What quantity is actually being shared?” Draw a dependency chain: original → remaining → per group.
Only calculate a step when its input quantity has been established.
Retest: rotate the order of the story sentences while keeping the chronological events unchanged.
This tests whether the learner follows dependencies rather than sentence order alone.
25. Unit error: correct number, wrong measurement
A learner finds rectangle perimeter 25 but writes 25 cm².
Repair: ask what was measured: boundary length or area. The correct unit is cm.
A numerical answer and its unit form one mathematical statement.
Retest: ask for area and perimeter of the same figure and require the unit before calculating.
Do not add units only after the arithmetic is finished.
26. “Careless” calculation error: confirm the concept before simplifying the diagnosis
A learner sets up 246 × 23 correctly, labels both partial products correctly, estimates around 5,000 and then adds 4,920 + 738 as 5,648 instead of 5,658.
This is evidence of an execution error after the structure was controlled.
It is reasonable to repair the addition accuracy directly.
But earn the label by inspecting the working first. Do not call a conceptual error “careless” simply because it happened quickly.
A narrow execution repair is appropriate only when the earlier layers are genuinely secure.
27. A correction task should change one thing
If the error was decimal comparison, keep the arithmetic simple and change decimal lengths.
If the error was remainder interpretation, keep the same division but change the context from leftover items to capacity.
If the error was graph scale, keep the graph reading simple and change the interval size.
Changing too many features at once makes it difficult to know which repair succeeded.
A good correction isolates the weak relationship, then gradually reintroduces complexity.
28. Require an error explanation, not a copied correct solution
A useful correction note has three parts:
- What did I do?
- Why does that not match the problem?
- What will I check next time?
Example: “I treated 3.8 as 3.08 because I compared the digits after the decimal as whole numbers. I should compare tenths first or write 3.8 as 3.80.”
This note is short but diagnostic.
Copying the teacher’s full answer may leave the original misconception untouched.
29. Retest after a delay and with a new surface
Immediate correction shows whether the learner can follow the repair while it is fresh.
Later retrieval shows whether the repair remains available independently.
Change the representation or unknown on the retest. A fraction error corrected only in the identical numbers has weak evidence of transfer.
Record whether a prompt was needed.
The aim is not perfection after one explanation; it is increasing stability across time and surface.
30. Error laboratory: identify the first weak link
- 47,362 rounds to 47,300 nearest hundred because “3 is the hundreds digit”.
- 1,248 ÷ 6 = 28.
- 161 people, 8 per van → 20 vans remainder 1.
- 1/3 + 1/4 = 2/7.
- 3.18 > 3.8 because 18 > 8.
- 4.449 → 4.45 → 4.5 to one decimal place.
- Rectangle perimeter 25 cm².
- Acute angle measured as 132°.
- Graph labels 20 and 40 with four gaps → interval size 4.
- “A has four more than B” drawn as four copies of B.
- Two amounts total 90 → 45 each, with no equality condition.
- After removing items, original total is divided before finding the remainder.
For each, write the error family, first repair question and one changed retest.
31. Batch 7 handover: repair should return to mixed transfer
An error atlas prevents broad remediation. Once the first weak link is repaired, the relationship must return to a mixed environment where the learner decides when to use it.
Continue to Primary 4 Mathematics Mixed Problem Laboratory, Interleaving and Transfer.
Final checkpoint: can the learner explain why the old route failed, solve a changed version and select the repaired relationship later without being told which chapter it belongs to?
Source and editorial note
The content families are referenced against the MOE Primary Mathematics Syllabus, updated October 2025. Error families, examples, repair questions and retest routines are independently written and are not represented as a validated clinical or psychometric taxonomy.
Editorial approach: Wintour House V1.0 · CivDJ · eduKate Publishing. Find the first divergence, repair the smallest relationship and retest under rotation.
