Secondary Mathematics Tuition | How to Read Maths Questions: Translate Wording, Conditions and Diagrams into Mathematics

Secondary Mathematics Tuition · Reading, translation and question parsing

Jo reads the question twice.

She knows every formula that appears to be relevant.

She still cannot start.

The difficulty is not yet calculation.

It is translation.

Mira reads the same question and does something different.

She circles what must be found, separates the givens from the conditions, marks the unit, notices that “after a 20% discount” makes the stated price the final amount rather than the original, and writes one relationship before touching the calculator.

Ryan is fast at algebra but often answers the question he expected rather than the one that was written. He sees “triangle”, reaches for a familiar theorem, and only later notices that the condition authorising that theorem was never given.

These are not unusual failures.

Secondary Mathematics questions are written in several languages at once:

ordinary English or another natural language;

mathematical vocabulary;

symbols and notation;

tables;

graphs;

diagrams;

units;

conditions;

and answer-form instructions.

A learner can know the Mathematics and still lose the problem before calculation begins.

This worldwide guide owns that layer.

How to read Maths questions means learning how to convert question language into a reliable mathematical state: identify the target, preserve conditions, recognise relationships, choose a useful representation and understand what form the final answer must take.

It is not a global glossary of every mathematical word.

It is not a list of exam command words detached from actual problem solving.

It is not the full problem-solving system.

The ownership boundary matters. How SEC Mathematics Question Reading Works remains the Singapore SEC G1/G2/G3 owner. Secondary 1 Mathematical Vocabulary & Task Language remains the Secondary 1 owner. Singapore School Mathematics: Command Words, Answer Forms and Task Contracts remains Singapore-specific. Primary and PSLE question-reading owners remain separate.

This page also does not replace How to Solve Unseen Maths Problems Independently, which owns cue-free first attempts; How to Solve Math Problems, which owns the complete problem-solving architecture; or How to Show Working in Maths, which owns the visible mathematical route after the problem has been interpreted.

This page owns the input transformation:

QUESTION AS PRESENTED → TARGET → GIVEN INFORMATION → CONDITIONS → RELATIONSHIPS → REPRESENTATION → ANSWER CONTRACT.

Jo, Mira and Ryan are fictional recurring learners. Their scenes are explanatory examples rather than reported student cases or fixed ability labels.

The central reading system is:

ORIENT → NAME THE TARGET → SORT THE INFORMATION → MARK CONDITIONS → IDENTIFY RELATIONSHIPS → TRANSLATE THE WORDS, SYMBOLS OR DIAGRAM → CHECK UNITS AND DOMAIN → CHOOSE A FIRST REPRESENTATION → STATE THE ANSWER CONTRACT → ONLY THEN SOLVE.

This is not a ritual that every easy problem must display line by line. On a familiar one-step question, several actions collapse into seconds. On a long unfamiliar question, keeping them separate prevents the learner from solving the wrong problem efficiently.

Evidence and scope note: the parsing routines, fictional scenes, translation matrices and checkpoints in this article are original instructional tools, not validated reading tests or psychological instruments. The What Works Clearinghouse problem-solving guide for grades 4–8 gives strong-evidence recommendations to monitoring/reflection and visual representations within its stated grade scope; the NCETM Five Big Ideas emphasise representation and structure, mathematical thinking, fluency, variation and coherence. These sources support related teaching principles, not one universal wording-to-equation algorithm for all curricula.

1. Reading a Maths question is part of the Mathematics, not a prelude outside it

Students sometimes divide a problem into two phases.

First, “read the English”.

Then, “do the Maths”.

That division is too simple.

Reading a Mathematics question already contains mathematical decisions.

Consider:

“A number is increased by 20% and becomes 84. Find the original number.”

The natural-language sentence must become a mathematical relationship.

Original number = x.

After a 20% increase, final = 1.2x.

So 1.2x = 84.

The algebra is easy once the sentence has been interpreted correctly.

The real mathematical work was deciding:

which quantity was original;

which quantity was final;

what “increased by 20%” means;

and which direction the unknown lies.

Now change one phrase:

“A number increases to 120% of its original value and becomes 84.”

The relationship is still 1.2x = 84.

Now change another phrase:

“A number increases by 20 percentage points.”

That is not automatically the same operation as multiplying a quantity by 1.2. Percentage points compare percentages, not ordinary quantities.

Language controls structure.

The same is true in geometry.

“The diagram is not drawn to scale” is not a decorative note.

It tells the learner not to infer lengths or angles from visual appearance.

“AB is parallel to CD” authorises relationships that a merely parallel-looking drawing does not.

“The triangle is right-angled at B” makes B the location of a condition, not just a letter.

In probability:

“without replacement” changes the state after a draw.

“independent events” preserves one probability relationship.

“at least one” describes a set of cases and often invites a complement.

In statistics:

“combined mean” may require totals and counts.

“estimate” changes the expected precision.

“compare” requires a relational statement, not just two calculations.

Reading is therefore mathematical representation.

The learner converts one representation—the presented question—into another state that is easier to reason about.

A strong reader does not merely pronounce every word correctly.

They preserve what each word changes mathematically.

This leads to the first principle:

if changing a word changes the legal mathematics, that word is mathematical information.

2. Find the target before choosing the method

Many misreads begin because the learner notices a familiar object before noticing what must be found.

They see a rectangle and calculate area.

The question asked for perimeter.

They see a quadratic expression and factorise.

The question asked for the minimum value.

They see a graph and calculate gradient.

The question asked for the intercept.

They see two group means and average them.

The question asked for a combined total.

The target controls the route.

Use a target sentence:

“My final answer must be a ______.”

Examples:

length;

angle;

probability;

percentage;

equation;

coordinate;

number of items;

proof that a statement is true;

comparison;

estimate;

range of admissible values.

Jo reads:

“A rectangle has length 12 cm and width 8 cm. Find the amount of fencing required around it.”

Target:

boundary length.

That means perimeter.

The dimensions 12 and 8 could also support area, but area is not the target.

Mira reads:

“The mean of five values is 18. A sixth value is added and the new mean is 20. Find the sixth value.”

Target:

the added value.

That tells her to reconstruct totals before and after.

Ryan reads:

“Show that x² − 6x + 13 is always positive.”

Target:

a general justification of positivity, not simply one numerical evaluation.

Completing the square becomes attractive because:

x² − 6x + 13 = (x − 3)² + 4.

Now positivity is visible.

The target can be explicit:

find;

calculate;

solve;

determine;

show;

prove;

estimate;

state;

compare;

describe;

interpret.

But the command word alone is not enough.

“Find” could request a length, probability, equation or count.

The learner must identify the mathematical object.

A useful margin habit is to write one tiny target cue:

“length”

“x-values”

“probability”

“prove odd”

“compare spread”.

This is especially helpful in long questions where early calculations produce several intermediate quantities.

Without a target cue, the learner may stop at a useful intermediate result and believe the question is finished.

The target is the anchor for everything that follows.

3. Separate mathematical givens from context, decoration and numbers that may not need immediate use

Not every noun matters mathematically.

Not every number should be used immediately.

Word problems often contain a story because the Mathematics is being modelled in a context.

The reader must decide which details constrain the model.

Consider:

“A theatre with 500 seats sells adult tickets for 18 dollars and student tickets for 12 dollars. At one performance, 320 tickets are sold for a total of 4800 dollars. Find how many adult tickets were sold.”

The theatre capacity 500 is context and possibly a plausibility bound.

It is not required to form the two main equations if 320 tickets were sold.

The decisive givens are:

adult price = 18;

student price = 12;

tickets sold = 320;

revenue = 4800.

Let adult tickets = a and student tickets = s.

a + s = 320.

18a + 12s = 4800.

The 500-seat capacity can later verify that 320 is feasible, but forcing it into the equations creates noise.

Now consider:

“A tank can hold 100 litres. It currently contains 40 litres and fills at 5 litres per minute. How long until it is full?”

Here, capacity 100 is essential.

The same kind of number—capacity—can be decorative in one question and decisive in another.

The reader should not use the rule:

“Every number must appear in my calculation.”

Use instead:

“What role does each number play?”

Possible roles include:

target;

initial value;

rate;

constraint;

scale;

comparison reference;

maximum/minimum;

intermediate handoff;

or contextual information not needed for this part.

Mira uses a three-column mental sort:

Need now.

Might constrain later.

Context only.

She does not literally draw the table on easy questions.

On dense questions, the distinction prevents number-grabbing.

Jo learns another useful test:

if I remove this number, does the mathematical relationship become underdetermined?

If yes, it is likely essential.

If no, it may be redundant, checking information or context.

Ryan learns that unused information is not necessarily a trick.

Real modelling environments often contain more information than one calculation needs.

Strong mathematical reading includes restraint.

Use information because the relationship needs it, not because the question printed it.

4. Conditions and constraints decide which methods are legal

A condition is not merely another fact.

It often decides whether a method can be used at all.

Right angle.

Parallel lines.

Similar figures.

Independent events.

Without replacement.

Integer.

Positive.

Distinct.

Exact.

At most.

At least.

Within a given domain.

Conditions create a feasible set.

Consider:

“x is a positive integer and x² < 30.”

The inequality alone permits many real x values.

The positive-integer condition reduces the admissible values to:

x = 1, 2, 3, 4, 5.

Consider:

“Find the length x if x² = 49.”

Algebra gives x = ±7.

If x is a physical length, the context restricts the answer to 7 units.

Consider:

“Two counters are drawn without replacement.”

The second-stage denominator changes.

Ignore the condition and the probability model is wrong before arithmetic begins.

Consider:

“AB is parallel to CD.”

That condition can authorise angle relationships.

A diagram that merely looks parallel does not.

Ryan’s reading error is often method-first.

He thinks:

“This looks like Pythagoras.”

A better order is:

“What condition would authorise Pythagoras?”

Then:

“Do I have it?”

This produces a general rule:

method selection should be condition-backed, not appearance-backed.

Mark conditions early if they are easy to forget.

For algebra:

x ≠ 3.

For modelling:

0 ≤ t ≤ 10.

For geometry:

∠B = 90°.

For probability:

no replacement.

For counts:

n ∈ positive integers.

The exact notation can vary with course level.

The important job is preserving the restriction through the solution.

A condition that disappears from the reader’s state can make a later line look mathematically reasonable while actually being invalid.

5. Read the command word and the answer contract together

Command words tell the learner what kind of response is being requested.

The answer contract tells the learner what the finished response must contain.

These are related but not identical.

“Calculate” may require a number, unit and precision.

“Solve” may require all values satisfying an equation or inequality.

“Show that” may give the destination and ask for a valid route.

“Explain” may require a reason rather than more calculation.

“Compare” requires a relational statement, not two isolated descriptions.

“Estimate” signals that approximation is acceptable or required.

“State” may invite a short answer if no justification is requested.

“Hence” often asks the learner to use an earlier result rather than rebuild from scratch.

These conventions vary across curricula and examinations, so official local instructions remain authoritative.

The transferable reading skill is to combine the verb with the target object and required form.

Jo reads:

“Calculate the gradient of the line, giving your answer to 2 decimal places.”

Her contract is:

object = gradient;

form = numerical;

precision = 2 decimal places.

Mira reads:

“Show that n² + n is even for every integer n.”

Her contract is:

object = general statement;

form = justification valid for all integers n;

not enough = checking n = 1, 2, 3.

