BTT Mathematics / Primary Mathematics Learning Hub / Working Mathematically

Primary Mathematics: Reflection, Self-Monitoring and Metacognitive Checking | Worked Learning Guide

Reflection is not asking “Was I right?” after the page is finished. It is noticing what you understand, what you are assuming, where you are stuck and what mathematical move would improve the next attempt.

A learner can know many methods and still use them without monitoring. A familiar-looking bar model is drawn before the relationship is read. A fraction answer feels wrong but the learner continues anyway. A long calculation produces an impossible negative number and is copied into the final line. A student knows they are stuck but cannot say whether the difficulty is vocabulary, representation, operation choice or arithmetic.

Self-monitoring makes those internal states visible enough to act on. It does not replace mathematical knowledge. It helps the learner decide when knowledge is missing, when a strategy is mismatched, when an answer deserves checking, and when a second approach is worth trying.

This guide organises reflection around four recurring questions: What is this problem asking? What do I already know that connects? What strategy fits and why? Does my solution make sense, and what should I change if it does not?

Before solving · During solving · When stuck · After solving · Learn from errors · 24 tasks · Worked guidance

1. Before solving: say what the problem is about

Question: A tank is 3/5 full. After 24 L is added, it becomes 4/5 full. Find capacity.

A useful first reflection is not “Which formula?” but “The 24 L increase represents the change from 3/5 to 4/5, which is 1/5 of the tank.”

That sentence reveals the structure before any arithmetic.

Comprehension check

• What is known?
• What is unknown?
• What does each number measure?
• Which condition links the known and unknown quantities?

2. Predict the shape of the answer

Before calculating 498×21, estimate about500×20=10,000.

This creates an expectation. If a later exact calculation gives1,045 or104,580, the learner has a reason to stop and inspect rather than accept the written digits automatically.

Prediction is part of monitoring because it gives the final answer something to be compared against.

3. Rate certainty about the relationship, not confidence about the final answer

A learner may feel “very confident” because a method is familiar. That feeling is not evidence of correctness.

A more useful self-rating separates:

• I understand what the question asks.
• I know why this method matches the relationship.
• I can perform the arithmetic.
• I know how I will check the result.

One of these can be uncertain while the others are secure.

4. During solving: monitor whether each step still belongs to the problem

Suppose a ratio problem has A:B=3:5 and total64.

The learner divides64 by8, obtaining8. A monitoring question is: “What does this8 represent?”

It is one equal ratio unit. Therefore A=3×8=24 and B=5×8=40.

If the learner cannot name the8, the method may be remembered procedurally without a stable model.

5. Use a checkpoint after a representation change

When 0.6 is rewritten as0.60 to compare with0.58, ask: “Did the value change?”

No. Six tenths equals sixty hundredths.

When1.35 m becomes135 cm, ask the same question. Representation and unit changed; physical length did not.

6. Ask whether the method is getting closer

A useful method should reduce uncertainty.

If random trial in a digit puzzle creates more cases and more duplicates, the learner should notice that the method is not narrowing the problem. Switching to an organised list may be more productive.

Persistence does not require staying loyal to an ineffective representation.

7. Track units through long calculations

A multi-stage journey has120 km at60 km/h, then90 km at45 km/h.

Stage times are2 h and2 h. If there is a30 min stop, total elapsed time is4.5 h.

A monitoring question after each stage is: “What unit is my current number?” This prevents adding distance to time or reporting a duration as a speed.

8. Compare the next step with the original condition

A class has53 pupils and vans hold8 each. 53÷8=6 remainder5.

If the learner writes “6 vans”, a context check catches the failure: six vans hold only48 pupils. The remainder represents five people still needing transport, so7 vans are required.

9. “I am stuck” is not yet a diagnosis

There are several different stuck states:

• I do not understand a word or condition.
• I know the quantities but not the relationship.
• I know the relationship but cannot choose a representation.
• I chose a method but cannot execute one arithmetic step.
• I can calculate but do not know whether the answer makes sense.

Naming the stuck state suggests a different next move.

Worked example: representation stuck

Problem: A is18 more than B and total74.

If the learner understands “18 more” but cannot form an equation, a bar model may be the right next move. If the learner cannot interpret “18 more” itself, drawing more bars will not fix the relationship.

10. Use a smallest-next-question routine

Instead of asking “How do I solve the whole problem?”, ask:

1. What can I determine immediately?
2. What one quantity would make the next step possible?
3. Can I create a smaller version?
4. Can I rename the quantities?

For a 6×8 rectangle-counting problem, a smaller 1×2 grid may reveal the counting structure before the large case is attacked.

11. Check whether you are repeating the same failed move

Trying 17 random values after 16 random values is not necessarily new information.

If repeated attempts fail for the same reason, change the method: organise cases, draw a model, use a bound, reverse the process or ask which condition has not yet been used.

