Singapore School Mathematics Operating Manual · Chapter 35
A final answer is only as trustworthy as the chain that produced it.
Was a number printed in the question, measured from a diagram, calculated from an earlier part, assumed by the model, rounded from an exact value, copied from a calculator display, inferred from a graph, or introduced as a temporary variable?
Those origins matter.
This chapter develops mathematical provenance: the ability to trace a result backward through its sources, assumptions, transformations and checks. Result lineage is the dependency trail showing how one intermediate value became input to another. Traceability means a reader can audit that trail without guessing what happened.
The goal is not bureaucratic working. It is mathematical accountability.
1. Every important value should have an origin
If radius=5 cm, where did 5 come from?
Was it given directly? Was diameter 10 cm given? Was it calculated from area? Was it measured from a diagram?
The numerical value alone does not reveal its authority.
2. Provenance separates given facts from derived facts
Suppose a triangle has printed angles 50° and 60°.
The third angle 70° is derived from the angle sum.
All three may be used later, but their origins differ.
That difference matters when diagnosing an error.
3. A derived result inherits the assumptions used to derive it
If travel time is calculated using a constant-speed model, the result inherits that model assumption.
Later calculations using the time also inherit it.
An assumption can therefore travel downstream even if it is not repeated on every line.
4. Result lineage is a dependency chain
Suppose:
diameter → radius → area → cost.
If the radius is wrong, area and cost may both be wrong.
Lineage shows why early errors can affect many later conclusions.
5. High-lineage values deserve stronger checking
An intermediate quantity used once may have limited impact.
A quantity used in five later steps is high leverage.
Checking it early can prevent an entire branch of downstream error.
6. Labels improve provenance
“4” is weak.
“radius=4 cm” is stronger.
“radius=4 cm from diameter 8 cm” is stronger still when the route matters.
Good labels preserve meaning across handoffs.
7. Equation numbers or named results can support long solutions
In a complex derivation, label a key relationship as Equation (1).
Later write “using (1)” instead of reconstructing the source mentally.
This creates a visible audit trail.
8. A “hence” question explicitly reuses lineage
An earlier result becomes part of the evidence base for a later part.
The lineage is intentional and visible.
The Linked Parts and Result Handoffs chapter develops the examination structure behind this.
9. Temporary assumptions should be marked as temporary
Suppose a proof by contradiction begins “assume the opposite”.
Every line inside that branch depends on the temporary assumption.
When contradiction is reached, the branch is discharged.
The temporary premise must not survive into later reasoning as an established fact.
10. Case-specific results have local lineage
If x≥0, then √(x²)=x.
That result belongs to the non-negative case.
It cannot be copied into the x<0 branch where √(x²)=−x.
Lineage includes scope.
11. Rounded values should carry approximation lineage
If √2≈1.414, the decimal inherits approximation status.
A later result computed from 1.414 is also approximate unless an exact cancellation restores exactness by another route.
Do not silently turn approximate lineage back into exact equality.
12. Measurement lineage carries uncertainty
If a length was measured to nearest millimetre, every area or volume derived from it inherits measurement uncertainty.
Calculator precision does not erase the origin of the data.
13. Diagram-derived information needs explicit authority
A line looking perpendicular is not sufficient provenance.
A right-angle marker, statement or valid deduction is.
Visual appearance is a weak source unless the problem authorises measurement.
14. Calculator output has method lineage
If a solver obtains an angle using sin⁻¹, record whether degree or radian mode was intended.
The result’s validity depends on calculator state as well as arithmetic.
15. Graph-read values are approximate unless exact structure is given
If an intersection is read from a sketch, report it as an estimate unless the graph is intended to encode an exact labelled point.
The provenance source controls precision language.
16. Algebraic transformations should preserve lineage through equivalence
If x+5=9 becomes x=4 by subtracting 5, the new equation is equivalent.
The solution set lineage is preserved.
If both sides are squared, the transformed equation may have weaker lineage because extra candidates can appear.
17. One-way transformations should be marked mentally as candidate-generating
Squaring, cancelling variable factors and clearing denominators can change the candidate set.
The final verification must return to the original source equation.
This is a provenance reset: the original problem remains the final authority.
18. A proof should expose premise lineage
Each conclusion should follow from earlier accepted facts, definitions or theorems.