Ryan reads:

“Compare the distributions.”

He calculates two means and stops.

The contract requires a comparison:

which centre is larger;

which spread is larger;

what evidence supports the statement;

and perhaps a limitation, depending on the data representation.

A small answer-contract checklist is:

What? Which mathematical object?

How many? One value, all values, interval, ordered pair?

Unit? Is one required?

Precision? Exact, decimal, significant figures, nearest unit?

Reason? Is justification required?

Context? Must the result be interpreted?

Reading this contract before solving prevents a common failure:

doing correct Mathematics that does not finish the asked task.

6. Read mathematical nouns, verbs and relationships as different jobs

Dense questions become easier when the learner stops treating every word as one undifferentiated sentence.

Mathematical language often contains three kinds of information.

Nouns identify objects or quantities.

Length.

Angle.

Gradient.

Mean.

Probability.

Radius.

Integer.

Function.

Sequence.

Verbs identify operations or requested actions.

Increase.

Decrease.

Double.

Compare.

Estimate.

Solve.

Factorise.

Substitute.

Construct.

Relationship words connect quantities.

More than.

Less than.

Twice.

Per.

Of.

Difference between.

Directly proportional.

Parallel.

Similar.

Consecutive.

Without replacement.

The same sentence may contain all three.

“The length of a rectangle is 4 cm more than twice its width.”

Nouns:

length;

width.

Relationship:

4 more than twice.

Translation:

if width = w, then length = 2w + 4.

A common misread is:

length = 2(w + 4).

Why?

The learner notices “twice” and “4 more” but does not parse their grammatical order.

Now compare:

“The length is twice the quantity that is 4 cm more than the width.”

That really is:

length = 2(w + 4).

Small language changes alter bracket structure.

This is why keyword-only reading is fragile.

“More” does not always mean add immediately.

“Of” does not always mean multiply without considering the object.

“Difference” does not tell you subtraction order unless the relationship or target does.

“Per” usually signals a rate but still requires compatible units.

Use a relationship sentence before symbols when wording is dense.

For example:

“The larger quantity equals twice the smaller plus 4.”

Then:

L = 2S + 4.

That intermediate natural-language paraphrase can prevent bracket errors.

Mira does this when a sentence contains nested relationships.

Ryan skips it when the structure is simple.

Jo uses it only until the pattern becomes easy to recognise.

The principle is:

translate relationships, not keywords.

7. Comparative language controls direction: more than, less than, older than, longer than, cheaper than

Comparisons are a major source of translation errors because the order of words and the order of subtraction are not always the same.

Consider:

“A is 7 more than B.”

Correct:

A = B + 7.

Now:

“A is 7 less than B.”

Correct:

A = B − 7.

Now:

“B exceeds A by 7.”

Correct:

B = A + 7

or

B − A = 7.

These translations are equivalent.

A learner who relies on the word order “B … A … 7” may construct the wrong subtraction.

Use a reference rule:

identify the quantity being described, then identify what it is being compared with.

“A is 7 more than B.”

A is being described.

B is the reference.

Start at B and add 7.

A = B + 7.

Now multiplicative comparison:

“A is three times B.”

A = 3B.

“A is 300% of B.”

A = 3B.

But:

“A is 300% more than B.”

A = B + 3B = 4B.

The word “more” changes the relationship.

This difference is easy to underestimate.

“150% of” means 1.5 times.

“150% more than” means 2.5 times.

Comparison language also appears in geometry.

“The radius is 3 cm less than the diameter” is mathematically possible only for a circle with diameter 6 cm and radius 3 cm; the usual defining relationship diameter = 2 radius provides an additional condition.

In age or number problems:

“Mira is 4 years older than Jo.”

M = J + 4.

If asked for age difference, M − J = 4.

In data:

“Group A has a mean 5 points higher than Group B.”

mean A = mean B + 5.

This does not tell us anything directly about spread or individual values.

Comparative language should therefore be read in two layers:

which quantity is described;

what operation or relation connects it to the reference quantity.

A useful self-check is:

if the reference quantity increases, should my translated expression move in the same way the sentence describes?

This catches many reversed comparisons.

8. “Per”, rates, ratios and proportions must be read with units and reference quantities attached

The word “per” often signals division or a rate.

But the exact mathematical object depends on what is measured per what.

60 kilometres per hour.

5 dollars per item.

12 grams per litre.

3 students per computer.

Each rate is a relationship between two quantities with units.

Jo reads:

“A machine produces 240 parts in 6 hours at a constant rate.”

Rate = 240 parts / 6 hours = 40 parts/hour.

If the target is time for 500 parts:

time = 500 parts / (40 parts/hour) = 12.5 hours.

The units help decide which way to divide.

Mira reads:

“A recipe uses flour and sugar in the ratio 5:2. There are 350 g of flour. How much sugar is used?”

The ratio gives parts:

5 parts flour = 350 g

1 part = 70 g

2 parts sugar = 140 g.

Ryan sometimes sees 5:2 and immediately computes 350 × 5/2.

That ignores which ratio term corresponds to the given quantity.

Ratio reading requires correspondence.

Now direct proportion:

“y is directly proportional to x.”

This means:

y = kx

for some constant k, within the model’s intended domain.

It is stronger than merely saying y increases when x increases.

A line y = 3x + 5 increases with x but is not directly proportional in the usual school sense because it does not pass through the origin.

Inverse proportion likewise names a specific relationship:

y = k/x

in its standard form.

Do not translate “inversely related” from ordinary speech automatically into formal inverse proportion unless the context establishes it.

Rate language can also hide a fixed component.

“A taxi charges 4 dollars plus 1.20 dollars per kilometre.”

This is not direct proportion between cost and distance.

It is:

C = 4 + 1.2d.

The phrase “plus 4 dollars” changes the model family.

Good reading attaches the unit and reference:

1.20 dollars per kilometre.

Not just 1.20.

This produces a reliable rate test:

What quantity is in the numerator? What quantity is in the denominator? Is there a starting amount? Are the units compatible?

Those four questions prevent many rate and ratio errors before calculation begins.

9. Percentage language is mostly about identifying the correct 100% base and the direction of change

Percentage problems often become difficult because learners calculate before deciding what quantity represents 100%.

Consider:

“A jacket originally costs 120 dollars and is discounted by 15%.”

Original = 100%.

Final = 85% of original.

Final = 0.85 × 120 = 102.

Now reverse the information:

“After a 15% discount, a jacket costs 102 dollars. Find the original price.”

The 102 is not 100%.

It is 85%.

0.85P = 102.

P = 120.

The same numbers can support opposite operations depending on which quantity is known.

Now:

“A value increases from 80 to 100. Find the percentage increase.”

Increase = 20.

Base = original 80.

Percentage increase = 20/80 × 100% = 25%.

A common misread divides by 100 because 100 is visually prominent.

The correct base comes from the meaning of “increase from”.

Now compare:

“The pass rate rises from 60% to 72%.”

Increase = 12 percentage points.

Relative percentage increase = 12/60 × 100% = 20%.

These are different statements.

Now compare:

“A quantity is 20% less than B.”

A = 0.8B.

“B is 20% more than A.”

B = 1.2A.

These do not imply that A is 20% less than B and B is 20% more than A symmetrically in both directions. If B = 1.2A, then A is 1/1.2 = 83.33…% of B, about 16.67% less than B.

Percentage comparisons are reference-dependent.

Jo writes the base before the arithmetic when the wording is easy to reverse.

Mira writes:

100% = original

or

100% = reference group.

Ryan uses a verbal test:

“Percentage of what?”

That question should be automatic.

A reliable reading sequence is:

identify the 100% base → identify direction → convert the percentage relationship → then calculate.

10. “At least”, “at most”, “more than”, “no more than” and “between” describe sets, not single answers

Inequality language is easy to misread because ordinary speech is flexible while mathematical boundaries are exact.

“At least 5” means 5 or more:

x ≥ 5.

“More than 5” means strictly greater:

x > 5.

“At most 5” means 5 or less:

x ≤ 5.

“Less than 5” means strictly less:

x < 5.

“No more than 12” means:

x ≤ 12.

“No fewer than 12” means:

x ≥ 12.

“Between 3 and 8” can be ambiguous in ordinary language about whether endpoints are included. A mathematical question should ideally specify the intended inclusion, or the context may make it clear. If precision matters, the reader should not invent endpoint inclusion from habit.

Probability language uses similar structures.

“At least one success” includes one, two, three, and so on, up to the maximum possible.

Its complement is “no successes”.

“At most one success” includes zero or one.

“Exactly one” includes only one.

These are different events.

Counting language also matters.

“More than 20 students” means at least 21 if the count is integer-valued.

“At least 20 students” includes exactly 20.

For continuous quantities, there may be infinitely many values between bounds.

The learner should preserve quantity type.

Now consider optimisation:

“The cost must not exceed 500 dollars.”

Cost ≤ 500.

This is a constraint, not necessarily the target.

Another part of the problem may ask to maximise output subject to that constraint.

Mira marks inequality phrases because their small words matter.

Jo translates them into a symbol early.

Ryan checks boundary inclusion by testing the exact endpoint against the sentence.

Would 5 satisfy “at least 5”?

Yes.

Would 5 satisfy “more than 5”?

No.

That simple boundary test is powerful.

The reading principle is:

inequality words define a set of allowed states; preserve whether the boundary belongs to the set.

11. Define variables from quantities, not from convenient letters alone

Word problems become easier when symbols are attached to meanings before equations are built.

The variable name itself does not matter much.

The definition does.

Consider:

“The sum of two consecutive integers is 51.”

Let the smaller integer be n.

Then the next integer is n + 1.

Relationship:

n + (n + 1) = 51.

The variable definition carries the word “consecutive” into the algebra.

A weak translation is:

x + y = 51.

That equation loses the fact that y must be exactly one more than x.

It uses two variables where one relationship can reduce the state to one.

Now:

“The length of a rectangle is 4 cm more than its width and its area is 96 cm².”

Let width = w cm.

Length = w + 4 cm.

Then:

w(w + 4) = 96.

The definition creates the model.

Now a ticket problem:

“Adult tickets cost 15 dollars and student tickets cost 9 dollars. A total of 40 tickets brings in 492 dollars.”

Let a = number of adult tickets.

Let s = number of student tickets.

a + s = 40.

15a + 9s = 492.

The variable definitions make the coefficients readable.

15 multiplies the number of adult tickets because it is dollars per adult ticket.

Good variable definitions also prevent unit confusion.

Let t = time in hours.

Let d = distance in kilometres.

Let C = cost in dollars.

The units need not be repeated on every algebraic line, but they should be known.

Jo chooses variable names that remind her of the quantity.

Mira sometimes uses x and y but writes the definitions explicitly.

Ryan once uses x for width in one line and then unconsciously treats x as length in another. A clear definition prevents symbol drift.

A strong variable question is:

What quantity can I name so that the remaining quantities can be expressed from it?

This can reduce the number of unknowns.

If two quantities are related by “4 more than”, do not automatically give them unrelated variables.

Use the relationship.

If the problem truly contains two independent unknowns, define two.

The goal is not to minimise variables at all costs.

It is to preserve the structure in the simplest state that remains faithful to the question.

12. Translate words into equations by preserving relationships one clause at a time

Long sentences are easier to translate when they are decomposed into mathematical clauses.

