Why G2 Mathematics Needs More Than Procedures | Algebraic Thinking, Mathematical Reasoning and SEC K210 Readiness

Why does G2 Mathematics need more than procedures? Because a procedure answers only one part of a mathematical problem: how to carry out a known operation once the relevant structure has already been recognised. G2 Mathematics also asks the learner to decide what the situation means, choose a representation, connect topics, formulate an equation, interpret a graph, justify a statement and decide whether the result makes sense. Those decisions cannot be replaced by faster execution of an irrelevant method.

Alicia can factorise a quadratic when the heading says “factorise” but may not see that a rectangle problem creates the same quadratic. Tricia can calculate a gradient but may not connect it to rate in a cost or distance model. Kai Kai can produce a correct answer with a calculator and struggle when asked to justify the mathematical statement. All three know procedures. Their next growth depends on mathematical thinking around those procedures.

This article owns the G2 Mathematics explanatory “why” job. The new How G2 Mathematics Works page remains the mechanism owner. The existing Secondary 4 G2 Mathematics K210 guide remains the specific final-year destination. This page explains why G2 Mathematics deliberately requires representation, connection, reasoning and communication in addition to routine proficiency.

For 2027 school candidates, SEAB lists G2 Mathematics as K210. The official assessment objectives allocate approximately 60% to standard techniques, 30% to solving problems in varied contexts and 10% to reasoning and communicating mathematically. AO2 explicitly includes identifying relevant concepts, translating information between forms, connecting topics, formulating problems mathematically, selecting techniques and interpreting results in context. AO3 explicitly includes justification, explanation and mathematical argument.

That balance explains the thesis of this article. Procedures matter enormously; they occupy the largest assessment share and provide the tools from which more complex work is built. But procedures become mathematically powerful only when the learner can select, combine, explain and check them. The problem is not “procedures versus thinking”. The problem is procedures without enough thinking to know when they apply.

Alicia, Tricia and Kai Kai are fictional learners. Every task, case, dialogue, error log and worked example below is original teaching material, not an official SEAB question, specimen-paper reproduction, grade boundary or private route-selection test.

1. What a procedure can and cannot do

A mathematical procedure is a reliable sequence for carrying out a particular operation. Long division, factorisation, solving a linear equation, calculating a mean and applying a trigonometric ratio are all useful because they compress tested relationships into repeatable actions.

Procedures reduce cognitive load

Once standard techniques are fluent, the learner does not need to reconstruct arithmetic or algebra from first principles every time. Attention can be spent on the larger problem.

A procedure assumes the problem has been classified

The steps for reverse percentage are useful after the learner has recognised that the given amount is a percentage of an unknown original. If the learner thinks the task is a forward discount, perfect execution gives the wrong answer.

Procedures do not determine relevance

Area and perimeter procedures can both be correct for a rectangle. The physical question decides which one matters.

Procedures do not interpret the output

A calculator can produce 3.4 buses. The learner must decide that four are required if capacity must be sufficient.

Procedures do not prove every claim

Testing three numbers can illustrate a pattern and cannot establish a universal theorem. General reasoning remains necessary.

The educational goal is usable fluency

Procedures should become sufficiently reliable that they support selection, modelling and reasoning. They should not become rituals detached from meaning.

2. Why method selection is mathematical knowledge

In a topic exercise, the page heading often tells the learner which method to use. In a mixed or real-world problem, that selection becomes part of the Mathematics.

Selection begins with relationships

Is the change additive or multiplicative? Is there a fixed plus variable component? Are two quantities proportional? Is there a right triangle? Is the sample space changing?

Keywords are insufficient

The word increase can describe a fixed increase, percentage increase, gradient or general trend. The surrounding quantities determine the structure.

Several methods can be valid

A break-even point can be found algebraically, graphically or through a table. The learner should choose a method suited to accuracy, efficiency and the information given.

Selection can be trained directly

Use short mixed prompts where the student names the likely representation without solving. This reveals whether the concept is available before execution begins.

Selection is a bridge to unfamiliarity

The learner does not need to have seen the exact story before if they can identify a familiar relationship beneath it.

3. Why representation changes the problem

A representation is not a decorative restatement. It can expose relationships that are hard to see in the original form.

A table reveals repeated change

Two pricing plans become easier to compare when several usage levels are aligned.

An equation compresses a general relationship

C=20+0.05x describes every usage value under the pricing rule, not one example.

A graph makes rate and threshold visible

Gradient, intercept and intersection turn algebra into spatial features.

A diagram reduces verbal load

Lengths, angles, directions and unknowns can be placed where they belong instead of held entirely in working memory.

A probability tree preserves sequence

Branching makes ordered paths and changing denominators explicit.

Representation should reduce the important uncertainty

The most sophisticated-looking form is not automatically best. Good mathematical judgement chooses the form that makes the decisive relationship easier to inspect.

4. Why topics must connect

K210 is organised into strands for clarity, while the official real-world guidance explicitly notes that problems may integrate ideas from more than one topic. Connections are therefore part of competence, not optional enrichment.

Percentage connects to algebra

An unknown original value turns a percentage relationship into an equation.

Algebra connects to graphs

Linear expressions become lines; roots become intercepts; simultaneous equations become intersections.

Geometry connects to algebra and coordinates

Lengths can generate quadratics; gradients can prove parallel relationships; scale factors interact with area and volume.

Statistics connects to percentage and argument

A percentage must retain its denominator and sample scope before it becomes a defensible claim.

Probability connects to representation

Lists, tables and trees make the sample space visible.

Connection is tested by transfer

If the learner knows two topics separately and cannot combine them when a problem requires both, the connection has not yet become usable.

5. Why reasoning must be visible

Reasoning is the relationship between claim and justification. It tells another person why the conclusion follows and gives the learner a way to inspect their own work.

Working can expose a hidden error

A final answer may be wrong while the model and first stages are correct. Visible working lets feedback target the actual step.

Justification protects geometry

“Angles are equal” becomes mathematically meaningful when the property is named and its conditions are satisfied.

Reasoning protects general claims

Examples can suggest a conjecture. Proof or a valid theorem establishes the general case; a counterexample can disprove it.

Context needs explanation

Rounding 6.2 vehicles to 7 should be justified by the capacity requirement rather than treated as ordinary nearest-integer rounding.

Communication is part of mathematical control

The learner who can state why a method fits is less likely to apply it mechanically in the wrong place.

Transfer bank: twenty cases where a correct procedure can still produce a weak mathematical answer

Case 1 · The wrong percentage base

A learner knows how to calculate 20% but uses the sale price rather than the original price when asked for percentage decrease.

Case 2 · The right rate with the wrong unit

Distance 18 km, time 30 minutes. Learner writes 18/30=0.6 km/h.

Case 3 · Direct proportion from a non-zero intercept

Cost C=5+2x is called directly proportional to x.

Case 4 · Reverse percentage treated forward

$96 is after a 20% discount. Learner calculates 96×1.20.

Case 5 · Compound growth treated additive

A value grows 5% for two years and the learner adds 10% once.

Case 6 · Algebraic cancellation across addition

(x+5)/x becomes 5.

Case 7 · Correct factorisation, forgotten restriction

(x²−9)/(x−3)=x+3 is reported for all x.

Case 8 · Solving before defining

A word problem contains two unknowns; the learner writes x and y without stating what they represent and later swaps their meanings.

Case 9 · Quadratic root without domain

Roots are 8 and −13 for a width problem; both are reported.

Case 10 · Inequality solved as equality

“At most 50” becomes x=50.

Case 11 · Gradient without meaning

Gradient 4 is calculated from a cost graph and left unitless.

Case 12 · Intercept outside context

A graph is extended to x=0 even though the real process begins at x=10.

Case 13 · Theorem used from appearance

A diagram looks isosceles and the learner assumes equal base angles.

Case 14 · Similarity ratio with wrong correspondence

Proportional arithmetic is correct but mismatched sides are compared.

Case 15 · Sine chosen because an angle appears

A right-triangle problem includes adjacent and hypotenuse, but the learner uses sine automatically.

Case 16 · Area scale uses linear factor

Length factor 4, area multiplied by 4.

Case 17 · Mean calculated correctly, chosen badly

Dataset has one extreme outlier; learner reports mean as the typical value.