12. Distinguish productive pause from avoidance

A useful pause has a question attached: “I need to decide whether order matters,” or “I need a common unit before adding.”

An unstructured pause does not change the mathematical state. Self-monitoring should lead to a next action.

13. After solving: verify more than the arithmetic

For A:B=3:5 total64, candidate A=24,B=40 should be checked against both original conditions:

• 24+40=64.
• 24:40 simplifies to3:5.

Checking one condition is incomplete when the problem supplied two.

14. Ask whether a different method agrees

Calculate25×36 by quarter-of-100 reasoning: 100×36÷4=900.

Check with written multiplication or36×(20+5)=720+180=900.

A structurally different method is useful because it is less likely to reproduce the exact same procedural error.

15. Separate answer quality from method quality

A correct answer can result from an invalid method by coincidence. An incorrect final answer can follow a correct model with one arithmetic slip.

Reflection should therefore ask two separate questions:

• Was the mathematical relationship and strategy valid?
• Was the execution accurate?

16. Compare two successful methods

For199+386:

• Written addition works.
• Compensation gives200+385=585.

Both are correct. Reflection asks which is more efficient here and whether the learner understands why compensation preserves the total.

The goal is not to shame the longer method. It is to build method choice.

17. Classify the first error, not only the final wrong answer

Suppose a learner answers 3/4+1/8=4/12.

The first error is not “bad addition”. It is treating denominator values as counts that can be added directly even though quarters and eighths are different-sized units.

Repair: convert3/4 to6/8, then6/8+1/8=7/8.

Classifying the error tells us what to teach next.

18. Build an error log that records causes, not embarrassment

Observed error	First weak link	Repair	Delayed check
0.7<0.65	decimal place value	rename0.7 as0.70	compare0.8 and0.79 later
6 vans for53 pupils at8 each	remainder interpretation	check capacity of6 vans	new capacity context
AB and BA both counted	ordered/unordered distinction	canonical pair order	new selection task

The purpose is to make the next learning action more precise.

19. High-confidence errors deserve attention

If a learner confidently states that 0.7<0.65 because “65 is bigger than7”, the misconception may be stable. A single correction may not be enough. Use a place-value representation, a number line and a delayed new comparison.

Confidence should be calibrated against explanation and repeated evidence.

20. Low-confidence correct answers also contain information

If a learner correctly says3/5 of45=27 but believes it was luck, the concept may not yet be reliably retrievable.

Ask for the reasoning: one fifth=9; three fifths=27. Then return later with a changed surface.

21. Reflection should produce a next move

After a practice set, do not write only “revise fractions”. Write something actionable:

• “Compare fractions with the same numerator using unit size.”
• “Check the percentage base before multiplying.”
• “Use a stage table for multi-step time problems.”
• “When order is irrelevant, prevent reversed duplicates.”

A useful reflection narrows future practice.

22. Use connection questions

After solving a percentage problem, ask: “What earlier idea does this resemble?”

25% of80 resembles1/4 of80, a fraction-of-quantity problem. Recognising the connection can create a simpler second method.

23. Use strategy questions

Before solving, ask: “Which of my available strategies fits and why?”

For an open-ended finite counting problem, an organised list may fit better than a bar model. For additive comparison, a bar model may fit better than random trial.

24. Use reflection questions

After solving:

• Does the solution make sense?
• Can it be solved another way?
• Where did I get stuck?
• What changed when I corrected it?
• What would I look for first in a similar problem tomorrow?

Reflection closes one problem while preparing the next.

25. Common self-monitoring errors

• Checking only whether the final answer matches an answer key.
• Using confidence as evidence of correctness.
• Calling every difficulty “careless”.
• Repeating the same method after evidence that it is not helping.
• Reflecting so generally that no next action follows.
• Changing several parts of the method at once and losing track of what fixed the problem.
• Overchecking simple secure work until time and attention are wasted.

26. A four-phase metacognitive routine

1. Comprehend: What is the problem really asking?
2. Connect: What do I already know that resembles this structure?
3. Select: Which strategy is appropriate, and what risk should I watch?
4. Reflect: Does the solution make sense, what evidence supports it, and what will I change next time?

27. Practice: 24 original self-monitoring tasks

Tasks 1–8: Before solving

1. For “3/5 of a tank is72 L”, identify known quantity, unknown and relationship before solving.
2. Estimate498×21 before exact calculation.
3. For A:B=3:5 total64, state what64÷8 represents.
4. For53 pupils in vans of8, predict whether the final number of vans can be fractional.
5. For0.6 versus0.58, name one representation that could clarify the comparison.
6. For a 2×3 rectangle-counting task, describe a smaller case that may reveal structure.
7. For a word problem containing five numbers, explain why using every number automatically is poor monitoring.
8. Write one comprehension question you could ask yourself before solving a percentage problem.