Consider:

“A company’s monthly charge is 12 dollars plus 0.08 dollars for each message above the included allowance.”

Clause 1:

fixed charge = 12.

Clause 2:

additional charge = 0.08 per chargeable message.

If m is the number of chargeable messages:

C = 12 + 0.08m.

Now:

“A number is three less than twice another number.”

Let the other number be x.

Twice it = 2x.

Three less than that = 2x − 3.

Now:

“The square of a number is 10 more than three times the number.”

x² = 3x + 10.

Each phrase maps to a relationship.

Do not hunt for an operator word and immediately perform it.

Build the structure in order.

Nested phrases should trigger brackets.

“Three times the sum of x and 4”:

3(x + 4).

“The sum of three times x and 4”:

3x + 4.

Same words.

Different grammar.

Different Mathematics.

Now rates:

“The total distance is 30 km more than twice the distance travelled on Monday.”

If Monday = m:

total = 2m + 30.

Not:

2(m + 30).

Now proportions:

“The number of red counters is 3/5 of the number of blue counters.”

R = (3/5)B.

But:

“The ratio of red to blue counters is 3:5.”

R/B = 3/5

or

R = (3/5)B.

Those two statements happen to produce the same relationship.

Now:

“Red counters are 3/5 of all counters.”

R = (3/5)(R + B).

The reference quantity changed from blue counters to total counters.

This is why nouns matter as much as fractions.

A useful translation discipline is:

underline the quantity each number or fraction modifies.

3/5 of what?

20% of what?

4 more than what?

Twice what?

That question prevents many syntactic misreads.

13. Multi-sentence word problems should be converted into a small relationship map before full calculation begins

Long word problems feel difficult partly because information arrives over several sentences.

The learner may hold the first sentence mentally while reading the third.

By the end, relationships have blurred.

A relationship map externalises the state.

Consider:

“A water tank contains 20 litres. Water enters at 6 litres per minute. At the same time, 2 litres per minute leaves through an outlet. The tank can hold 80 litres. Assuming the rates remain constant, how long until the tank is full?”

Do not calculate after the first rate appears.

Map:

initial = 20 L.

inflow = 6 L/min.

outflow = 2 L/min.

net rate = 4 L/min.

target volume = 80 L.

relationship:

20 + 4t = 80.

Then solve:

t = 15 min.

The relationship map prevents the learner from using the inflow alone.

Now:

“A shop buys notebooks for 4 dollars each. It sells them at a 25% markup. During a sale, the marked price is discounted by 20%. Find the sale price.”

This is not a single percentage.

Map the stages.

cost = 4.

after 25% markup:

4 × 1.25 = 5.

after 20% discount:

5 × 0.8 = 4.

The final price happens to return to 4, but the two percentage changes do not cancel because 25 − 20 = 5. They act on different bases.

Now linked motion:

“A cyclist travels 18 km at 12 km/h and then 24 km at 16 km/h. Find the average speed for the whole journey.”

A weak reader averages 12 and 16.

The relationship map shows:

total distance = 42 km.

time 1 = 18/12 = 1.5 h.

time 2 = 24/16 = 1.5 h.

total time = 3 h.

average speed = 42/3 = 14 km/h.

In this particular case, 14 also equals the simple average because the times are equal.

But the correct method came from total distance/total time, not from guessing that rates should be averaged.

Multi-sentence reading should therefore delay arithmetic until the relationships are assembled.

A useful page note is:

given → relation → target.

For long questions, this may be faster than repeated rereading because the learner no longer has to reconstruct the state each time.

14. Read a table by asking what each row, column and cell represents before looking for a pattern

Tables are compressed language.

They remove sentences and place relationships spatially.

That makes them efficient.

It also creates reading risks.

Before calculation, identify:

row variable;

column variable;

units;

whether entries are values, frequencies, probabilities, intervals or categories;

and whether totals are given or must be reconstructed.

Consider:

x: 0, 1, 2, 3

y: 5, 8, 11, 14.

A learner may notice “+3” and say the rule is y = 3x.

That ignores the value at x = 0.

Better reading:

constant first difference = 3;

when x = 0, y = 5;

so y = 3x + 5.

Now a frequency table.

Value x: 1, 2, 3, 4.

Frequency f: 2, 5, 1, 2.

The row “frequency” does not contain more data values of the same type as x.

It tells how many times each x occurs.

Mean requires:

Σfx / Σf.

Reading the table structure comes before formula choice.

Now grouped data:

0–10, 10–20, 20–30.

Frequencies: 4, 7, 9.

If the task asks for an estimated mean, class midpoints may be used in a standard school method.

The word “estimated” matters because exact individual values are unavailable.

Now a two-way table.

Rows = year group.

Columns = transport mode.

The learner must decide whether a requested probability is joint, conditional or marginal by identifying which population forms the denominator.

Mira always reads table labels before numbers.

Jo checks totals if probabilities are involved.

Ryan once assumes the first row is x and second y without reading headings. That habit fails as soon as the table type changes.

The table-reading rule is:

name the variables and units before searching the cells for operations.

15. Read a graph from axes and scale outward; shape comes after meaning

A graph is not a picture first.

It is a coordinate representation.

Start with axes.

What does x represent?

What does y represent?

What are the units?

What is the scale?

Does the graph begin at zero?

Is the horizontal variable continuous, discrete, categorical or time-based?

Only then interpret shape.

Consider a graph where the vertical axis begins at 90 rather than zero.

A small numerical difference can look visually dramatic.

A learner who reads shape before scale may overstate the change.

Consider a distance-time graph.

Gradient represents speed if distance is vertical and time horizontal.

A horizontal section means distance is unchanged: stationary.

On a speed-time graph, a horizontal section means constant speed, not stationary unless the speed level is zero.

Same visual shape.

Different axis meaning.

Now a straight line:

y = 3x + 5.

Gradient = 3.

Intercept = 5.

If x is number of items and y is total cost in dollars, the interpretation may be:

fixed cost 5 dollars;

variable cost 3 dollars per item.

If axes represent completely different quantities, the same numbers mean something else.

Graphical intersection also depends on meaning.

Two cost lines intersect where the two costs are equal.

Two motion graphs intersect where the represented position quantities are equal at the same time.

Two functions intersect where they have the same y-value for the same x.

Now graphical approximation:

If a root is read at x ≈ 2.4, the answer is usually approximate to the resolution of the graph.

Writing x = 2.4 exactly may claim more than the representation supports.

Ryan used to identify “increasing graph” without saying what increased.

He now writes a full interpretation:

“As time increases, distance increases at a constant rate.”

Mira checks scale before gradient.

Jo reads any legend or multiple-series key before comparing lines.

A reliable graph scan is:

axes → units → scale → domain → key points → shape → relationship → target.

That order prevents the visual impression from outrunning the mathematics.

16. Read a diagram as a set of stated and derived relationships, not as a drawing to be trusted by eye

Diagrams are powerful because they compress spatial relationships.

They are dangerous when learners assume that visual appearance is mathematical evidence.

A line may look horizontal without being stated horizontal.

Two lengths may look equal without equality marks.

An angle may look like 90° without a right-angle symbol.

A point may look like a midpoint without being defined as one.

The reader should distinguish three states.

Given: explicitly stated or marked.

Derived: justified from given information.

Looks like: visual impression only.

Only the first two belong in mathematical reasoning.

Consider a triangle drawn with one angle that appears right-angled.

If the question says “not drawn to scale” and gives no right-angle condition, Pythagoras is not automatically authorised.

Now suppose a small square marks the angle.

The condition is given.

Now Pythagoras or right-triangle trigonometry may be available depending on the known quantities and target.

Consider parallel lines.

Arrow markings or a statement AB ∥ CD create a relationship.

The learner may then derive angle equalities.

The angle equality is derived working, not a visual guess.

Now similarity.

Two triangles may look similar.

That is not sufficient.

The reader looks for angle relationships or side proportionality conditions appropriate to the course.

Once similarity is established, correspondence matters.

A rotated diagram can make this difficult.

Mira writes a correspondence line before ratios:

A ↔ D, B ↔ E, C ↔ F.

Then:

AB/DE = AC/DF.

Jo marks the target directly on the diagram if it is not already labelled.

Ryan used to copy every number from the diagram into his working immediately. He now asks what each label means and whether it belongs to the relevant shape.

For composite figures, isolate the smallest relevant sub-diagram.

A busy diagram may contain several triangles, circles or rectangles.

Redrawing only the relevant part can reduce reading load.

But the redraw must preserve all conditions that matter.

If a right angle or parallel relation disappears in the redraw, the learner may lose method authority.

A reliable diagram scan is:

target → given marks → labels → conditions → derived relationships → relevant subfigure → method candidates.

The central rule is:

read markings and statements as evidence; treat unmarked appearance as provisional only.

17. Units and dimensions are part of the reading state because they tell you what kind of quantity the answer can be

Units often reveal relationships before formulas do.

Distance in kilometres.

Time in hours.

Speed in kilometres per hour.

Area in square centimetres.

Volume in cubic metres.

Cost in dollars.

Rate in dollars per item.

A question can be misread even when every number is copied correctly if unit meaning is lost.

Consider:

“A runner travels at 5 m/s for 3 minutes.”

The rate uses seconds.

The time uses minutes.

The first mathematical reading action is:

3 min = 180 s.

Then:

distance = 5 × 180 = 900 m.

Now:

“A rectangle measures 12 cm by 8 cm. Find its area.”

Target type = two-dimensional measure.

Answer unit = cm².

If the learner writes 96 cm, the arithmetic may be right while the quantity type is wrong.

Now density:

mass / volume.

If mass is in grams and volume in cm³, density is g/cm³.

The compound unit helps the learner decide numerator and denominator.

Units can also expose impossible operations.

Adding 5 cm to 3 cm² has no ordinary geometric meaning because the dimensions differ.

Subtracting 5 dollars from 12 dollars is coherent.

Comparing 5 dollars/item with 12 dollars requires context because the types differ.

Ryan uses a unit prediction before calculation:

“My final answer should be hours.”

If his algebra ends in kilometres, he knows something has gone wrong.

Mira marks conversions before arithmetic.

Jo notices that scale factors change dimension:

length scale k;

area scale k²;

volume scale k³.

The reading state therefore includes dimensional type.

A useful unit scan is:

what unit is given → what unit is required → what conversion is needed → what compound unit does the relationship imply?

This is one of the cheapest pre-calculation checks available.

18. Read exactness, estimation and rounding language before deciding what kind of numerical answer is allowed

Mathematical questions do not always ask for the same numerical status.

Some require exact values.

Some require decimal approximations.

Some ask for estimates.

Some specify significant figures or decimal places.

The reader should know this before deciding when to round.

Consider:

“Give the exact value of √50 in simplest surd form.”

Target:

exact expression.

√50 = 5√2.

A decimal such as 7.071 is not the requested form.

Now:

“Give your answer correct to 3 significant figures.”

The learner may preserve full calculator precision internally and round the final result.

Now:

“Estimate the value of…”

The task may intentionally invite convenient approximations rather than calculator-exact evaluation.

The course or assessment context determines what type of estimation is expected.

Now a graph:

“Use the graph to estimate the root.”

The answer is inherently approximate to the graph’s resolution.

Writing an unnecessary trail of decimal digits does not increase accuracy.

Now measurements:

if a length is given as 12.4 cm to the nearest 0.1 cm, some courses may ask for bounds. The phrase “to the nearest 0.1 cm” then carries interval information, not merely display formatting.