Case 18 · Percentage from selected sample becomes population claim

90% of ten club members support a plan, reported as 90% of students.

Case 19 · Probability tree with frozen denominator

Without replacement, second denominator remains unchanged.

Case 20 · Correct total, wrong decision

A budget question asks whether a plan is affordable. Learner computes $742 and stops; budget is $700.

Transfer bank continued: twenty cases where reasoning changes the result

Case 21 · Estimate catches calculator entry

49.5×20.1 is entered as 49.5×201 and accepted.

Case 22 · Rounding changes a later answer

A trigonometric intermediate length is rounded to one decimal place before being squared.

Case 23 · Formula selected without checking units

Speed calculation uses km and seconds while answer is labelled km/h.

Case 24 · Same mean, different reliability

Two delivery services both average 30 minutes. One ranges 29–31; the other 10–50.

Case 25 · Weighted groups treated equally

A class of 10 averages 60 and a class of 40 averages 80; learner reports 70 overall.

Case 26 · Truncated axis overstatement

Bars at 98 and 100 on an axis from 97 to 101 are described as “twice as large”.

Case 27 · Correlation turns into cause

A scatter graph rises and the learner says x causes y.

Case 28 · Experimental probability becomes certainty

A spinner landed red 40% of 100 trials, so the learner predicts exactly 40 red in the next 100.

Case 29 · Repeated coin result creates gambler’s fallacy

Five heads in a row means tails “must” come next.

Case 30 · Examples mistaken for proof

2+4, 6+8 and 10+12 are even, so learner writes “therefore all even plus even is even”.

Case 31 · Counterexample not checked against condition

To disprove “all multiples of 4 are even”, learner gives 6.

Case 32 · Coordinate result without geometry

Equal gradients are found but the learner does not state parallelism.

Case 33 · Locus drawn without condition

A circle is drawn because the previous question used one.

Case 34 · Real-world model ignores threshold

A linear delivery price is extrapolated beyond a bulk-discount point stated in the problem.

Case 35 · “Best” has no criterion

One plan is cheapest, another fastest; learner declares the cheapest objectively best.

Case 36 · Insufficient information is treated as a challenge to invent

Usage rates are missing from two subscription plans, but the learner chooses one anyway.

Case 37 · Calculator solves but working does not reveal the model

The correct answer appears from a solver function with no equation shown.

Case 38 · Formula sheet reduces recall but not selection

A learner finds the area formula and applies it to a fencing problem.

Case 39 · Worked example causes false familiarity

A near-identical problem is solved immediately after the model; a week later a changed context fails.

Case 40 · One mark becomes identity

A low paper score is described as evidence the learner “cannot reason”.

The shared lesson

In every case, the procedure is either available or easily taught. The decisive weakness lies in choosing, interpreting, limiting or justifying it. This is the reason G2 Mathematics needs more than procedural fluency while still depending on procedural fluency.

Reasoning workshop: twenty-five integrated cases where the learner must explain the choice

Each case requires a calculation and a sentence explaining why the representation or conclusion fits. That second sentence is the point of the workshop.

Workshop 1 · Percentage plus fixed fee

A $400 service receives a 12% discount and then a fixed $18 administration fee. Find final cost.

Working: 400×0.88+18=$370. Reasoning: the percentage acts on the original service price; the fixed fee is added after the percentage change.

Workshop 2 · Reverse percentage plus tax

A pre-tax sale price after 20% discount is $240. Find original price, then add 9% tax to the sale price.

Answer: original=$300; taxed sale price=$261.60. The reverse percentage and tax use different bases.

Workshop 3 · Ratio and total

Boys:girls=4:5, total 72. Find counts.

Answer: 9 parts; one part=8; counts 32 and 40. The total determines the scale factor.

Workshop 4 · Direct proportion test

y values are 6,12,18 for x 2,4,6. Is direct proportion plausible?

Reasoning: y/x=3 throughout, so y=3x fits the data and passes through the origin under the model.

Workshop 5 · Inverse proportion test

Pairs (x,y)=(2,12),(3,8),(4,6). Explain the relationship.

Reasoning: xy=24 throughout, so y=24/x is an inverse-proportion model.

Workshop 6 · Algebraic identity

Show (x+4)²−(x−4)²=16x.

Working: expand or use difference of squares: [(x+4)−(x−4)][(x+4)+(x−4)]=8×2x=16x.

Reasoning: two valid representations lead to the same identity.

Workshop 7 · Simultaneous equations from a context

A shop sells 60 notebooks: standard $4, premium $7, total $303. Find each.

Model: s+p=60; 4s+7p=303. Solution p=21, s=39. The equations represent total count and total revenue.

Workshop 8 · Inequality as planning

A budget of $900 includes $180 fixed cost and $24 per participant. Find maximum whole participants.

Answer: 180+24p≤900 → p≤30. The inequality represents every affordable count, not only the boundary.

Workshop 9 · Quadratic root filter

A rectangle area 160, length 6 greater than width.

Model: w(w+6)=160 → w²+6w−160=0=(w+16)(w−10). Width 10, length 16. The negative root violates the physical domain.

Workshop 10 · Graph model comparison

A: y=10+2x; B: y=4x. Find intersection and explain.

Answer: 10+2x=4x → x=5, y=20. The intersection is the input at which both models have the same output.

Workshop 11 · Similarity and area

Corresponding lengths are 5 and 8; small area 75. Large area?

Answer: 75×(8/5)²=192. Area uses the square of the linear factor because two dimensions scale.

Workshop 12 · Similarity and volume

Linear scale small:large=2:3, small volume 240.

Answer: 240×(3/2)³=810. Volume scales in three dimensions.

Workshop 13 · Trigonometry from side roles

Right triangle, adjacent=12, hypotenuse=15, find angle.

Answer: cosθ=12/15=0.8, θ≈36.9°. Cosine is chosen because adjacent and hypotenuse are known.

Workshop 14 · Coordinate proof

A(0,0),B(4,1),C(8,2). What can be said?

Reasoning: gradient AB=1/4 and BC=1/4, so the three points are collinear.

Workshop 15 · Bounds

Length 6.8 cm nearest 0.1; width 4.2 nearest 0.1. Give lower area bound.

Answer: 6.75×4.15=28.0125 cm². Lower bounds combine for positive dimensions.

Workshop 16 · Median under an outlier

3,4,5,5,6,7,40. Which centre better represents the ordinary cluster?

Answer: median=5; mean=10. The median is less distorted by the extreme value, though the outlier remains important evidence.

Workshop 17 · Weighted mean

Group A 15 people mean 62; Group B 35 mean 74.

Answer: (930+2590)/50=70.4. Group size determines weight.

Workshop 18 · Percentage with non-response

60 of 150 invited respond; 42 support.

Answer: 70% of respondents support; response rate 40%. The views of 90 non-respondents are unknown.

Workshop 19 · Graph scale judgement

A chart starts at 49 and shows 50 versus 52. How should the difference be described?

Answer: numerically two units, or 4% relative to 50 if that comparison is relevant. Do not infer magnitude from the displayed bar ratio.

Workshop 20 · Probability without replacement

5 red, 3 blue; two draws. P(blue then red).

Answer: 3/8×5/7=15/56. The second denominator is seven because one item has been removed.

Workshop 21 · Complement

A fair die rolled three times. P(at least one 6).

Answer: 1−(5/6)³=91/216. Complement is efficient because “no six” is one simple event.

Workshop 22 · Correlation statement

A scatter graph shows higher study time associated with higher marks.

Reasoning: report positive association. Do not claim study time alone causes marks without stronger design.

Workshop 23 · Model domain

Temperature model T=90−4t. At t=30, T=−30. What should the learner ask?

Reasoning: whether the linear model remains physically meaningful that long. Real cooling may approach ambient temperature rather than continue linearly below it.

Workshop 24 · Insufficient criterion

Option A cheaper, B faster, C more reliable. Choose best.

Reasoning: no unique mathematical winner until priorities or a combined criterion are specified.

Workshop 25 · Full reasoning capstone

A club compares Venue A: $250+$6/person, Venue B: $430 flat, for 45 expected people. Venue B has area 60 m² and equipment occupies 12 m²; plan uses 1.1 m²/person. A prior survey of 24 finds 15 prefer Venue B’s location.