A proof that quietly assumes the target has circular lineage.
Backward planning may use the target privately, but forward justification must trace to premises.
19. Theorem use should include condition lineage
Pythagoras requires a right triangle.
The cosine rule applies more generally.
A formula’s provenance includes the conditions under which it is valid.
Using a theorem without its condition breaks the audit trail.
20. Units are provenance markers
A value 60 km/h carries rate lineage.
A value 60 km carries distance lineage.
Units tell the reader what kind of quantity the number came from and what operations are legal next.
21. Variable definitions create local provenance
Let n be the number of boxes.
Every later occurrence of n inherits that definition until scope changes.
Reusing n for a different quantity corrupts lineage.
22. Data cleaning creates provenance decisions
If an outlier is removed, record why.
The resulting mean is a statistic of the cleaned data, not the original data set.
Later interpretation should not forget the exclusion.
23. Model outputs inherit model choices
A fitted straight line inherits the assumption that linear summarisation is appropriate.
An exponential projection inherits its growth assumptions.
Downstream predictions should not be presented as assumption-free facts.
24. A provenance graph can be drawn for difficult work
Use arrows:
given dimensions → derived radius → area → tile count → cost.
Attach conditions such as “nearest cm”, “positive”, “whole number” or “assuming constant rate” to the relevant node.
The graph makes dependencies auditable.
25. Provenance improves error diagnosis
If cost is wrong, trace backward.
Was price wrong? Tile count? Area? Dimensions?
Instead of recomputing everything, inspect the lineage.
26. Provenance distinguishes source error from processing error
If a given value was copied incorrectly, the source entered the calculation wrongly.
If it was copied correctly but manipulated badly, the processing failed.
Those require different repairs.
27. Independent methods can create independent provenance
If area is calculated by two decompositions, the routes have different lineage.
Agreement is stronger evidence than repeating one route twice.
28. Cross-checks should not secretly depend on the same mistake
If both methods use the same wrong radius, agreement may be false reassurance.
Good verification seeks as much independence of lineage as practical.
29. Provenance matters in multi-source data
If one quantity comes from a table and another from a graph, record which is exact and which is approximate.
Different sources may have different precision, units or definitions.
30. A final answer should preserve necessary lineage without overwhelming the reader
Not every arithmetic step needs a narrative.
But critical assumptions, domain restrictions, approximations and high-leverage intermediate results should remain visible.
The goal is a solution that can be trusted and audited.
31. A provenance audit
Ask:
Where did each important value come from? Which assumptions does it depend on? Which transformations produced it? Is it exact, approximate, measured or modelled? What scope does it belong to? Which later results depend on it? Can I trace the final answer back to the original problem without a missing link?
32. Independent practice
1. A radius 5 cm is obtained from a stated diameter 10 cm. Describe its provenance.
2. Why should 1.414 derived from √2 carry approximation status?
3. A proof assumes x>0 inside one case. Can that result automatically be used in x<0?
4. Why is a right-angle marker stronger provenance than a diagram that merely looks right-angled?
5. A result depends on constant-speed modelling. What should later users of the result remember?
6. Why can two checks share a hidden common error?
7. What makes an intermediate value high leverage?
8. Why should candidates after squaring be checked in the original equation?
9. What does a unit contribute to provenance?
10. What is the purpose of a provenance audit?
33. Worked answers
1. It is a derived exact geometric value: radius=diameter/2=5 cm, based on the given diameter.
2. The decimal was produced by rounding an irrational exact value and is not exactly equal to √2.
3. No. The result has local case scope.
4. The marker is explicit mathematical information; visual appearance alone is not proof.
5. The downstream conclusion remains conditional on that modelling assumption.
6. If both methods depend on the same incorrect source value, they can agree while both are wrong.
7. Many downstream results depend on it.
8. Squaring can create candidates that were not solutions of the original problem.
9. It records quantity type and scale, helping show what the number represents and how it can be used.
10. To ensure the final conclusion can be traced back through valid sources, assumptions and transformations.
34. Continue through Batch 09
Use Observability and Identifiability to inspect whether the source information determines the hidden state, Compression and Sufficient Summaries to understand what information a representation retained, and Error Budgets and Cumulative Uncertainty to track approximate lineage quantitatively.
Return to the BTT Mathematics Hub for Batch 09.