Jo reads precision instructions at the beginning and again before the final line.

Mira keeps exact forms through intermediate calculations when practical.

Ryan used to round every line to two decimal places because he thought tidy numbers were safer. He learned that early rounding can accumulate error.

The reading principle is:

answer precision is part of the question contract, not a cosmetic choice after the Mathematics is finished.

19. Linked question parts create result handoffs: read what has already been established before rebuilding the Mathematics

Multi-part questions often create an internal dependency chain.

Part (a) produces a result.

Part (b) asks the learner to use it.

Part (c) may reinterpret or extend it.

The reader should preserve these handoffs.

Suppose part (a) asks:

“Show that the equation can be written as x² − 5x + 6 = 0.”

Part (b):

“Hence solve the equation.”

The word “hence” strongly suggests using the established quadratic form.

The learner should not reconstruct the original modelling step unless needed for checking.

Factor:

(x − 2)(x − 3) = 0.

So x = 2 or 3.

Now geometry.

Part (a) finds an angle.

Part (b) uses that angle in trigonometry.

The learner should carry the angle with its meaning, not only its number.

Write:

∠ABC = 42° from (a).

Then use it.

Now statistics.

Part (a) estimates a mean.

Part (b) compares it with another distribution.

The earlier value becomes evidence, not a new target to recompute.

Linked parts create two common failures.

Failure 1 — ignoring the handoff.

The learner starts part (b) from scratch and wastes time.

Failure 2 — using a handoff without preserving its conditions.

An earlier result may depend on a domain, approximation or assumption. Those limitations still matter later.

For example, if part (a) gives an approximate value, later calculations using it remain approximate.

If part (a) established a model only for 0 ≤ t ≤ 10, part (b) should not use t = 15 without questioning the domain.

Ryan boxes only results that later parts may reuse.

Mira writes a tiny provenance note:

“from (a)”

so the source remains visible.

Jo rereads the final sentence of the previous part before starting the next.

The linked-part reading rule is:

treat earlier results as typed mathematical objects with conditions, not just numbers available for reuse.

20. Learn to leave information unused when the target and relationships do not need it

Many learners feel uncomfortable when a number remains unused.

They assume every printed value must enter a calculation.

That belief makes them vulnerable to irrelevant information.

Consider:

“A rectangular sports court is 28 m long and 15 m wide. It is surrounded by 200 seats. Find the area of the court.”

The 200 seats do not matter.

Area = 28 × 15 = 420 m².

Forcing 200 into the calculation damages the model.

Now:

“A rectangular sports court is 28 m long and 15 m wide. Two hundred seats are placed around it at equal spacing. Find the average number of seats per metre of perimeter.”

Now 200 is essential.

The same datum changes role because the target changed.

This proves why “unused number” is not a permanent property of a number.

Use relevance testing.

Ask:

Does this datum define the target?

Does it define a relationship?

Does it constrain the solution?

Does it provide a scale or conversion?

Does it become useful only in a later part?

If none apply, leave it unused unless a new need emerges.

Irrelevant information can also be verbal.

A story may say the journey occurred on Tuesday.

The day may have no mathematical role.

A graph may include another series that is not relevant to the requested comparison.

A diagram may contain a length belonging to a different subfigure.

Strong reading is selective.

Jo once tried to use every number because she worried that ignoring one meant missing something.

Her tutor changed the question:

“What mathematical job would this number perform?”

If no job can be stated, do not force one.

Mira keeps potentially useful information visible but inactive.

Ryan uses irrelevant-information practice deliberately: questions where one or two values are true but unnecessary.

This trains restraint.

The principle is:

information belongs in a solution only when a mathematical relationship or constraint gives it a job.

21. A strong reader can recognise when the given information is insufficient, inconsistent or deliberately incomplete

Not every question with numbers has one determinate numerical answer.

Sometimes the correct mathematical reading is:

there is not enough information;

the conditions are inconsistent;

there are several admissible answers;

or a range is possible rather than one value.

Consider:

“A triangle has two sides of length 7 cm and 10 cm. Find the third side.”

There is no unique third side from those two lengths alone.

Triangle inequality gives:

3 < x < 17.

If the question asks for one exact length without another condition, the problem is underdetermined.

Now add:

“The two given sides meet at a right angle.”

Now the third side is determined by Pythagoras:

x = √149 if it is the hypotenuse.

One condition changes the problem from a range to a unique value.

Now consider simultaneous equations:

x + y = 10

2x + 2y = 20.

The second equation does not add an independent relationship.

Infinitely many pairs satisfy both.

A learner who expects every pair of equations to produce one ordered pair may continue manipulating until they manufacture an answer.

Good reading recognises dependence.

Now:

x + y = 10

x + y = 12.

The conditions are inconsistent.

No pair satisfies both.

Now modelling.

A problem may state that a machine produces at a constant rate but omit the rate and provide no time-output pair.

If the target depends on the missing rate, the model cannot be completed numerically.

Now probability.

“A bag contains red and blue counters. One counter is chosen. Find the probability of red.”

Without counts or a distribution rule, there is not enough information.

Strong readers do not respond to missing information by guessing a hidden assumption.

They ask:

What relationship is missing?

Could a range be given instead?

Is there a condition elsewhere in the diagram or previous part?

Is an assumption explicitly intended by the course?

Ryan used to interpret “I cannot solve this” as personal failure.

He now distinguishes:

“I cannot solve it because I have not found the relationship”

from

“the information does not determine one answer.”

That distinction is mathematical maturity.

The reading principle is:

before searching harder for a number, check that the question state actually determines the object being requested.

22. Train near-neighbour wording because one small phrase can change the entire method

Reading improves fastest when learners compare questions that look almost the same but differ in one decisive phrase.

This makes the language-to-structure boundary visible.

Pair 1 — increase by / increase to.

“Increase 80 by 25%.”

80 × 1.25 = 100.

“Increase 80 to 125% of its value.”

Same result.

But “increase to 25% of its value” would reduce it to 20, not increase it by 25%.

Pair 2 — 20% of / 20% more than.

20% of B = 0.2B.

20% more than B = 1.2B.

Pair 3 — at least / more than.

x ≥ 5 versus x > 5.

Pair 4 — at most / less than.

x ≤ 5 versus x < 5.

Pair 5 — with replacement / without replacement.

Stable probability state versus changed state.

Pair 6 — directly proportional / linear.

y = kx versus y = mx + c with possibly nonzero c.

Pair 7 — area / perimeter.

Same dimensions; different target type and units.

Pair 8 — exact / approximate.

Same underlying value; different answer form.

Pair 9 — show that / verify for these examples.

General proof versus finite checking.

Pair 10 — average of rates / average rate over a journey.

The latter depends on total distance and total time; the simple average is valid only under special conditions.

Pair 11 — consecutive integers / consecutive even integers.

n, n + 1 versus a structure with difference 2 and parity constraints.

Pair 12 — similar-looking / mathematically similar.

Visual resemblance versus established similarity conditions.

For each pair, ask three questions:

What words stayed the same?

What phrase changed?

What mathematical decision changed because of that phrase?

Mira keeps a “one-word switch” set.

Ryan predicts whether the method changes before solving.

Jo writes the translated relationship only; she does not need to finish every calculation.

This is efficient practice because the target skill is reading discrimination rather than arithmetic volume.

The learner becomes less dependent on memorised keywords because the same keyword appears in contrasting structures.

The goal is not to become suspicious of every word.

It is to notice the words that change mathematical legality, reference or target.

23. Tutors should diagnose where reading breaks before explaining the entire solution

A tutor sees a student stuck on a word problem.

The fastest response is often to say:

“This is simultaneous equations.”

Then define the variables and start the model.

The student may complete the algebra.

But the tutor has carried the reading layer.

To build independence, locate the first failed reading responsibility.

Can the learner state the target?

Can they identify the quantities?

Can they separate givens from conditions?

Can they paraphrase the relationship?

Can they choose variables?

Can they translate one sentence?

Can they read the diagram markings?

Can they identify the answer contract?

Use the smallest prompt that restores the missing job.

Examples:

“What exactly must your final answer describe?”

“Which quantity is the 100% base?”

“Which word tells you the second draw changes the state?”

“What is given, and what are you only assuming from the picture?”

“What does 3 dollars per item mean if there are n items?”

“Which part of this table is a frequency rather than a value?”

“What unit should the final answer have?”

Jo’s tutor notices that she can solve equations but needs help translating prose.

So lessons include short translation-only tasks.

No full solving.

Just:

define variables;

write relationships;

state the target.

Mira is strong in words-to-equations but weak on diagram evidence.

Her tutor uses geometry questions with deliberately misleading visual appearance.

Ryan reads too fast under time and drops small conditions.

His tutor trains a five-second condition scan before method choice.

Feedback should name the mechanism.

Not:

“Read carefully.”

But:

“You read the target correctly, but you treated the final price as the 100% base.”

Or:

“Your algebra is fine. The error entered when ‘at most’ became ‘less than’.”

Specific diagnosis teaches what careful reading actually means.

The tutor’s goal is not to make every question easier to read forever.

It is to make the learner’s internal parser stronger.

24. Parents can support question reading without becoming the person who translates every problem

At home, a child says:

“I don’t understand what the question wants.”

A parent may be tempted to rephrase the entire problem.

Sometimes clarification is appropriate.

Repeated full translation can create dependence.

A stronger support sequence uses neutral prompts.

What is the question asking you to find?

Which numbers describe the same quantity?

Which word changes the condition?

What does the unit tell you?

Can you draw or define the quantities?

What part do you understand already?

Where exactly does the wording stop making sense?

These questions return the reading work to the learner.

A parent does not need to know the method to ask:

“What is your final answer supposed to be—a length, time, cost, percentage or something else?”

That prompt can often restore orientation.

Parents can also avoid unhelpful habits.

Do not say:

“Use all the numbers.”

Some information may be irrelevant.

Do not say:

“This is the same as the question yesterday.”

One changed condition may make it different.

Do not immediately point to a keyword and name an operation.

Keyword rules can work on simple exercises and fail badly on transfer.

Instead, ask for the relationship in a sentence.

Parents can look for progress signals.

The learner rereads less because they build a small relationship map.

They ask more precise questions.

They stop using every number automatically.

They mark conditions before calculating.

They identify unit mismatches without being told.

They distinguish “I do not know the method” from “I do not understand the wording”.

These are meaningful reading gains.

If the language itself is genuinely unfamiliar—for example, a learner is working in a second language—clarifying vocabulary can be necessary. The goal is still to preserve the mathematical relationship rather than turning every problem into an adult translation exercise.

Home support is strongest when it helps the learner build a repeatable reading routine and then gets out of the way.

25. AI can clarify wording, but it should not silently perform the entire mathematical translation before the learner tries

AI tools can be useful when a question’s wording is confusing.

They can also remove the exact reading decisions the learner needs to practise.

If the learner pastes a question and asks:

“Solve this,”

the AI may supply:

target;

variable definitions;

equations;

method;

calculation;

check.

The learner reads a clean route but has not practised parsing.

A better AI sequence preserves ownership.

Stage 1 — learner first.

Write:

target;

givens;

conditions;

one attempted representation.

Stage 2 — ask for clarification, not solution.

Examples:

“Restate this question in simpler language without solving it.”