Working: A=$520, B=$430. Usable simple area=48 m², giving floor(48/1.1)=43 people, so B fails the simplified capacity requirement for 45. Survey preference=62.5% of a small prior sample. Reasoning: cheaper cost does not override insufficient capacity under the stated model; preference data do not repair the constraint.

Why explanations stay short

The objective is not to surround every calculation with an essay. One sentence identifying the relationship, condition, domain or interpretation often makes the difference between procedural output and mathematical communication.

Hidden-bottleneck clinics: thirty places where “practise the procedure again” is the wrong first response

Clinic 1 · Correct calculation, wrong denominator

The learner can calculate percentages but chooses the final amount as the base for percentage change. Repair: name original, change and final before arithmetic.

Clinic 2 · Formula known, quantity unknown

Speed formula is recalled, but the learner cannot identify whether average speed should include a stop. Repair: define the time interval the average is meant to describe.

Clinic 3 · Ratio method tied to recipes

Ratio succeeds with ingredients and fails with map scales. Repair: compare invariant multiplicative structure across contexts.

Clinic 4 · Percentage-point language is missing

A rate rises from 40% to 50% and is called a 10% increase. Repair: distinguish 10 percentage points from 25% relative increase.

Clinic 5 · Algebra begins from numbers rather than relationships

Every visible number is combined immediately. Repair: write what each quantity means and how they relate before operating.

Clinic 6 · Bracket errors cluster after subtraction

The learner knows expansion but loses signs after a leading minus. Repair: treat subtraction as addition of the opposite or multiplication by −1 across the bracket.

Clinic 7 · Factorisation succeeds and roots remain invisible

The learner treats factorisation as the end even when solving is required. Repair: connect zero-product property to roots.

Clinic 8 · Algebraic fraction simplification erases restrictions

Cancelled factors make excluded values disappear from the learner’s final statement. Repair: write original restrictions before simplification.

Clinic 9 · Equation solving is fluent but variable meanings drift

Repair: define variables with words and units and use them again in the conclusion.

Clinic 10 · Simultaneous equations are chosen because there are two numbers

Repair: identify two unknown quantities and two independent relationships, not merely “two things”.

Clinic 11 · Quadratic method selected without checking structure

A learner expands a simple factorised relationship into a harder quadratic unnecessarily. Repair: ask which form makes roots or constraints easiest to see.

Clinic 12 · Inequality answer lacks interval meaning

x<5 is obtained and reported as x=5. Repair: plot or test values to reinforce solution sets.

Clinic 13 · Graphs are pictures instead of models

The learner recognises “straight line” but cannot explain rate or fixed value. Repair: interpret gradient and intercept with units.

Clinic 14 · Graph intersection not connected to equality

Repair: write both functions equal at the intersection and verify the shared output.

Clinic 15 · Examples are mistaken for proof

Repair: represent the general integer or variable and derive the statement for all cases.

Clinic 16 · Counterexample fails the premise

Repair: check the proposed counterexample satisfies the original condition before using it to reject the conclusion.

Clinic 17 · Geometry follows visual familiarity

Repair: hide the picture orientation, mark stated conditions and require theorem-condition language.

Clinic 18 · Similarity arithmetic is correct with wrong correspondence

Repair: match angles and sides before writing any ratio.

Clinic 19 · Trigonometry uses the wrong reference angle

Repair: mark side roles relative to the requested angle each time.

Clinic 20 · Bearing measured from destination

Repair: draw north at the starting point and measure clockwise from there.

Clinic 21 · Unit conversion happens after multiplication

Repair: standardise units before combining quantities.

Clinic 22 · Bounds are treated as another rounding rule

Repair: interpret the rounded measurement as an interval of possible actual values.

Clinic 23 · Mean chosen by habit

Repair: ask which feature the decision needs: centre, typical value, spread or weighted total.

Clinic 24 · Range is treated as enough description of spread in every case

Repair: examine the full distribution or the syllabus-appropriate measure; understand what extremes do and do not reveal.

Clinic 25 · Graph scale ignored

Repair: describe actual numerical change before verbal magnitude.

Clinic 26 · Sample becomes population

Repair: write sample, population and selection method beside the percentage.

Clinic 27 · Probability multiplication without event meaning

Repair: label branches and paths before multiplying.

Clinic 28 · Independence assumed automatically

Repair: ask whether the first event physically changes the second probability or whether the model states independence.

Clinic 29 · Correct answer creates false mastery

The tutor supplied the representation and the learner performed the routine. Repair: record support and retest method selection independently.

Clinic 30 · Low score triggers more full papers

Repair: classify first failures and choose focused changed tasks before returning to paper-length integration.

How the clinics change teaching

The clinics move the conversation from “Which chapter is weak?” to “Which mathematical decision is not yet reliable?” The chapter still matters because content knowledge is necessary. The decision matters because it determines whether that content can be used outside the exact form in which it was taught.

Advanced cases: twenty-five examples where the explanation matters as much as the answer

Advanced Case 1 · A percentage that hides a changing base

A fee increases from $200 by 10%, then by another 10%.

Answer: $242, not $240. Explanation: the second increase is $22 because it acts on $220. Repeated percentage change is multiplicative.

Advanced Case 2 · A fixed fee hidden in a graph

A cost graph crosses the y-axis at 18 and rises $0.25 for each unit.

Model: C=18+0.25x. Explanation: the intercept is the cost at zero usage; the gradient is the unit rate.

Advanced Case 3 · Exchange-rate sequence

SGD 700 converts at 0.76 X/SGD; reconversion later is SGD 1.25/X.

Answer: 532 X then SGD 665. Explanation: reciprocal-looking rates at different times need not reverse the first conversion exactly.

Advanced Case 4 · Ratio after addition

Red:blue=3:4. Add 6 red and ratio becomes 9:8.

Model: (3k+6)/(4k)=9/8 → 24k+48=36k → k=4. Original 12 red,16 blue.

Explanation: one scale factor describes the original ratio; addition changes only one part.

Advanced Case 5 · Algebraic fraction with two restrictions

Simplify (x²−4)/(x²−x−6).

Working: (x−2)(x+2)/[(x−3)(x+2)]=(x−2)/(x−3), with x≠−2,3.

Explanation: cancelling a factor changes the written form, not the original domain.

Advanced Case 6 · Quadratic model from geometry

A rectangle has perimeter 34 and area 60.

Model: L+W=17, LW=60. Let W=x, L=17−x: x(17−x)=60 → x²−17x+60=0 → (x−5)(x−12)=0. Dimensions 5 and 12.

Explanation: both roots correspond to swapping which side is called length.

Advanced Case 7 · Inequality after a negative operation

Solve −3x+5>17.

Working: −3x>12, so x<−4.

Explanation: dividing by a negative reverses the inequality. Test x=−5 to check: 20>17.

Advanced Case 8 · Intersection as decision

Plan A=40+2x, Plan B=5x.

Break-even: 40+2x=5x → x=13⅓.

Explanation: if x is a whole number of uses, compare 13 and 14 to make the actual decision.

Advanced Case 9 · Proof through structure

Show the difference of squares of two consecutive integers is odd.

Proof: (n+1)²−n²=2n+1, which is odd for every integer n.

Explanation: the algebra covers all integers rather than selected examples.

Advanced Case 10 · Counterexample with condition

Claim: if ab>0 then a>0 and b>0.

Counterexample: a=−2,b=−3 gives ab=6>0 while neither is positive.

Explanation: the counterexample satisfies the premise and violates the conclusion.

Advanced Case 11 · Similarity under rotation

Triangle PQR is similar to XYZ. PQ=6 corresponds to XY=15; QR=10 corresponds to YZ.

Answer: scale factor 2.5, so YZ=25. Explanation: correspondence is determined by matching vertices, not orientation.

Advanced Case 12 · Area scaling and cost

A logo is enlarged by linear factor 1.8. Printing cost is proportional to inked area.

Answer: area/cost factor=1.8²=3.24.

Explanation: two dimensions scale, so a modest linear enlargement can more than triple area.

Advanced Case 13 · Trigonometric sanity check

A right triangle has hypotenuse 9 and calculated opposite side 11.

Conclusion: impossible. Explanation: the hypotenuse must be longest; check ratio selection or calculator entry before continuing.

Advanced Case 14 · Bearing from correct origin

B is on bearing 135° from A.

Representation: draw north at A and turn clockwise 135°. Explanation: bearings are measured at the starting point, not from page horizontal.