Tasks 9–16: During solving and stuck states

9. A learner has tried six random digit arrangements and keeps finding duplicates. Diagnose the stuck state and choose a new method.
10. A learner understands “18 more” but cannot form an equation. Suggest a representation move.
11. A learner writes120/60=2 in a speed problem. What monitoring question should be asked immediately?
12. A learner converts1.35 m to135 cm. What should be checked about the represented quantity?
13. A calculation produces−12 pupils. What should happen before the answer is accepted?
14. A learner cannot remember whether to multiply or divide in ratio. Write the smallest next question that could restore the model.
15. A long bar model has become unreadable in a four-stage process. What kind of strategy switch may help?
16. Give one example of a productive pause sentence and one unproductive “stuck” statement.

Tasks 17–24: After solving

17. Check A=24,B=40 against ratio3:5 and total64.
18. Verify25×36=900 using two different methods.
19. A learner gets the correct answer to x+17=45 after an invalid equality chain. What should reflection distinguish?
20. Classify the first weak link in 3/4+1/8=4/12.
21. Classify the first weak link in0.7<0.65 because65>7.
22. Write an actionable reflection after repeatedly using the wrong percentage base.
23. Write a connection question after solving25% of80.
24. Write a reflection question that would help decide whether a second method is worth trying.

28. Worked guidance

1. Known:3/5 equals72 L; unknown: whole tank capacity; relationship:72 represents three equal fifths. 2. About10,000. 3. One equal ratio unit. 4. No; van count is discrete and must be whole.

5. Rename0.6 as0.60 or use a number line. 6. A1×2 or2×2 grid. 7. Some numbers may be irrelevant, descriptive or used only in another part; the relationship determines relevance. 8. Example: “What quantity is the percentage of?”

9. Search-organisation failure; switch to ordered cases or a table. 10. Use a labelled bar model showing one common B-sized part plus an18 segment. 11. “What unit does2 have?” Answer hours. 12. Check that physical length is unchanged.

13. Stop and inspect model and constraints because a count of pupils cannot be negative. 14. Ask “How many equal ratio units are there altogether, and what quantity do they represent?” 15. A stage table or equation can reduce clutter. 16. Productive: “I know the total but not whether order matters; I will test that.” Unproductive: “I can’t do it.”

17. 24+40=64 and24:40=3:5. 18. Example quarter-of-100:25×36=900; written multiplication also gives900. 19. Separate correctness of the final candidate from validity of the method used to obtain it. 20. Fraction-unit misconception: unlike denominators name different-sized parts.

21. Decimal place-value comparison error. 22. Example: “Before multiplying by a percentage, label the original reference whole.” 23. “How is25% of80 the same as1/4 of80?” 24. “Would a second method reveal a different risk than the first one?”

29. Reflection laboratory: one error, four possible causes

A learner answers a ratio problem incorrectly. Do not begin with “revise ratio”. Test four possibilities:

1. Can the learner explain what3:5 means using equal units?
2. Can the learner identify that total64 represents8 units?
3. Can the learner calculate64÷8 accurately?
4. Can the learner rebuild3 units and5 units and check the total?

The first failed step is more useful than the final wrong answer. It determines whether the repair belongs to concept, representation, arithmetic or checking.

30. Parent and tutor guide

Ask the learner to describe where they became uncertain before providing help. If the answer is vague, offer categories: “Was the wording unclear, the relationship unclear, the method unclear, or the calculation difficult?”

Keep reflection brief enough that it supports mathematics rather than replacing it. One precise sentence about the first weak link is often more useful than a long diary entry.

After correction, use a changed question later. Immediate success can reflect short-term imitation; delayed success in a new surface gives stronger evidence that the monitoring routine has become usable.

31. Mastery receipt

• I identify the mathematical job before solving.
• I create an expected size or form for the answer when useful.
• I know what each intermediate number represents.
• I can name different kinds of “stuck”.
• I switch methods when evidence shows the current one is unproductive.
• I check original conditions, not only arithmetic.
• I classify first errors by cause.
• I turn reflection into a specific next action.

Sources and scope

NSW Department of Education — Metacognition: a key to unlocking learning discusses comprehension, connection, strategic and reflection questions in mathematical problem solving and reviews evidence on metacognitive teaching.

University of Cambridge NRICH — Integrating Rich Tasks: Activity 1.4 describes problem solving as a cycle rather than a single linear procedure.

For Singapore subject scope, use the MOE Primary Mathematics Syllabus P1–P6, updated October 2025.

Continue the capability batch

• Problem Posing, Questioning and Creating Variations
• Mathematical Modelling, Assumptions, Context and Validation
• Productive Struggle, Resourcefulness and Persistence
• BTT Primary Mathematics Learning Hub

The Quiet Return

Reflection earns its place when it changes the next mathematical decision. Know what you are doing, notice when it stops working, and make the next attempt more intelligent than the last.