“Tell me what ‘at most’ means here, but do not choose the method.”

“Check whether I identified the 100% base correctly.”

“Do these variable definitions match the quantities in the question?”

“What information in the diagram is explicitly given rather than visually assumed?”

Stage 3 — learner translates.

Convert the clarified wording into Mathematics.

Stage 4 — AI critiques.

Ask whether the equation faithfully represents the sentence.

Stage 5 — changed problem without AI translation.

Test transfer.

Mira uses AI to generate near-neighbour wording pairs.

Jo asks AI to create a version with one irrelevant datum.

Ryan asks for three ways the same relationship can be worded.

These uses can strengthen reading discrimination.

AI output must still be checked.

A system may paraphrase too aggressively, assume a curriculum convention or miss ambiguity.

If a question is from a formal assessment, the original wording remains authoritative.

The tool should help the learner inspect language.

It should not become a permanent preprocessor that turns every question into ready-made equations.

The long-term direction is:

AI CLARIFIES → LEARNER TRANSLATES → AI CRITIQUES → LEARNER TRANSFERS WITHOUT AI.

26. Practise question reading as a separate skill instead of hiding it inside full problem sets

If a learner misreads questions, giving more full questions may not repair the reading layer efficiently.

The student may spend most practice time on algebra or arithmetic while repeating the same translation error.

Question reading can be isolated.

Use short tasks where the learner does not solve the whole problem.

Task type 1 — target only.

Read ten questions and state only the required mathematical object.

Length.

Probability.

Equation.

Comparison.

Proof.

Task type 2 — conditions only.

Mark the words or diagram features that authorise or restrict methods.

Task type 3 — variable definitions only.

Define quantities without solving.

Task type 4 — translation only.

Turn one sentence into an equation, inequality, table, diagram or relationship.

Task type 5 — relevant versus irrelevant.

Classify numbers by job.

Task type 6 — answer contract.

State unit, precision, number of solutions or explanation requirement.

Task type 7 — near-neighbour contrast.

Compare two almost-identical questions and identify the phrase that changes the Mathematics.

Task type 8 — representation reading.

Interpret tables, graphs and diagrams without calculation.

Jo spends one ten-minute block a week on translation-only practice.

She becomes faster at word-to-equation mapping without doing more algebra volume.

Mira uses contrast pairs.

Ryan uses timed target-and-condition scans because his errors appear when pace increases.

A simple progression is:

direct wording → paraphrase variation → irrelevant information → changed representation → near-neighbour mix → linked parts → timed scan.

The goal is not to make reading slow and ceremonial.

It is to make the important distinctions automatic.

At first, the learner may underline and annotate heavily.

Later, much of the parsing becomes internal.

Practice has succeeded when the learner can read quickly without losing mathematical state.

27. Under examination time, read for high-value information rather than rereading the entire question repeatedly

Time pressure changes reading behaviour.

Some students rush.

Others reread the full question three or four times because they do not trust their first parse.

Both can be expensive.

A timed reading routine should extract high-value state early.

First pass: target and broad structure.

What kind of problem is this?

What must be found?

Second pass: conditions, units, relationships and constraints.

What words could change method legality?

Third action: externalise the state.

Write the variable, equation, diagram mark or target cue.

Once the state is externalised, stop rereading everything unless the route later conflicts with the problem.

Ryan used to reread the whole paragraph after every algebraic line.

Now he writes:

target = original price;

final = 102;

discount = 15%;

0.85P = 102.

The question has been converted into a state he can work from.

Jo circles “without replacement” once.

She does not need to reread that phrase at every probability branch because the state change is already represented.

Mira labels graph axes and units before calculating gradient.

This reduces late discovery that the scale was misread.

Use a leave-and-return rule for reading stalls.

If the learner cannot form a useful representation after a reasonable attempt, mark the unresolved point:

“unclear base”

“which theorem?”

“missing relationship?”

Then move on if the assessment strategy permits and return later.

The 140 unseen-problem and examination-craft owners handle broader paper strategy.

For question reading, the timed principle is:

extract enough state that rereading becomes selective rather than total.

28. When reading keeps failing, identify the first interpretation error instead of telling the learner to “read more carefully”

“Read carefully” is advice without a mechanism.

A better diagnosis names the first broken reading job.

Failure 1 — target substitution

The learner answers a familiar question rather than the stated one.

Repair: target sentence before method choice.

Failure 2 — keyword reflex

A word such as “more”, “per” or “triangle” triggers an operation without relationship analysis.

Repair: paraphrase the relationship, then symbolise.

Failure 3 — reference-base error

Percentage or ratio is applied to the wrong quantity.

Repair: write “100% = ?” or “ratio term ↔ quantity”.

Failure 4 — condition loss

The learner notices a condition initially but forgets it during solving.

Repair: mark it visibly next to the representation.

Failure 5 — diagram assumption

The learner trusts visual appearance.

Repair: classify given, derived and looks-like information.

Failure 6 — scale blindness

The learner reads graph shape before axes and scale.

Repair: axes→units→scale scan.

Failure 7 — every-number compulsion

The learner forces all data into the model.

Repair: require a mathematical job for each used datum.

Failure 8 — premature calculation

Arithmetic begins before relationships are assembled.

Repair: relationship map first on multi-sentence problems.

Failure 9 — answer-contract loss

The learner obtains a mathematically useful quantity but does not provide the requested form.

Repair: target/unit/precision cue before solving.

Failure 10 — linked-part amnesia

Earlier results are ignored or reused without their conditions.

Repair: handoff note “from (a)” plus state/precision.

Failure 11 — language comprehension gap

The learner does not know the ordinary or mathematical meaning of a word.

Repair: teach the vocabulary explicitly, then retest in a new sentence.

Failure 12 — knowledge gap mistaken for reading failure

The learner parses the problem correctly but has never learned the required relationship.

Repair: teach the missing Mathematics. Do not keep rephrasing a question whose mathematical content is genuinely unknown.

This last distinction matters.

Reading support cannot replace missing content knowledge.

Likewise, content reteaching cannot fix a persistent translation error if the learner already knows the underlying method.

The first failed layer determines the next intervention.

29. Frequently asked questions about reading Maths questions

Why can I do the Maths when someone explains the question, but not by myself?

Your execution may be stronger than your translation. Practise identifying the target, quantities, conditions and relationships before anyone names the method. Use changed wording and small translation-only tasks.

Should I underline keywords?

Underline high-value words if it helps, but do not treat keywords as automatic operations. “More”, “per”, “at least” and “similar” have meaning only inside the relationship and condition.

What should I read first?

Find the target, then scan givens, conditions, units and answer-form instructions. On graphs and diagrams, also read axes, scale, markings and labels before interpreting appearance.

Do I have to use every number?

No. Use a number only when it has a mathematical job: defining a relationship, constraint, conversion, target, scale or later handoff.

How do I know which formula to use?

Question reading comes before formula choice. Translate the state and conditions first. Then the method-selection layer can compare candidate relationships.

Why do “more than” questions confuse me?

Identify which quantity is being described and which is the reference. “A is 7 more than B” means start from B and add 7: A = B + 7.

What does “at least” mean?

It includes the boundary. “At least 5” means x ≥ 5. “More than 5” means x > 5.

Why are reverse-percentage questions harder?

The final quantity is given but the original 100% base is unknown. Write the relationship final = multiplier × original before deciding the operation.

How do I read graphs better?

Read axes, units, scale, domain and legend before shape. A horizontal segment means different things on distance-time and speed-time graphs because the vertical variable differs.

How do I stop trusting the diagram?

Separate given markings and derived facts from visual appearance. Use only stated or justified relationships.

What if the wording is ambiguous?

Do not silently invent a condition. Check whether the course convention, diagram, previous part or context resolves it. If not, the mathematically honest response may be that more information is needed.

Should I rewrite every word problem in simpler English?

No. Paraphrase only when the relationship is unclear. The long-term goal is to translate the original wording directly into mathematical state.

Can I draw a diagram even if none is given?

Yes, when a diagram helps represent the relationships. Mark only what the question gives or what you can derive; do not add assumptions from appearance.

How do units help reading?

Units identify quantity type, reveal conversions and often indicate which way a rate relationship should be arranged.

How do I know if I should use an equation or a table?

Choose the representation that makes the relationship and target easier to inspect. There is no universal requirement that word problems become equations first.

What does “hence” mean?

It usually signals that an earlier result should be used. Exact conventions vary by assessment, but the reading job is to preserve the earlier result and its conditions rather than restart blindly.

How can I practise without doing hundreds of questions?

Use short parsing drills: target only, variables only, relationships only, condition spotting, near-neighbour wording and representation reading. These train the bottleneck directly.

Can AI help me understand wording?

Yes. Ask for clarification or a simpler paraphrase without a solution, then do the translation yourself and test a changed problem without AI help.

Why do I misread more under time pressure?

Pace encourages premature method choice and skipped conditions. Train a brief high-value scan and externalise the state so you do not need to reread the whole question repeatedly.

What is the single best reading habit?

Before calculating, be able to say what must be found, what relationships are given, and what conditions make those relationships legal.

30. The complete question-reading system: convert the presented problem into a mathematical state before asking the mathematics to solve it

Return to Jo, Mira and Ryan.

Jo’s difficulty was not a lack of formulas.

Her problem was that the question remained verbal for too long.

Once she learned to extract target, quantities and relationships, she could use the Mathematics she already knew.

Mira’s strength was not that she read slowly.

She read selectively.

She noticed the words and markings that changed the legal structure and ignored decorative detail until it became relevant.

Ryan learned that speed without parsing is expensive.

A five-second condition scan often saved minutes of wrong-route work.

The complete system is:

ORIENT TO THE TASK → IDENTIFY THE TARGET OBJECT → SORT GIVEN INFORMATION → MARK CONDITIONS AND CONSTRAINTS → ATTACH UNITS AND DOMAINS → IDENTIFY RELATIONSHIPS → DEFINE VARIABLES IF USEFUL → TRANSLATE WORDS, TABLES, GRAPHS OR DIAGRAMS INTO A MATHEMATICAL REPRESENTATION → STATE THE ANSWER CONTRACT → CHECK WHETHER THE INFORMATION IS SUFFICIENT → HAND THE STATE TO METHOD SELECTION AND PROBLEM SOLVING.

This page stops there deliberately.

How to Solve Unseen Maths Problems Independently owns what happens when no route cue is obvious.

How to Solve Math Problems owns the full representation-route-execution-recovery-verification system.

How to Show Working in Maths owns the visible route after translation.

How to Improve Maths Accuracy owns prospective error prevention.

How to Check Maths Answers owns verification.

The question-reading owner has one job:

make sure the learner is solving the mathematical problem that was actually presented, not the problem they expected to see.

Public references: What Works Clearinghouse — Improving Mathematical Problem Solving in Grades 4 Through 8; NCETM — Five Big Ideas in Teaching for Mastery.

Appendix A — Reading and translation checkpoint: twenty-four cases

This checkpoint is original instructional material. It is not a standardised assessment and has no validated cut score. The useful evidence is where interpretation first breaks.

For each case, use:

Target → Key givens → Conditions → Translation → Answer contract → Reading risk.

Task 1 — “More than” direction

Question: A number A is 7 more than a number B. Write A in terms of B.

Target: an expression for A.

Key relationship: A is being described relative to B.

Translation: A = B + 7.