Advanced Case 15 · Coordinate geometry establishes a parallelogram

A(0,0),B(4,1),C(6,5),D(2,4).

Gradients: AB=1/4, DC=(5−4)/(6−2)=1/4; BC=(5−1)/(6−4)=2, AD=4/2=2. Opposite sides parallel, so ABCD is a parallelogram.

Advanced Case 16 · Bounds in area

L=10.2 m nearest 0.1, W=6.5 m nearest 0.1.

Intervals: 10.15≤L<10.25, 6.45≤W<6.55. Lower area=65.4675; upper<67.1375 m².

Explanation: a rounded measurement represents an interval, not one exact quantity.

Advanced Case 17 · Outlier and decision purpose

Travel times 18,19,19,20,20,21,55.

Mean:24.57; median:20. If planning typical journey time, median may be more representative; if planning worst-case buffer, the 55-minute event matters strongly.

Advanced Case 18 · Weighted performance

Morning shift produces 200 units at 4% defect; evening 800 at 1% defect.

Overall defects:8+8=16 of 1000=1.6%.

Explanation: averaging 4% and 1% to 2.5% ignores different production volumes.

Advanced Case 19 · Sample percentage and non-response

300 invited, 60 respond, 48 agree.

Finding:80% of respondents agree; response rate 20%. Limit: views of 240 non-respondents are unknown.

Advanced Case 20 · Graph and causal claim

A scatter graph shows exercise frequency associated with reported wellbeing.

Conclusion: positive association may be described. Limit: observational data do not establish exercise frequency as the sole cause; other variables and measurement quality matter.

Advanced Case 21 · Conditional probability structure

Bag 4 red,6 blue. Given first draw is red and not replaced, find P(second red).

Answer:3/9=1/3. Explanation: conditioning tells us the first event occurred and changed the sample space.

Advanced Case 22 · Complement and reliability

Independent failure probability 0.02 for each of 5 components. Find P(at least one failure).

Answer:1−0.98⁵≈0.0961. Explanation: complement avoids listing one, two, three, four and five failures separately.

Advanced Case 23 · Real-world budget and whole-number threshold

Budget $1200, fixed $285, $17.50 each. Maximum people?

Working:(1200−285)/17.5≈52.29, so maximum 52. Check:52 costs $1195; 53 costs $1212.50.

Advanced Case 24 · Piecewise pricing

First 100 units cost $0.20 each; extra units $0.12; fixed fee $15. Find cost for 180 units.

Answer:15+20+9.60=$44.60. Explanation: one rate cannot be applied across both usage bands.

Advanced Case 25 · Integrated decision with competing constraints

A training day has budget $2400. Venue X costs $500+$8/person and holds 80. Venue Y costs $950 flat and holds 120. Food $12/person. Expected 75 participants.

Cost X:500+600+900=$2000. Cost Y:950+900=$1850. Y is $150 cheaper and has enough capacity. If expected attendance rises above 80, X also fails capacity. Reasoning: both cost and capacity support Y under stated conditions, but availability/location could add constraints not provided.

What a reasoned answer sounds like

A reasoned answer does not need to be long. It might be “I used the median because one extreme value raises the mean,” “I rejected the negative root because x is a length,” or “I cannot generalise the percentage because the sample is self-selected.” These sentences expose the mathematical decision that the procedure alone cannot show.

Evidence room: thirty post-test conversations that identify the real mathematical job

Each conversation starts from a visible error. The tutor’s task is to identify whether the procedure itself is missing or whether the difficulty lies before or after it.

Conversation 1 · “I forgot how to do percentages.”

Alicia calculates 15% of any number accurately, but reverse percentage fails. The more precise diagnosis is not “forgot percentages”; it is “does not yet represent the final amount as a percentage of the original”. Practice should compare forward and reverse arrows.

Conversation 2 · “I cannot do rates.”

Tricia calculates distance/time correctly when both are in compatible units and fails only when minutes must become hours. The active target is unit conversion inside a rate, not the speed relationship itself.

Conversation 3 · “Ratio is weak.”

Kai Kai handles equivalent ratios and fails when one part changes after an addition. He needs algebraic scale-factor modelling, not another page simplifying ratios.

Conversation 4 · “Careless with standard form.”

The coefficient is often 32 or 0.42 rather than between 1 and 10. Teach the defining form A×10ⁿ and link decimal movement to exponent adjustment.

Conversation 5 · “Indices disappear under pressure.”

Positive integer laws are secure; negative indices become negative coefficients. The conceptual repair is reciprocal meaning, followed by mixed retrieval.

Conversation 6 · “Algebra is messy.”

Errors cluster only when a minus sign precedes a bracket. Isolate sign distribution, then return it to longer expressions.

Conversation 7 · “Factorisation is slow.”

The learner finds factors but does not recognise useful identities. Compare expansion and factorisation as inverse transformations rather than adding more random quadratics.

Conversation 8 · “Algebraic fractions are impossible.”

Multiplication/division works; addition with unlike denominators fails. Focus on common denominators and factor structure.

Conversation 9 · “Equations are fine, word problems are not.”

This is classic formulation evidence. Separate variable definition and equation construction from solving.

Conversation 10 · “Simultaneous equations are slow.”

Both methods are known, but the learner does not select elimination when coefficients already cancel. Teach method economy, not new algebra.

Conversation 11 · “Quadratics are wrong.”

Algebra is accurate; negative physical root is reported. The active target is domain interpretation.

Conversation 12 · “Inequalities are confusing.”

Linear inequalities work until dividing by a negative. Build number-line/test-value justification for sign reversal.

Conversation 13 · “Graphs are easy except word questions.”

Plotting is secure; gradient and intercept are not contextualised. Attach units and boundary-state meaning to graph features.

Conversation 14 · “I know the graph but not the equation.”

Use two points to derive gradient and intercept; connect the line to y=mx+c rather than treating equation and graph as separate topics.

Conversation 15 · “Geometry depends on the picture.”

Theorem recognition fails when diagrams rotate. Teach invariant relationships and varied orientation.

Conversation 16 · “Pythagoras is overused.”

The learner knows the formula too well and condition too weakly. Require right-angle evidence before substitution.

Conversation 17 · “Similarity ratios are inconsistent.”

Arithmetic is accurate; correspondence changes midway. Mark matching vertices and keep one orientation of ratios.

Conversation 18 · “Trigonometry formulas get mixed.”

The mnemonic is memorised; side roles are not. Hide the mnemonic temporarily and require labelled side relationships first.

Conversation 19 · “Bearings are off by ninety degrees.”

The learner measures from east or from the arrival point. Rebuild the convention: north at starting point, clockwise angle.

Conversation 20 · “Coordinates are a separate topic.”

The learner can calculate gradient and cannot use it to prove parallelism. Teach coordinate quantities as geometric evidence.

Conversation 21 · “Units are an afterthought.”

Correct numeric values are attached to wrong dimensions. Carry units through major operations and use them to check the formula.

Conversation 22 · “Bounds make no sense.”

The learner sees a rounded measurement as exact. Use physical intervals and a number line before deriving product bounds.

Conversation 23 · “Statistics is fine; interpretation loses marks.”

Mean, median and range are calculated accurately. The learner needs purpose-driven selection and comparison language.

Conversation 24 · “The graph was misleading.”

The learner notices a truncated axis only after being told. Teach an automatic scale audit before describing magnitude.

Conversation 25 · “Percentages in data are easy.”

The percentage is right and the denominator changes from respondents to all students in the conclusion. The target is sample scope.

Conversation 26 · “Probability trees are okay, but I forget cases.”

Paths are not labelled, so exactly-one events miss an order. Label events before probabilities.

Conversation 27 · “I know probability but at-least questions are long.”

Complement recognition is the efficiency target.

Conversation 28 · “I always run out of time.”

Observe where time goes. If method selection consumes time, faster arithmetic practice will not address the cause.

Conversation 29 · “I did it with my tutor yesterday.”

Check what cue the tutor supplied and whether the learner can reconstruct the representation after delay.

Conversation 30 · “My score proves I am bad at Maths.”

Replace the identity claim with a mechanism profile. A paper can show current performance under stated conditions; it cannot define mathematical potential.

Evidence-room rule

The tutor should be able to finish the review with one sentence beginning, “The next useful thing to practise is…” If the sentence names an entire subject or all three strands, the diagnosis is probably still too broad.

Twelve-week build: move from procedure ownership to mathematical independence

This sequence is an instructional framework, not an official MOE timeline, route rule or grade guarantee. It is designed to help teachers see whether procedural fluency is becoming transferable reasoning.

Weeks 1–2 · Separate procedure from selection

Use paired tasks. One explicitly names the method; the other hides the same structure in a context. Compare performance. If the first is strong and the second weak, the procedure is available while selection remains dependent.

Collect evidence from all three strands. One algebraic expression, one geometric relationship, one data/probability task and one mixed problem are enough to begin. The purpose is not a full diagnostic exam but a map of where cues are currently doing mathematical work.

Weeks 3–4 · Build representation choice

Practise turning the same relationship into words, tables, equations and graphs. Ask which form makes a particular question easiest. The learner should see representation as a tool, not a teacher preference.

For geometry, move from words to diagram and from coordinates back to shape. For probability, move from a physical sample space to a list or tree. Record which representations the learner creates independently.

Weeks 5–6 · Make conditions visible

Focus on theorem conditions, domain, denominator, sample and assumptions. Ask “When is this method allowed?” before “How do I perform it?”

Use near-miss examples deliberately: a straight line that is not direct proportion, a triangle that is not right-angled, a percentage with the wrong base, a population claim from a narrow sample. These examples teach boundaries.

Weeks 7–8 · Connect topics

Use problems requiring two mathematical ideas. A percentage becomes an algebraic equation. Geometry creates a quadratic. A graph checks simultaneous equations. A survey percentage needs statistical interpretation.

Ask the learner to state both components and the bridge between them. “I need percentage and algebra because the original amount is unknown.” This makes cross-topic connection explicit before it becomes automatic.

Weeks 9–10 · Add reasoning and realistic timing

Use short mixed sets under moderate time. Record whether time is spent understanding, selecting, executing, explaining or checking. Timing should target the actual bottleneck.

Require short justifications on selected items: why a root is rejected, why a statistic is chosen, why a theorem applies, why a claim is limited. Do not require an essay around every routine calculation.

Week 11 · Integrated decision case

Use one case with money, graph or geometry, data and a practical constraint. The learner should state assumptions and identify the result that controls the final decision.

Compare the mathematical answer with the operational answer. A venue can be cheapest and too small. A model can predict a negative quantity outside its domain. This week makes interpretation unavoidable.

Week 12 · Delayed unseen review

Return to the original mechanisms using different surfaces. Remove topic headings and most prompts. Compare first representation, procedure accuracy, reasoning, interpretation and checking with Week 1.

Retire mechanisms that transfer reliably. Keep only current targets. The diagnostic system should become lighter as the learner becomes more independent.

Three learner trajectories

Alicia: strong procedures, weak transfer. Her sequence emphasises changed-context formulation and representation. Tricia: strong conceptual choice, fragile execution. Her sequence emphasises smaller transformation steps and checking. Kai Kai: fast procedures, weak justification. His sequence emphasises conditions, domain and concise argument.

What counts as improvement

Improvement includes fewer prompts, faster recognition of structure, more stable working, better contextual closure and checks that catch errors before feedback. A score can reflect these changes and should not be the only evidence used to describe them.

Family and tutoring guide: keep the mathematics larger than the mark

For students

After each correction, write what changed in the decision: “I used the final value as the percentage base,” “I forgot the theorem condition,” or “I reported the sample as the population.” Avoid error records that say only “wrong Q7”.

For parents

Ask whether the student chose the method independently, whether the working made sense and whether the same issue repeats. A lower score can arise from a few high-cost mechanisms; a higher score can still hide dependence on familiar question forms.

For tutors

Model less once understanding is secure. Use a general prompt before a method cue. Collect an individual changed task after group discussion. Do not convert tutorial success into an official progression recommendation; formal subject arrangements remain with the school.

For school conversations

Bring representative scripts and describe mechanisms, support and change over time. “Needs a theorem cue in unfamiliar geometry” is more useful than “bad at geometry”. “Selects algebra independently but loses signs in long transformations” separates reasoning from execution.

Weekly dashboard

The dashboard is a teaching record, not an official K210 rubric.

Deep-transfer problems: thirty new surfaces for familiar mathematical structures

The first line after each problem names the invariant structure. Hide that line during independent practice; reveal it only during review.

1 · Parking charges

$6 entry plus $2.50/hour.

Invariant: fixed-plus-variable linear model C=6+2.5h.

2 · Mobile data

$12 base fee plus $0.03/MB over a free allowance.

Invariant: piecewise model with a threshold; one linear rule does not apply everywhere.

3 · Population multiplier

A population falls 4% annually.

Invariant: repeated multiplier 0.96ⁿ, not repeated subtraction of a fixed amount.

4 · Exchange percentage

An amount after a 6% fee is known.

Invariant: final=0.94×original when the fee is percentage-based.

5 · Recipe scale

Ingredients for 6 must serve 17.

Invariant: direct proportional scale factor 17/6 under the recipe model.

6 · Shared work

Time changes with number of identical workers.

Invariant: inverse-proportion model if total worker-hours remain constant.

7 · Taxi threshold

Two pricing rules cross.

Invariant: solve equality or graph intersection to locate break-even.

8 · Printing budget

Fixed setup plus per-page cost under a maximum budget.

Invariant: linear inequality and whole-number interpretation.

9 · Ticket mix

Total count and total revenue for two ticket prices.

Invariant: two simultaneous linear relationships.

10 · Rectangular field

Area given, length related to width.

Invariant: quadratic formulation and domain filter.

11 · Pattern tiles

Figure number and tile count follow a linear pattern.

Invariant: nth-term expression and graph.

12 · Profit graph

Profit rises linearly after a fixed initial loss.

Invariant: negative intercept plus positive gradient; zero crossing is break-even.

13 · Roof triangle

Right triangle hidden inside a roof cross-section.

Invariant: diagram construction followed by Pythagoras/trigonometry as conditions permit.

14 · Shadow similarity

Object and shadow triangles share angle relationships.

Invariant: similarity and proportional corresponding sides.

15 · Map enlargement

Linear scale changes and area is asked.

Invariant: square the linear factor.

16 · Tank model

Volume decreases at constant rate.

Invariant: linear model with physical domain ending at zero.

17 · Coordinate route

Two locations on a grid and a midpoint meeting place.

Invariant: coordinate midpoint plus unit conversion from grid to real distance.

18 · Parallel road segments

Coordinates describe two road lines.

Invariant: equal gradients as evidence of parallelism.

19 · Fencing versus turf

Same rectangle, two purchase decisions.

Invariant: perimeter versus area selected by physical job.

20 · Packaging scale

A model box enlarged.

Invariant: length, area and volume factors differ by powers of the scale factor.

21 · Waiting-time centre

One extreme delay among ordinary waits.

Invariant: compare mean and median before choosing “typical”.

22 · Delivery reliability

Two services have equal means and different ranges.

Invariant: centre alone does not describe consistency.

23 · Unequal class sizes

Two means need combining.

Invariant: weighted mean based on counts.

24 · Survey preference

Respondents are only a fraction of those invited.

Invariant: sample percentage plus response-rate and population-scope limits.

25 · Advertisement chart

Small numerical difference shown with cropped axis.

Invariant: numerical scale before visual magnitude.

26 · Bag draw

Two colours drawn without replacement.

Invariant: changing sample space and ordered paths.

27 · Quality-control events

Independent fault probability across several items.

Invariant: complement for “at least one fault”.

28 · Repeated trial frequency

Observed success rate used to forecast.

Invariant: experimental probability as estimate, not guaranteed exact future frequency.

29 · Product choice

Cheapest, lightest and longest-lasting options differ.

Invariant: no unique optimum until criteria are defined.

30 · Missing information

A comparison omits one variable charge.

Invariant: under-specification; identify missing data rather than invent it.

Use the invariant after the attempt

The invariant label is feedback, not a pre-question hint. If it is visible before the attempt, it performs part of AO2 selection. When the learner can generate a similar label independently, the underlying structure is becoming available across contexts.

“Why this step?” clinic: twenty-five explanations that make procedures meaningful

Why multiply by 1.08 for an 8% increase?