Reading risk: writing B = A + 7 because B appears later in the sentence.

Check: choose B = 10. Then A should be 17. The expression A = B + 7 matches the words.

Task 2 — “Less than” with nested multiplication

Question: The length L is 3 less than twice the width w. Write L in terms of w.

Target: expression for L.

Translation: first twice the width = 2w, then 3 less = 2w − 3.

Answer: L = 2w − 3.

Reading risk: L = 2(w − 3), which means twice a quantity that is already 3 less.

Transfer: compare with “L is twice the amount that is 3 less than w”, which would be 2(w − 3).

Task 3 — Reverse percentage

Question: After a 20% discount, a jacket costs 72 dollars. Find the original price.

Target: original price.

Given: final price = 72; discount = 20%.

Condition: final is 80% of original.

Translation: 0.8P = 72.

Reading risk: calculating 20% of 72 or multiplying 72 by 0.8.

Answer contract: price in dollars.

Check: original 90 gives 20% discount 18, final 72.

Task 4 — Percentage points versus percent increase

Question: A pass rate rises from 50% to 65%. State the increase in percentage points and the relative percentage increase.

Target: two different measures.

Percentage-point increase: 65 − 50 = 15 percentage points.

Relative percentage increase: 15/50 × 100% = 30%.

Reading risk: reporting 15% for both.

Lesson: the reference base differs between the two descriptions.

Task 5 — At least versus more than

Question: A ride requires passengers to be at least 1.4 m tall. Write an inequality for height h.

Target: allowed-height set.

Translation: h ≥ 1.4 m.

Reading risk: h > 1.4, which wrongly excludes exactly 1.4 m.

Boundary check: would a person exactly 1.4 m tall be allowed? Yes.

Task 6 — No more than

Question: A lift can carry no more than 900 kg. Write the constraint on total load W.

Translation: W ≤ 900 kg.

Reading risk: W < 900, which excludes the stated maximum.

Role: this may be a constraint inside a larger optimisation or counting problem, not the main target.

Task 7 — Consecutive integers

Question: The sum of two consecutive positive integers is 41. Form an equation.

Target: a model, not necessarily the final integers yet.

Variable: let smaller integer = n.

Condition: next integer = n + 1; n positive integer.

Translation: n + (n + 1) = 41.

Reading risk: x + y = 41 without encoding consecutiveness.

Task 8 — Consecutive even integers

Question: The product of two consecutive even positive integers is 168. Form a model.

Target: model preserving evenness and consecutiveness in the even sequence.

One representation: let smaller even integer = 2n, next even integer = 2n + 2.

Translation: (2n)(2n + 2) = 168.

Alternative: let smaller even integer = x with x even, then x(x + 2) = 168.

Reading risk: x(x + 1), which models ordinary consecutive integers instead.

Task 9 — Ratio correspondence

Question: Red:blue counters are in the ratio 3:5. There are 45 red counters. How many blue counters?

Target: blue count.

Correspondence: 3 parts red = 45.

One part: 15.

Blue: 5 × 15 = 75.

Reading risk: multiplying 45 by 3/5 because the ratio order is not mapped to quantities.

Task 10 — Part-to-total rather than part-to-part

Question: Red counters are 3/5 of all counters. There are 30 red counters. How many counters altogether?

Target: total.

Reference: 30 represents 3/5 of total, not 3/5 of blue.

Translation: (3/5)T = 30.

Answer: T = 50.

Reading risk: treating 3:5 as red:blue and obtaining a different total.

Task 11 — Fixed fee plus rate

Question: A service charges 8 dollars plus 2.50 dollars per hour. Write the total cost C for h hours.

Target: cost model.

Given: fixed fee 8; rate 2.50 dollars/hour.

Translation: C = 8 + 2.5h.

Reading risk: C = 8h + 2.5 or C = 10.5h.

Near neighbour: a pure 2.50 dollars per hour charge with no fixed fee would be direct proportion C = 2.5h.

Task 12 — Average speed

Question: A cyclist travels 30 km at 15 km/h, then 30 km at 30 km/h. Find the average speed for the entire journey.

Target: total distance divided by total time.

Distance: 60 km.

Time: 30/15 + 30/30 = 2 + 1 = 3 h.

Average speed: 60/3 = 20 km/h.

Reading risk: simple average (15 + 30)/2 = 22.5, which would only be justified under equal-time conditions, not equal-distance conditions.

Translation lesson: “average speed” invokes the definition total distance/total time, not “average the displayed speed numbers”.

Task 13 — Without replacement

Question: A bag contains 5 red and 4 blue counters. Two counters are drawn without replacement. Write the probability of drawing two reds.

Target: probability of the event red then red.

Given: 5 red, 4 blue, total 9.

Condition: without replacement means the state changes after the first draw.

Translation: P(RR) = 5/9 × 4/8.

Reading risk: 5/9 × 5/9, which silently treats the second draw as if the first counter were returned.

Answer contract: probability between 0 and 1; simplify if appropriate.

Transfer: if the wording changes to “with replacement”, the second factor returns to 5/9. One phrase changes the state model.

Task 14 — At least one

Question: The probability of success on each of three independent trials is 0.4. Find the probability of at least one success.

Target: event containing one, two or three successes.

Condition: trials independent; success probability remains 0.4.

Useful translation: P(at least one) = 1 − P(no successes).

Calculation state: P(no success) = 0.6³.

Reading risk: interpreting “at least one” as exactly one.

Answer contract: probability, so 1 − 0.216 = 0.784.

Transfer: “at most one” would mean zero or one success, a different event.

Task 15 — Diagram not drawn to scale

Question: A triangle is shown with side lengths 7 cm and 10 cm. The diagram looks right-angled but states “not drawn to scale”. No right angle is marked. Can Pythagoras be used from the diagram alone?

Target: determine whether a method is authorised, not calculate a side.

Given: two side lengths.

Condition: no right angle is stated or marked.

Translation: visual appearance is not mathematical evidence.

Answer: no, not from the diagram alone.

Reading risk: assuming that a familiar visual layout supplies a condition the question never gave.

Transfer: add a right-angle mark and the method status changes immediately.

Task 16 — Graph scale before gradient

Question: A graph has x-axis labels 0, 5, 10, 15 and y-axis labels 0, 2, 4, 6. A line passes through the labelled points (5,2) and (15,6). Find its gradient.

Target: gradient.

Representation reading: coordinates come from axis values, not square counts.

Translation: m = (6 − 2)/(15 − 5).

Answer: 4/10 = 0.4.

Reading risk: counting grid intervals as 1 unit each and obtaining a different gradient.

Check: the axes and units must be read before applying the formula.

Task 17 — Distance-time versus speed-time

Question A: On a distance-time graph, a segment is horizontal from t = 4 to t = 7. What does it mean?

Translation: distance is unchanged while time passes, so the object is stationary.

Question B: On a speed-time graph, a segment is horizontal at 12 m/s from t = 4 to t = 7. What does it mean?

Translation: speed remains constant at 12 m/s; the object is moving.

Reading risk: interpreting graph shape without reading the vertical-axis variable.

Lesson: the same geometric shape can encode different physical meanings because the axes define the representation.

Task 18 — Estimated mean from grouped data

Question: A grouped frequency table gives intervals 0–10, 10–20 and 20–30 with frequencies 4, 7 and 9. Estimate the mean.

Target: estimated mean, not exact mean.

Representation reading: individual values within each interval are unknown.

Translation: use class midpoints 5, 15 and 25 in the standard school estimate.

Calculation state: estimated total = 4(5) + 7(15) + 9(25).

Reading risk: reporting the result as exact, or treating interval endpoints as actual data values without justification.

Answer contract: an estimate, with suitable numerical precision for the context.

Task 19 — Exact versus decimal form

Question: Solve x² = 18 and give the positive solution exactly.

Target: positive exact solution.

Translation: x = √18 = 3√2.

Reading risk: giving 4.24 because a calculator decimal feels more complete.

Answer contract: exact form; no decimal approximation is required.

Near neighbour: if the question asks for 3 significant figures, x ≈ 4.24 would then be appropriate.

Task 20 — Hence as a handoff

Part (a): Show that the model leads to x² − 7x + 12 = 0.

Part (b): Hence find the possible values of x.

Target: solve the established quadratic.

Handoff: x² − 7x + 12 = 0 from part (a).

Translation: factorise (x − 3)(x − 4) = 0.

Answer: x = 3 or 4, subject to any original contextual restrictions.

Reading risk: rebuilding the model unnecessarily, or forgetting restrictions inherited from part (a).

Lesson: a prior result carries provenance and conditions, not just a number or expression.

Task 21 — Irrelevant datum

Question: A rectangular floor is 9 m long and 6 m wide. The room contains 24 chairs. Find the floor area.

Target: area.

Needed givens: length 9 m, width 6 m.

Irrelevant for this target: 24 chairs.

Translation: A = 9 × 6 = 54 m².

Reading risk: assuming every number must appear in the calculation.

Transfer: if the target became “chairs per square metre”, 24 would become essential.

Task 22 — Insufficient information

Question: A rectangle has perimeter 30 cm. Find its area.

Target: area.

Given relationship: 2L + 2W = 30, so L + W = 15.

Reading state: many positive length-width pairs satisfy this.

Conclusion: area is not uniquely determined without another condition.

Examples: 7 cm by 8 cm gives 56 cm²; 5 cm by 10 cm gives 50 cm².

Reading risk: assuming a square or inventing equal sides because one exact area is expected.

Lesson: a well-read problem can legitimately end at “insufficient information”.

Task 23 — Compare distributions

Question: Distribution A has median 52 and interquartile range 8. Distribution B has median 47 and interquartile range 15. Compare the distributions.

Target: a relational comparison, not two isolated calculations.

Reading state: centre and spread information are already given.

Translation: A has the higher median; B has the larger interquartile range.

Possible response: A has a higher typical central value by the median, while B is more spread out by the IQR.

Reading risk: reporting only “52 and 47” or “8 and 15” without comparison.

Answer contract: words that connect the two distributions using the supplied evidence.

Task 24 — Read the target before the familiar object

Question: A rectangle has length 12 cm and width 8 cm. The diagonal is drawn. Find the length of the diagonal.

Target: diagonal length.

Given: perpendicular rectangle sides 12 cm and 8 cm.

Representation: diagonal forms a right triangle with the sides.

Translation: d² = 12² + 8².

Answer: d = √208 = 4√13 cm, or an appropriate decimal if requested.

Reading risk: seeing “rectangle” and automatically calculating perimeter or area because those are more familiar rectangle tasks.

Lesson: object recognition should never replace target recognition.

How to interpret the checkpoint

Do not add the twenty-four tasks into one reading score. The important result is the first translation responsibility that repeatedly fails.

Target failure: the learner solves a related but different problem.

Reference failure: the learner attaches a percentage, ratio or comparison to the wrong base quantity.

Condition failure: a small phrase or diagram mark that controls method legality disappears.

Boundary failure: strict versus inclusive inequalities are confused.

Representation failure: tables, graphs or diagrams are read from appearance rather than labels and structure.

Unit failure: quantity type or conversion is lost.

Precision failure: exact, estimated and rounded states are confused.

Relevance failure: all printed information is forced into the model.

Sufficiency failure: the learner assumes one answer must exist even when information is insufficient or inconsistent.

Answer-contract failure: the learner performs correct Mathematics but does not provide the requested object, form or comparison.