Because the final quantity contains the original 100% plus 8%, or 108%=1.08 of the original. The multiplier is a compact representation of the percentage relationship.

Why divide by 0.8 for an original price after 20% discount?

Because the final price is 80% of original. Division reverses the multiplication by 0.8; adding 20% to the final uses the wrong base.

Why is y=3x+4 not direct proportion?

Direct proportion requires y=kx, so y=0 when x=0. The +4 creates a fixed component and a non-zero intercept.

Why add exponents when multiplying like bases?

x²×x³ contains two factors of x followed by three more, giving five factors: x⁵.

Why can a factor cancel but not an added term?

Cancellation divides a common multiplicative factor from numerator and denominator. In x+3, x is not a factor of the entire sum.

Why does an equation stay balanced after the same operation on both sides?

Equality states the two expressions have equal value. Applying the same valid invertible operation preserves that equality under its conditions.

Why reverse an inequality when multiplying or dividing by a negative?

Negative multiplication reverses order on the number line. For example 2<5, but −2>−5.

Why reject a negative length root?

The algebraic equation allows it; the variable’s physical domain does not. Context restricts which mathematical solutions are admissible.

Why use simultaneous equations for some two-quantity problems?

Two unknown quantities require enough independent relationships to determine them uniquely. Count plus cost is a common pair.

Why can graph intersection solve two equations?

At the intersection, one coordinate pair satisfies both relationships simultaneously, so the y-values are equal for the same x.

Why does gradient need units?

Gradient is change in vertical quantity per change in horizontal quantity. Units identify the contextual rate.

Why does the y-intercept often represent fixed cost?

It is the y-value when x=0. In a cost model, that is cost before any units are used, if zero usage is inside the domain.

Why must Pythagoras wait for a right-angle condition?

The theorem describes the side relationship of right triangles. Without a right angle, the equation need not hold.

Why square a similarity scale factor for area?

Area combines two independent length dimensions. If each is multiplied by k, their product is multiplied by k².

Why cube the factor for volume?

Volume combines three length dimensions, so k×k×k=k³.

Why identify opposite and adjacent relative to the angle?

Those side names change when the reference angle changes. The trigonometric ratio is defined by the chosen angle.

Why measure bearings clockwise from north?

That is the convention defining three-figure bearings. Using page horizontal changes the quantity being measured.

Why use the median when one value is extreme?

The median depends on order, not the magnitude of every value, so one extreme observation moves it less than the mean. Whether that is desirable depends on the decision.

Why weight means by group size?

A group of forty contributes four times as many observations as a group of ten. Equal averaging would give each group equal influence despite unequal counts.

Why read a graph’s scale before describing magnitude?

Visual distances depend on axis range. Numerical differences are defined by scale, not by the fraction of the screen occupied.

Why cannot a sample percentage automatically describe a population?

The sample may be small or selected in a way that differs from the population. Generalisation depends on sampling evidence, not arithmetic alone.

Why does the denominator change without replacement?

The first draw physically removes an item, so fewer possible items remain for the second event.

Why use a complement for “at least one”?

The opposite event “none” often has fewer paths. Since at-least-one and none are exhaustive and mutually exclusive, subtracting from 1 is valid.

Why can one counterexample disprove a universal statement?

A universal claim says the conclusion holds for every case satisfying the premise. One genuine case where it fails is enough to show “every” is false.

Why is another full paper not always the next step?

If the same first failure repeats, a full paper hides it inside many unrelated demands. Focused repair followed by changed transfer can be more efficient, after which full papers test integration and timing again.

What the “why” answers reveal

A learner who can perform a procedure and explain the relationship has a more robust representation than one who remembers only the sequence. The explanation does not need to become a speech during every question; teaching uses it to verify understanding until the reasoning becomes internal and reliable.

Dialogue cases: twelve conversations that expose mathematical reasoning

These fictional dialogues are designed for small-group teaching. The useful moment is not who answers first. It is which relationship becomes visible and whether each learner can reconstruct it individually afterwards.

Dialogue 1 · Percentage base

Alicia: The price after discount is $80, so twenty per cent is $16 and the original is $96.

Tricia: But the discount was twenty per cent of the original, not of $80.

Kai Kai: Then $80 is eighty per cent. We should divide by 0.8.

The dialogue reveals the key issue: the base of a percentage. A teacher should not stop at the corrected $100. Ask Alicia to explain why her $96 uses the wrong reference amount, then give a different reverse-percentage context.

Dialogue 2 · Direct proportion

Kai Kai: It is a straight line, so it is direct proportion.

Alicia: It crosses the y-axis at 5.

Tricia: Direct proportion would have zero output when input is zero.

The group moves from a visual cue to a structural criterion. An individual exit task should use another straight line with a non-zero intercept.

Dialogue 3 · Forming equations

Tricia: I can solve simultaneous equations, but I don’t know what to write here.

Kai Kai: We know the total number of tickets and the total money.

Alicia: Then those are two different relationships about the same adult and student counts.

The discussion identifies formulation rather than algebraic manipulation as the bottleneck. The tutor can ask Tricia to define variables and form equations for a changed shop context without solving.

Dialogue 4 · Negative quadratic root

Kai Kai: The roots are 6 and minus 11, so the answers are 6 and minus 11.

Tricia: But x is the width.

Alicia: The algebra has two roots; the rectangle does not have a negative width.

This distinction between equation solution and contextual solution is central. The negative root is not “bad algebra”; it is outside the model domain.

Dialogue 5 · Gradient units

Alicia: I got gradient 2.4.

Tricia: Two point four what?

Kai Kai: The vertical axis is dollars and horizontal is kilometres, so $2.40 per kilometre.

Units convert a number into a contextual rate. This is a small habit with large checking value.

Dialogue 6 · Similarity under rotation

Tricia: I matched the left sides because both are on the left.

Alicia: One triangle is rotated. Which angles correspond?

Kai Kai: If we match the equal angles first, the side ratio becomes consistent.

Page orientation is replaced by invariant geometry. The next task should rotate the figures differently again.

Dialogue 7 · Trigonometry selection

Kai Kai: I always start with sine.

Tricia: We know adjacent and hypotenuse here.

Alicia: So cosine uses exactly the sides we have.

The mnemonic is not rejected; it is moved after side identification. This reduces formula roulette.

Dialogue 8 · Mean versus median

Alicia: The average wait is fourteen minutes.

Tricia: Most waits are around six or seven; one is forty-five.

Kai Kai: The mean is fourteen, but the median may describe the ordinary wait better.

The lesson is not “median is better”. It is that the purpose and distribution choose the useful summary.

Dialogue 9 · Survey denominator

Kai Kai: Eighty per cent of students agree.

Alicia: Eighty per cent of the twenty who responded.

Tricia: If two hundred were invited, the other one hundred eighty views are unknown.

The mathematical calculation and the claim are separated. This is statistical reasoning, not merely language editing.

Dialogue 10 · Probability without replacement

Tricia: Why did the denominator become nine?

Alicia: Because one token is already out of the bag.

Kai Kai: So the fraction changes because the physical sample space changes, not because the formula says so.

This physical explanation helps the tree diagram retain meaning.

Dialogue 11 · Counterexample

Alicia: To disprove “all multiples of three are odd”, I’ll use 4.

Tricia: Four isn’t a multiple of three.

Kai Kai: Use 6. It satisfies the premise and violates oddness.

A counterexample must enter through the statement’s front door: satisfy the condition, then break the conclusion.

Dialogue 12 · Real-world model

Kai Kai: Venue B is cheaper, so choose B.

Tricia: It only holds forty people and we have forty-six.

Alicia: Then the cheapest feasible option matters, not the cheapest number in isolation.

The conversation shows interpretation as a mathematical layer. A cost result can be correct and irrelevant if a capacity constraint is violated.

How to use dialogue without hiding individual evidence

After the discussion, give each learner a different short problem with the same reasoning demand. The group may discover the relationship collaboratively; the individual exit task shows what each person can now reconstruct. Rotate who explains, challenges and checks so that one confident speaker does not permanently perform the reasoning for everybody else.

Practical reasoning set: twenty-five questions where the result must be interpreted

1 · Ticket packages

A venue sells tickets in packs of 12 for $90. A group needs 50 tickets. Four packs give 48, so five packs are required, cost $450. Interpretation: package constraints override ordinary rounding.