The next practice block should isolate that reading responsibility rather than adding more full-solution volume indiscriminately.

Appendix B — Question-reading routing matrix: symptom → likely translation failure → smallest repair → next test

This matrix turns vague advice such as “read carefully” into an observable instructional decision. It is not a validated diagnostic scale. Start from the learner’s first wrong interpretation, not from the final wrong answer. A late arithmetic error and an early reading error need different repairs.

Symptom: the learner gives a correct calculation for the wrong target

Likely failure: target substitution.

The learner recognises a familiar object—rectangle, graph, percentage, quadratic—and performs the most familiar operation instead of the requested one.

Smallest repair: before calculation, complete one sentence: “My final answer must be a ___.”

Example: “fencing around the garden” → perimeter, not area.

Next test: present three questions using the same diagram but asking for three different targets. Require only the target object, not the solution.

Advance when: the learner changes method intention when the target changes, even though the visible object remains the same.

Symptom: the learner starts calculating as soon as a number appears

Likely failure: premature operation selection.

Smallest repair: prohibit calculation for the first thirty seconds on selected practice questions. Require target, quantity names and relationship first.

Example: write “final = 80% of original” before 72 ÷ 0.8.

Next test: use a word problem containing several numbers, including one irrelevant number. Ask the learner to label each number’s job before any arithmetic.

Advance when: arithmetic begins from a stated relationship rather than from number proximity.

Symptom: every printed number is forced into the solution

Likely failure: relevance discrimination.

Smallest repair: for each used datum, ask “What mathematical relationship or constraint gives this number a job?”

Next test: create paired questions where the same datum is irrelevant in one target and essential in another.

Advance when: the learner can leave true information unused without anxiety and can explain why.

Symptom: “more than” and “less than” relationships are reversed

Likely failure: comparative reference direction.

Smallest repair: identify the described quantity and reference quantity separately.

Example: “A is 7 less than B” → start from B, subtract 7, so A = B − 7.

Next test: use several synonymous phrasings: “exceeds by”, “is lower than by”, “is 4 older than”, “is 5 shorter than”.

Advance when: the learner can create a quick numerical example to test the direction independently.

Symptom: brackets are wrong in translated expressions

Likely failure: nested-language parsing.

Smallest repair: translate one clause at a time and say the relationship in simpler natural language before symbolising.

Contrast: “three more than twice x” = 2x + 3; “twice the quantity three more than x” = 2(x + 3).

Next test: ask only for expressions from pairs with the same nouns/numbers but different grammatical grouping.

Advance when: bracket placement changes correctly with sentence structure.

Symptom: ratios are applied to the wrong quantity

Likely failure: correspondence or reference-set failure.

Smallest repair: write each ratio term beside its quantity before using a scale factor.

Example: red:blue = 3:5; 45 red means 3 parts = 45, not 5 parts.

Next test: contrast part-to-part with part-to-total language.

Advance when: the learner can state “fraction/ratio of what?” before calculating.

Symptom: percentage arithmetic is fluent but reverse-percentage questions fail

Likely failure: 100% base and direction.

Smallest repair: write “100% = ___” and “final = multiplier × original”.

Next test: mix forward percentage, reverse percentage, percentage increase and percentage-point change.

Advance when: the learner chooses the base from meaning rather than from the most visually prominent number.

Symptom: “at least” and “more than” are treated as identical

Likely failure: boundary inclusion.

Smallest repair: test the boundary value against the sentence.

Example: does exactly 5 satisfy “at least 5”? Yes. Does it satisfy “more than 5”? No.

Next test: alternate at least, at most, more than, less than, no more than and no fewer than with integer and continuous quantities.

Advance when: inclusive versus strict inequality symbols are selected correctly without keyword memorisation alone.

Symptom: the learner sees a triangle and chooses a theorem from appearance

Likely failure: condition authority.

Smallest repair: split diagram information into given, derived and looks-like.

Next test: include misleading diagrams where visual appearance contradicts or fails to establish the required condition.

Advance when: the learner can state the exact condition that makes the theorem legal before using it.

Symptom: the learner reads graph shape but misses scale or axes

Likely failure: representation semantics.

Smallest repair: enforce axes → units → scale → key/legend before shape.

Next test: show two graphs with identical visual shapes but different axis meanings or scales.

Advance when: interpretations name the variables and units, not just “increasing”, “flat” or “steep”.

Symptom: a table pattern is inferred incorrectly

Likely failure: row/column role or baseline reading.

Smallest repair: name each row/column variable and inspect the value at a meaningful reference such as x = 0 where available.

Next test: alternate function tables, frequency tables and two-way tables so the learner cannot assume every table has the same structure.

Advance when: formula selection follows table type rather than visual layout.

Symptom: a correct formula is used with incompatible units

Likely failure: unit-state parsing.

Smallest repair: predict the final unit and convert inputs before substitution.

Next test: mix minutes/seconds, metres/kilometres, area/length and rates with compatible and incompatible units.

Advance when: the learner detects a mismatch before numerical calculation.

Symptom: the learner gives a decimal when an exact form is requested

Likely failure: answer-contract precision.

Smallest repair: mark exact/estimate/decimal-place/significant-figure instruction at the start.

Next test: use the same underlying value with three different answer contracts: exact, 3 s.f., graph estimate.

Advance when: the mathematical route can stay the same while the final form changes appropriately.

Symptom: the learner solves part (b) from scratch despite “hence”

Likely failure: linked-part handoff recognition.

Smallest repair: write the relevant result from part (a) at the top of part (b) with its condition/precision.

Next test: use a three-part problem where each part consumes a previous result.

Advance when: the learner reuses established structure without unnecessary rebuilding and preserves inherited restrictions.

Symptom: the learner assumes every problem must have one numerical answer

Likely failure: sufficiency and solution-type judgement.

Smallest repair: ask “Do these givens determine the target uniquely?” before solving selected problems.

Next test: mix unique-solution, multiple-solution, no-solution and underdetermined questions.

Advance when: “insufficient information” or a range is accepted as a legitimate mathematical conclusion when justified.

Symptom: the learner understands after an adult paraphrases every question

Likely failure: translation remains externally owned.

Smallest repair: reduce adult support from full paraphrase to one prompt: target, relationship, reference base or condition.

Next test: a changed question with the same structure and no paraphrase.

Advance when: support shrinks from full restatement to a cue to none.

Symptom: the learner repeatedly asks “what does this word mean?”

Likely failure: vocabulary knowledge rather than deeper parsing.

Smallest repair: teach the term explicitly with an example and non-example.

Next test: reuse the term in a new sentence and representation.

Advance when: the learner can explain the term’s mathematical effect, not merely repeat a dictionary definition.

Symptom: the learner parses the question correctly but still cannot proceed

Likely state: reading may no longer be the bottleneck.

Smallest repair: route the problem onward: missing content, method selection, execution, fluency or unseen-problem independence.

Next test: ask the learner to state the translated state clearly. If it is correct, stop re-teaching reading and address the next layer.

Principle: a strong diagnostic system knows when to leave its own lane.

Symptom: the learner reads well untimed but misreads under a clock

Likely failure: reading controls are not yet cheap enough under load.

Smallest repair: preserve a short high-value scan: target, condition, unit, answer contract.

Next test: timed parsing-only sets before full timed solving.

Advance when: speed increases without a rise in target/condition/base errors.

Symptom: the learner rereads the entire question after every step

Likely failure: the question state has not been externalised.

Smallest repair: write a compact representation: variable definitions, relationship, target cue or labelled diagram.

Next test: once the representation is written, ask the learner to solve without rereading until a check or conflict requires it.

Advance when: rereading becomes selective rather than total.

Symptom: the learner treats AI’s paraphrase as authoritative even when it changes the problem

Likely failure: tool output has replaced source-text authority.

Smallest repair: compare the paraphrase clause by clause with the original: target, quantities, conditions and answer contract.

Next test: ask AI for two paraphrases and have the learner identify whether each preserves the same mathematical state.

Advance when: the learner can reject a convenient paraphrase that loses a condition.

Symptom: a second-language learner knows the Mathematics but repeatedly stalls on ordinary vocabulary

Likely failure: language-access gap rather than mathematical relationship knowledge.

Smallest repair: clarify the ordinary-language vocabulary while preserving the original mathematical terms and symbols.

Next test: give the same mathematical structure in several phrasings, including more direct wording, and observe whether translation becomes independent.

Advance when: vocabulary support can fade without changing the mathematical demand.

Symptom: the learner’s translated equation works numerically on one example but does not match the sentence generally

Likely failure: accidental fit rather than structural translation.

Smallest repair: test the translation with a different hypothetical value.

Example: if “A is 20% more than B” was translated as A = B + 20, choose B = 200 and compare the sentence meaning with the equation.

Next test: require one changed-value check on comparative/percentage translations before full solving.

Advance when: the relationship remains valid across plausible inputs.

How to use the routing matrix

Do not make the learner carry this full matrix forever. Its job is to make “careful reading” concrete enough to teach.

The compact internal version is:

What exactly is the target?

Which information has a mathematical job?

What conditions make methods legal?

What is the reference quantity?

What do the units and representation say?

What must the final answer look like?

Is there enough information?

Once these questions happen quickly and reliably, the matrix has completed its job.

Appendix C — Eight composite question-reading laboratories

These laboratories combine several reading responsibilities at once. The purpose is not to test difficult calculation. It is to show how target, wording, conditions, representations, units and answer contracts interact before method execution begins.

Laboratory 1 — The subscription plans: fixed fee, rate, equality and whole-number interpretation

Question:

Plan A charges 12 dollars per month plus 3 dollars for each unit of usage. Plan B charges 5 dollars per month plus 4.50 dollars for each unit. At what usage do the two mathematical cost models give the same cost? If usage must be a whole number, explain what the result means.

First reading pass — target.

The target is not the cost of either plan at a stated usage. It is the usage value at which the two costs are equal.

Second pass — quantities.

Usage = u.

Plan A cost = 12 + 3u.

Plan B cost = 5 + 4.5u.

Relationship phrase.

“Give the same cost” means set the two cost expressions equal:

12 + 3u = 5 + 4.5u.

The question-reading job is now largely complete. The algebra layer can solve:

7 = 1.5u

u = 14/3 ≈ 4.67.

Answer contract.

The question deliberately asks for interpretation if usage must be a whole number. Therefore 4.67 is the algebraic crossing of the continuous models, but no whole-number usage gives exact equality.

At 4 units:

A = 24; B = 23.

At 5 units:

A = 27; B = 27.5.

The cheaper plan changes between those whole-number states.

Reading risks:

treating 3 and 4.5 as total charges rather than rates;

forgetting fixed fees;

solving one plan rather than comparing both;

rounding 4.67 to 5 and claiming exact equality without checking the whole-number condition.

Transfer variation: change “same cost” to “Plan A costs no more than Plan B”. The equality becomes an inequality, and the target becomes a set of usage values rather than one crossing point.

Laboratory 2 — The marked-price chain: two percentages, two bases and one misleading shortcut

Question:

A shop buys an item for 80 dollars. It marks the item up by 25% from cost price. During a sale, the marked price is reduced by 20%. Find the sale price and state whether the two percentage changes cancel.

Target.

Sale price and an interpretation of whether the changes cancel.

Stage map.

Cost price = 80.

Markup is 25% of cost price, so marked price:

80 × 1.25 = 100.