2 · Data plan

Plan A $25+0.02x MB; Plan B $0.05x. Break-even 25=0.03x → x≈833.3 MB. Interpretation: compare whole or measured usage around the threshold rather than reporting a universal winner.

3 · Currency fee

Exchange SGD 500 at 0.74 X after a $5 fee: 495×0.74=366.3 X. Interpretation: a fixed fee changes the effective rate more for small transactions than large ones.

4 · Successive percentage

Value 300 rises 20%, falls 10%: 300×1.2×0.9=324. Interpretation: final is 8% above original, not 10%.

5 · Reverse tax

A total including 9% tax is $109. Original=109/1.09=$100. Interpretation: $109 is 109% of the untaxed amount.

6 · Constant-work model

8 workers ×6 hours=48 worker-hours. Under ideal inverse model, 12 workers need 4 hours. Limit: coordination and productivity may make real work non-inverse.

7 · Budget inequality

$750 budget, $210 fixed, $22 each → 210+22n≤750 → n≤24.54, so 24 whole participants. Check: 25 would exceed budget.

8 · Quadratic dimensions

Area 192, length 4 greater than width → w(w+4)=192 → w=12, length 16. Interpretation: reject negative algebraic root.

9 · Linear graph domain

C=60−5t models remaining credit. Zero at t=12. Interpretation: negative credit after 12 may not belong to the intended model if the service stops at zero.

10 · Piecewise fee

First 10 hours $8/h, later hours $5/h. For 16 hours: $80+$30=$110. Interpretation: one average rate cannot be applied before understanding the pricing bands.

11 · Similarity area

Linear scale factor 1.5, original area 80 → new area 180. Interpretation: area factor is 2.25.

12 · Volume scale

Linear factor 0.8, original volume 1000 → new volume 512. Interpretation: a 20% reduction in each length reduces volume by 48.8%, showing dimensional compounding.

13 · Ramp decision

Rise 1 m, run 6 m → angle arctan(1/6)≈9.5°. Interpretation: whether this is suitable requires an external design criterion not supplied by trigonometry alone.

14 · Coordinate meeting point

A(−2,4), B(8,10), midpoint=(3,7). Interpretation: midpoint is geometric halfway, not necessarily equal travel time if routes/speeds differ.

15 · Measurement bound

5.0 cm nearest 0.1 means 4.95≤L<5.05. Interpretation: the written measurement represents a range of possible actual lengths.

16 · Mean and median

2,3,4,4,5,6,25 → mean=7, median=4. Interpretation: the 25 dominates mean; purpose determines useful centre.

17 · Weighted mean

25 items average 8,75 average 12 → overall (200+900)/100=11. Interpretation: the larger group dominates the combined mean.

18 · Response rate

1000 invited,100 respond,70 agree. Interpretation: 70% of respondents, 10% response rate. Population support remains uncertain.

19 · Cropped graph

Values 49 and 51 on axis 48–52 look dramatically separated. Interpretation: actual difference is 2 units; visual height should not be converted into a claim of doubling.

20 · Probability path

3 red,7 blue; two without replacement. P(two red)=3/10×2/9=1/15. Interpretation: the second probability reflects the reduced sample space.

21 · Complement

Independent success probability 0.1 over 4 trials. P(at least one)=1−0.9⁴≈0.3439. Interpretation: 34.39% is a modelled probability, not a guaranteed frequency.

22 · Correlation

A graph shows higher revision time associated with higher marks. Interpretation: positive association; causation requires stronger evidence and control of other variables.

23 · Counterexample

Claim “if n² is even, n is odd.” n=2 gives n²=4 even while n is not odd. Interpretation: the universal claim is disproved.

24 · General proof

Even + even: 2a+2b=2(a+b), hence even. Interpretation: algebra covers all integer choices of a,b rather than selected examples.

25 · Multi-criterion decision

Route A costs $8 and takes 50 min; B costs $12 and takes 35 min. Interpretation: Mathematics provides the trade-off. A “best” route requires the traveller’s value for time, budget or reliability.

What makes these reasoning questions

Most calculations above are routine. The deeper demand lies in deciding how to model the situation and what the resulting number is entitled to say. That is why a learner can be procedurally strong and still benefit from explicit reasoning practice.

Twenty questions a G2 Mathematics learner should ask about their own working

1. What relationship did I identify before calculating?

If the answer is only “I saw the formula”, revisit the quantities and conditions.

2. Why is this the right denominator?

Percentage, rate, gradient and probability all depend on a reference quantity with meaning.

3. What does my variable represent?

Name the quantity and unit or count. Stable meaning protects the whole algebraic model.

4. Which information did I leave out, and why?

Irrelevant information should be omitted for a reason, not forgotten accidentally.

5. What assumption did I make?

Constant rate, independence, linearity or proportionality should be recognised as model assumptions where applicable.

6. What condition allows this theorem?

Name the right angle, parallel lines, similarity criterion or other requirement.

7. Why did I choose this representation?

A table, graph, equation or diagram should make a particular structure easier to inspect.

8. Is there another valid representation?

A second representation can provide insight or an independent check.

9. Does my answer belong to the domain?

Check signs, whole-number requirements, feasible ranges and physical meaning.

10. Are my units compatible?

Unit mismatch can reveal a hidden modelling or conversion error.

11. Is my accuracy appropriate?

Preserve precision during working and follow the stated final accuracy.

12. Did I prove the claim or only test examples?

Universal statements need general reasoning; examples support conjecture.

13. If I used a counterexample, does it satisfy the premise?

An invalid counterexample proves nothing about the claim.

14. Did the graph show cause or only association?

Study design, not visual slope, determines causal inference.

15. Who is represented in this percentage?

Name sample and denominator before generalising.

16. Has the probability sample space changed?

Replacement, conditioning and dependence can alter later probabilities.

17. Could a boundary value expose an error?

Try x=0, a threshold, or an extreme allowed value where it clarifies the model.

18. Does the magnitude make sense?

Use estimation and contextual scale to catch calculator errors.

19. What independent check can I use?

Inverse, substitution, estimate, units, graph or alternative method.

20. Could I recognise this structure in a different story?

If not, the procedure may still be attached to the original surface.

A student’s error log that leads to action

The error log is useful when it predicts what the learner should do differently next time. “Wrong Q12” does not.

Mathematical argument clinics: twenty-five ways to make a conclusion deserve belief

These clinics focus on the small piece of reasoning that turns a calculation into an argument. They are original teaching material, not official K210 questions. In each case, the learner should identify what is being claimed, what mathematical fact supports it and what condition limits the conclusion.

Argument 1 · Why the cheaper unit price is actually cheaper

Pack A costs $9 for 4 items; Pack B costs $13 for 6. A=$2.25/item; B≈$2.17/item. The conclusion “B is cheaper per item” follows because both totals have been converted to the same comparison basis. It does not prove B is the best purchase if the buyer needs only four items.

Argument 2 · Why equal percentage changes do not cancel

100 rises 20% to 120 and falls 20% to 96. The argument is not merely that 96 appears on the calculator; it is that the two 20% changes use different bases. The second reduction is 24, not 20.

Argument 3 · Why a direct-proportion graph passes through the origin

If y=kx, then x=0 gives y=0. A non-zero intercept represents an additional fixed component and therefore contradicts direct proportion under the model.

Argument 4 · Why reverse percentage divides

If a discounted value is 85% of the original, then final=0.85×original. Solving for original requires division by 0.85. The operation follows the relationship rather than a special reverse-percentage trick.

Argument 5 · Why average speed is not always the average of two speeds

Average speed is total distance divided by total time. If equal distances are travelled at two speeds, the times differ, so the two speeds do not receive equal time weight. A simple arithmetic mean is generally unjustified.

Argument 6 · Why a cancelled factor can leave a restriction

In (x²−4)/(x−2), the factor x−2 cancels after factorisation, but the original expression was undefined at x=2. Simplification preserves the same values where the original expression existed; it does not create a new valid input.

Argument 7 · Why simultaneous equations can determine two unknowns

Two independent linear relationships constrain the same two unknown quantities. Their common solution is the pair satisfying both simultaneously. If the equations are dependent, a unique solution may not exist; independence of the relationships matters.

Argument 8 · Why a negative root can be rejected

The rejection is not because negative numbers are undesirable. It is because the variable represents a physical quantity whose domain excludes negative values. A negative coordinate or financial balance could be valid in another model.