Discount is 20% of marked price, so sale price:

100 × 0.8 = 80.

Why this is a reading problem.

A learner may see +25% and −20% and combine them as +5%. That would treat the percentage changes as if they shared one base. They do not.

The first percentage acts on 80.

The second acts on 100.

In this particular numerical case, the multiplicative factors 1.25 and 0.8 multiply to 1, so the final price happens to equal the cost price.

Answer contract.

The learner must answer both parts: 80 dollars, and yes—the changes cancel here multiplicatively because 1.25 × 0.8 = 1, not because 25 − 20 = 0.

Reading risks:

using 20% of 80 for the discount;

subtracting percentage points without reference to the changing base;

stopping after finding the marked price;

reporting the coincidental return to 80 without explaining why.

Transfer variation: change the markup to 20% and discount to 20%. The final multiplier becomes 1.2 × 0.8 = 0.96, proving that equal percentage increase/decrease normally do not cancel.

Laboratory 3 — The delivery graph: identical visual shape, different mathematical meaning

Question:

A graph plots total delivery cost C dollars against distance d kilometres. The straight line passes through (0, 6) and (10, 26). Interpret the intercept and gradient, then write the cost model.

Read the axes before the line.

Horizontal: distance d in kilometres.

Vertical: total cost C in dollars.

The point (0,6) therefore means a 6-dollar cost even at zero kilometres. This is naturally interpreted as a fixed charge within the model.

Gradient.

(26 − 6)/(10 − 0) = 2 dollars/km.

Model.

C = 6 + 2d.

Reading risks:

calling the intercept 6 kilometres;

calling the gradient “2” without units or meaning;

assuming direct proportion because the graph is a straight line;

ignoring the nonzero intercept.

Why the visual shape is insufficient.

If the same line shape appeared on a temperature-time graph, the intercept and gradient would have different units and meanings. Representation semantics come from axes, units and context.

Answer contract.

The question asks for interpretation and an equation, so a complete answer should say:

fixed charge = 6 dollars;

distance charge = 2 dollars per kilometre;

C = 6 + 2d.

Transfer variation: change the intercept to zero. The model becomes C = 2d and is directly proportional to distance under the standard school definition.

Laboratory 4 — The rotated geometry diagram: condition, correspondence and target dimension

Question:

Two triangles are stated to be similar. Triangle ABC has sides AB = 6 cm and AC = 9 cm. In the second triangle, corresponding side DE = 10 cm. The second triangle is drawn rotated, and the stated correspondence is A↔D, B↔E, C↔F. Find DF.

Target.

DF, a length.

Condition.

Similarity is explicitly stated, so proportional corresponding sides are authorised.

Correspondence.

AB ↔ DE.

AC ↔ DF.

Scale factor from first triangle to second:

DE/AB = 10/6 = 5/3.

So:

DF = 9 × 5/3 = 15 cm.

Reading risks:

matching sides by page position instead of stated vertex correspondence;

using 10/9 because visually the sides appear adjacent;

treating a rotated diagram as a different structure;

squaring the scale factor even though the target is a length rather than area.

Transfer variation: ask instead for the area ratio. Now the linear scale 5/3 must be squared, giving 25/9. The target dimension changes the mathematical relationship even though the diagram is unchanged.

Laboratory 5 — The two-way table: denominator choice changes with the wording

A survey table records 100 students by year group and transport mode. There are 40 Secondary 1 students, of whom 18 travel by bus. There are 60 Secondary 2 students, of whom 24 travel by bus.

Question A: What is the probability that a randomly chosen student travels by bus?

Target population: all 100 students.

Total bus = 18 + 24 = 42.

P(bus) = 42/100 = 0.42.

Question B: Given that the chosen student is in Secondary 1, what is the probability they travel by bus?

Condition: “given Secondary 1” changes the reference set.

Denominator = 40 Secondary 1 students.

P(bus | Secondary 1) = 18/40 = 0.45.

Question C: What is the probability that a randomly chosen student is both Secondary 1 and travels by bus?

Target event: joint event inside the full population.

P(Secondary 1 and bus) = 18/100 = 0.18.

Reading lesson.

The same cell value 18 participates in three different fractions because the wording changes the reference population.

Reading risks:

using 18/100 for the conditional question;

using 42/100 for the joint question;

ignoring “given that” as decorative wording.

Transfer variation: reverse the conditioning: given bus, what is the probability of Secondary 1? The denominator becomes total bus travellers, 42, so the probability is 18/42.

Laboratory 6 — The linked multi-part model: preserve earlier results and their status

Part (a): A tank begins with 18 litres and fills at 4.5 litres per minute. Write a model for volume V after t minutes.

Reading: initial + rate × time.

V = 18 + 4.5t.

Part (b): Hence find when the tank reaches 72 litres.

Handoff: use the model from (a).

72 = 18 + 4.5t.

54 = 4.5t.

t = 12 minutes.

Part (c): The tank has maximum capacity 60 litres. Explain what must change about the model.

New condition: capacity contradicts unrestricted continuation to 72 litres.

The linear filling expression can describe the filling phase only until:

18 + 4.5t = 60.

4.5t = 42.

t = 28/3 ≈ 9.33 minutes.

After that, either filling stops, overflow occurs, or another physical rule must be specified.

Reading lesson.

Part (b)’s algebra is valid within the original simplified model, but part (c) adds a condition that changes the model’s physical domain.

Reading risks:

treating earlier equations as eternally valid after conditions change;

ignoring “hence” and re-deriving part (a);

treating capacity as an irrelevant number.

Transfer variation: state the capacity in part (a). The valid domain should then be carried from the start.

Laboratory 7 — The underdetermined rectangle: resisting the expectation of one neat number

Question:

A rectangle has perimeter 40 cm. Its length is greater than its width. Determine its area.

Target.

Area.

Given relationship.

2L + 2W = 40, so L + W = 20.

Condition.

L > W.

Does the state determine one area?

No.

Examples:

L = 11, W = 9 → area 99 cm².

L = 12, W = 8 → area 96 cm².

L = 15, W = 5 → area 75 cm².

All satisfy the perimeter and L > W.

Reading conclusion.

The problem is underdetermined if it asks for one exact area. Additional information is required.

Reading risk: assuming the rectangle is a square because that gives a clean answer, even though L > W explicitly rules out a square.

Higher-level extension: if the task instead asks for the greatest possible area among real positive side lengths subject to perimeter 40, the square would maximise area—but that violates the strict condition L > W if equality is disallowed, producing a supremum approached but not attained. The exact interpretation depends on the course level and wording.

Core lesson: do not let the expectation that “textbook questions have answers” override the mathematical information actually given.

Laboratory 8 — The examination instruction stack: target, precision, unit and explanation in one sentence

Question:

“Using the graph, estimate the time, in seconds, at which the two models have equal output, giving your answer to the nearest whole second, and explain what the intersection represents in context.”

This sentence contains several contracts.

Representation instruction: use the graph.

Target: time when outputs are equal.

Unit: seconds.

Status: estimate, because the value is read graphically.

Precision: nearest whole second.

Interpretation: explain the intersection in context.

A learner who reads only “find the intersection” may produce a coordinate pair and stop.

A complete answer might be:

t ≈ 37 s.

At about 37 seconds, the two models predict the same output.

Reading risks:

reporting both coordinates when only time is requested;

giving 37.428 s with false precision from a graph;

omitting seconds;

omitting the contextual explanation;

solving algebraically despite the explicit instruction “using the graph” where the assessment intends graphical reading.

Transfer variation: change “using the graph” to “solve algebraically”. The target may be the same, but the permitted or intended representation changes.

What the eight laboratories reveal

Question reading is not one isolated skill. It is a stack of state-preservation decisions.

In a tariff problem, the reader preserves fixed and variable components.

In percentage chains, the base changes between stages.

In graphs, axes and units define meaning.

In geometry, conditions and correspondence authorise ratios.

In tables, wording chooses the denominator population.

In linked parts, earlier results carry provenance and limitations.

In underdetermined problems, the reader must resist inventing conditions.

In examination instructions, the answer contract can contain representation, unit, precision and explanation requirements simultaneously.

The common principle is:

before solving, create a mathematical state that preserves every piece of information capable of changing the legal route or the required final answer.

Appendix D — The compact question-reading sheet

This scaffold is for learners who currently lose the problem before the Mathematics begins. It should fade as parsing becomes automatic.

1. Target

Write:

“Final answer must be…”

Examples: a length, an equation, all possible x-values, a probability, an estimated mean, a comparison, a proof.

2. Quantities

Name the important quantities and attach units where useful.

Do not calculate yet.

3. Conditions

Mark phrases or diagram evidence that change method legality or the solution set:

right-angled;

parallel;

positive integer;

without replacement;

at least;

exact;

not drawn to scale;

domain restrictions.

4. Reference

For ratios, percentages and comparisons, ask:

of what?

compared with what?

what is 100%?

which ratio term belongs to which quantity?

5. Relationship

State one sentence or symbolic relationship before doing arithmetic.

Examples:

final = 0.85 × original;

total cost = fixed fee + rate × usage;

larger = smaller + 7;

combined mean = combined total / combined count.

6. Representation

Choose a useful form:

equation;

inequality;

table;

graph;

labelled diagram;

tree;

ratio map;

stage map.

7. Sufficiency

Ask:

Does the information determine the target?

Are there hidden assumptions I am inventing?

Could several answers satisfy the same givens?

8. Answer contract

Record only what matters:

unit;

precision;

exact/estimated;

one value/all values;

explanation/proof/comparison;

required representation such as “use the graph”.

9. Handoff

Once the question state is stable, stop parsing and hand the problem to the next system:

method selection;

problem solving;

execution;

working;

checking.

This prevents over-reading. Reading is complete when the mathematical state is clear enough to act on.

How to fade the sheet

Start with all nine prompts only if the learner genuinely needs them.

Then reduce to:

Target → Conditions → Relationship → Answer contract.

Later reduce to a silent scan.

The scaffold has succeeded when the learner no longer needs to see it but still preserves the same information internally.

Appendix E — Reading receipts: short records that change the next practice decision

A receipt is useful only if it identifies a repeatable reading mechanism. Do not log every question forever.

Target error: “Calculated rectangle area when question asked for fencing. Next set: same diagram, three targets.”

Reference-base error: “Reverse percentage treated final price as 100%. Next set: write 100% base before arithmetic.”

Boundary error: “‘At least’ translated as >. Next set: boundary-value test for inclusive/strict phrases.”

Diagram error: “Used Pythagoras from appearance; no right-angle evidence. Next set: given/derived/looks-like classification.”

Graph error: “Gradient arithmetic correct, x-scale misread. Next set: axes and scale annotation before formula.”

Relevance error: “Forced seat-count datum into area problem. Next set: assign job to each used number.”

Linked-part error: “Recomputed model despite ‘hence’. Next set: write prior result + condition at start of next part.”

Successful transfer: “Fixed-fee model recognised in unfamiliar storage context; no keyword cue needed. Move to mixed representations.”

Reading lane complete: “Question translated correctly; stall began at method selection. Route to unseen/problem-solving practice rather than more paraphrasing.”

The final receipt is often the most important. Good diagnosis should prevent question-reading practice from swallowing problems that belong to another layer.

Reading has done its job when the learner can hand a faithful mathematical state to the solver without needing another person to rewrite the question first.