Argument 9 · Why an inequality solution is a set

x<5 describes every number below five, not a single boundary value. Substituting one allowed and one disallowed test value can make the set interpretation visible.

Argument 10 · Why gradient is a rate

Gradient compares vertical change with horizontal change. If vertical units are dollars and horizontal units are kilometres, the gradient is dollars per kilometre. The ratio of changes gives its contextual meaning.

Argument 11 · Why an intercept may be a fixed fee

For C=mx+c, c is the value of C when x=0. In a pricing model, if zero usage is meaningful, this can represent a fee charged before variable usage begins.

Argument 12 · Why Pythagoras proves a right-angle relation only under its condition

For a right triangle, the square of the hypotenuse equals the sum of squares of the other sides. Using the formula without establishing a right angle assumes the very condition the theorem requires.

Argument 13 · Why the converse can identify a right triangle

If the longest side c satisfies a²+b²=c², then the converse of Pythagoras establishes that the angle opposite c is a right angle. The relationship is now evidence for the condition.

Argument 14 · Why corresponding sides matter in similarity

A valid scale factor must compare parts occupying the same role in similar shapes. Matching by page position can pair unrelated sides and produce a numerically consistent but geometrically meaningless ratio.

Argument 15 · Why area scale is squared

If every length is multiplied by k, a rectangle with dimensions a and b becomes ka by kb. Area changes from ab to k²ab. The square arises from two scaled dimensions.

Argument 16 · Why volume scale is cubed

Three independent length dimensions scale by k, so volume is multiplied by k³. This explains why modest linear enlargement can greatly change capacity.

Argument 17 · Why trigonometric ratios depend on the chosen angle

Opposite and adjacent are defined relative to the reference angle. The same physical side can change role when a different acute angle is selected. Ratio choice must therefore follow the angle being solved.

Argument 18 · Why a median can be more representative

A median is determined by ordered position and is less affected by the magnitude of an extreme value. If the purpose is to describe the centre of a skewed cluster, this can make it more informative than the mean. The argument depends on purpose, not a universal rule that median is better.

Argument 19 · Why weighted mean respects group size

Each observation should contribute equally to the combined mean. Multiplying group means by their group sizes reconstructs the total contribution before division by the full count.

Argument 20 · Why a truncated axis can mislead without being false

The plotted values may be accurate while the chosen range makes a small difference occupy a large visual distance. The reader should judge magnitude from the scale and values rather than the apparent bar-height ratio.

Argument 21 · Why a sample percentage cannot automatically become a population percentage

The arithmetic describes the observed sample. Generalisation requires evidence that the sampling process makes the sample informative about the wider population. A selected club or low-response survey may not do so.

Argument 22 · Why association is not causation

A relationship between two variables does not by itself rule out confounding, reverse direction or common causes. Causal claims depend on design and evidence beyond a visual trend.

Argument 23 · Why the denominator changes without replacement

After one item is removed, the number of available outcomes decreases and the composition may change. The second probability describes a new physical state.

Argument 24 · Why complement works for “at least one”

“At least one” and “none” are mutually exclusive and together cover all possibilities. Therefore P(at least one)=1−P(none). The method follows event logic, not a memorised shortcut.

Argument 25 · Why the cheapest numerical option may not be the best feasible option

A venue can be cheapest and too small. A package can have the lowest unit cost and create unnecessary waste. Mathematics first tests constraints, then compares among feasible choices. The final judgement must state its criterion.

How to practise argument without slowing every question

Select one or two questions per lesson for full explanation. Ask the learner to name the decisive property in one sentence. The purpose is to make reasoning explicit while it is being learned, then allow it to become increasingly internal where the task does not require written justification.

Procedure or reasoning? Twenty-five mixed cases for G2 Mathematics

Each case can produce a wrong answer for more than one reason. The learner should decide whether the main repair is procedural fluency, representation, reasoning, interpretation or checking.

Case 1 · 18% of 450

The learner writes 450÷18. This is primarily a representation error: 18% should be represented as 0.18 or 18/100 before multiplication. Once the relationship is correct, the arithmetic is routine.

Case 2 · $135 after 10% discount

The learner finds 10% of 135 and adds it. This is a base/reasoning error: 135 is 90% of the original. Use 135=0.9P.

Case 3 · 3:5 ratio with total 64

The learner subtracts 5−3=2 and tries to distribute the total through the difference. This is a conceptual representation error. Total parts=8; one part=8.

Case 4 · 90 km in 75 minutes

The learner writes 90/75=1.2 km/h. The rate operation is correct but the unit reasoning is wrong. 75 minutes=1.25 h, giving 72 km/h.

Case 5 · x²x³=x⁶

The learner multiplies exponents. This is a procedure/concept error. Repeated multiplication shows that exponents add for multiplication of like bases: x⁵.

Case 6 · (x+4)/x cancels x

This is a structural reasoning error. x is not a common factor of the entire numerator; x+4 is a sum.

Case 7 · Solve 2(x−3)=14

The learner obtains x=4 after writing 2x−3=14. The main issue is distribution procedure. The bracket must be expanded as 2x−6.

Case 8 · Form equation from “five less than triple a number is 22”

The learner writes 5−3x=22. This is language-to-algebra representation, not equation-solving weakness. Correct form 3x−5=22.

Case 9 · Two ticket types, total count and revenue

The learner can solve simultaneous equations after they are given but does not form them. The active target is formulation.

Case 10 · x²−5x=0

The learner divides both sides by x and reports x=5, losing x=0. This is a structural procedure error. Factor x(x−5)=0 and use the zero-product property.

Case 11 · x²=16

The learner gives x=4 only. The active issue is solution completeness: x=±4 unless the domain restricts it.

Case 12 · x²=16 in a length problem

The learner gives ±4. The algebra is complete, while the context interpretation should retain only 4 cm.

Case 13 · y=2x+7 called direct proportion

This is model classification. A straight line is not enough; direct proportion must have zero intercept.

Case 14 · Gradient of line through two points

The learner swaps Δx and Δy and still obtains a number. This is procedure meaning: gradient is vertical change divided by horizontal change.

Case 15 · Geometry angle from a diagram

The learner assumes two lines are parallel because they look parallel. This is evidence/condition reasoning. Diagram appearance is not a theorem condition.

Case 16 · Similar triangles with rotated orientation

The learner pairs “left side with left side”. This is correspondence reasoning, not ratio arithmetic.

Case 17 · Right triangle, choose trig ratio

The learner knows SOHCAHTOA and chooses sine before identifying sides. This is selection reasoning. Mark opposite, adjacent, hypotenuse relative to the reference angle first.

Case 18 · 5.3 m by 420 cm rectangle

The learner multiplies 5.3×420. This is unit preparation. Convert to a common unit before calculating area.

Case 19 · Mean of 7,8,8,9,50

The learner computes 16.4 correctly and calls it “typical”. The issue is statistical interpretation; median 8 may better represent the central cluster depending on purpose.

Case 20 · Two groups with means 70 and 75

The learner says the second group is “better” without spread or criterion. This is claim reasoning. Higher centre does not answer every decision.

Case 21 · Survey 18 of 20 agree

The learner writes “90% of students agree” though the 20 are selected club members. This is sample-scope reasoning.

Case 22 · Probability without replacement

The learner multiplies 4/10 by 4/10 for two red draws. This is sample-space update. The second probability is 3/9 after one red is removed.

Case 23 · “At least one” event

The learner lists many cases and misses one. This is primarily a representation efficiency issue. Use the complement where appropriate.

Case 24 · Real-world model gives −20 litres

The equation is evaluated correctly after the tank should have emptied. This is domain interpretation, not arithmetic failure.

Case 25 · Correct answer, no working

A multi-stage K210 problem receives the right number with no visible model. The issue may be mathematical communication if the question requires essential working or justification. The repair is not necessarily more calculation; it is making the decisive relationship inspectable.

Classification rule

When a learner gets a question wrong, ask what would happen if the correct representation or method were supplied. If the rest becomes easy, the first weakness lies before procedure. If the method is selected independently and execution fails, procedural practice is appropriate. If the result is correct and the conclusion is wrong, interpretation or reasoning is the target. This classification prevents every error from receiving the same prescription: more questions.