There is a small thing I sometimes do after a student has solved a Mathematics question correctly.
I change one number.
Not the whole question.
Not the topic.
Not the method deliberately.
Just one part.
A 5 becomes a 6.
A positive sign becomes negative.
A right angle becomes an ordinary angle.
A repeated root becomes two distinct roots.
A scale factor of 2 becomes 3.
Then I ask:
“What changes?”
The student occasionally looks puzzled.
“But I already got the question right.”
She did.
That is exactly why the next question interests me.
A correct solution tells me that the student could solve that particular mathematical object under those particular conditions.
Changing one small feature lets me see something else.
Does she know which part of her reasoning depended on it?
Can she predict the direction of change before recalculating?
Does the same method still apply?
Does the answer merely change in size, or does the entire structure of the problem change?
Sometimes everything survives.
Sometimes one line needs adjusting.
Sometimes a tiny change makes the old method invalid.
After many years of teaching Mathematics, I have found these small perturbations unusually revealing.
They test something that another fresh worksheet does not always expose.
The student has already shown me that she can reach an answer.
Now I want to know how well she understands why that answer had the shape it did.
The direct answer
A student understands Mathematics more deeply when she can reason about how a solution should respond to a small change in the problem.
This is different from simply re-solving another similar question.
If I change one feature while holding most of the problem steady, the student can compare two nearby mathematical worlds.
She can ask:
What stayed the same?
What changed?
Which part of the method depended on the altered condition?
Should the new answer be larger or smaller?
Should there still be two solutions?
Does a formula still apply?
Has the problem merely changed numerically, or has it changed mathematically?
Those questions reveal structure.
They also reveal fragile learning very quickly.
A student who has memorised a route may solve the original question perfectly and then become lost when one apparently minor detail changes.
A student who understands the relationships underneath often has a prediction before touching the calculator.
That prediction is valuable.
It tells me the Mathematics is no longer attached only to one completed example.
Start with something very simple
Suppose a student solves:
3x + 5 = 20.
She writes:
3x = 15
so:
x = 5.
Correct.
Now I change only the 20:
3x + 5 = 23.
Before solving, I ask:
“Will x be larger or smaller than 5?”
It should be larger.
The left side must now reach a larger value while the coefficient and constant remain unchanged.
Solve:
3x = 18
so:
x = 6.
Nothing dramatic happened.
But the student has done more than execute algebra.
She has reasoned about the relationship between the equation and its solution.
Now change:
3x + 8 = 23.
Compared with:
3x + 5 = 23.
Should x increase or decrease?
The constant term has increased while the right-hand side stayed fixed.
Therefore x must decrease.
Indeed:
3x = 15
so:
x = 5.
The arithmetic is elementary.
The thinking is not entirely elementary.
The student is beginning to see equations as systems whose parts influence one another.
Prediction matters because calculation can hide weak judgement
If I simply give the second equation and ask for x, the student may calculate correctly.
I learn that she can solve linear equations.
Good.
But when I ask for a prediction first, I learn whether she has an expectation.
That expectation gives the later answer something to be checked against.
Suppose she predicts that x should decrease but obtains x = 9.
Now something feels wrong before I say anything.
The student has an internal contradiction.
This is useful.
Mathematical independence improves when students have reasons to distrust an answer themselves.
Without expectation, almost any calculator display can feel plausible.
With expectation, calculation becomes accountable to reasoning.
Percentages are especially good for this
Suppose an item costs $200 and receives a 10% discount.
Discount:
0.10 × 200 = 20.
New price:
$180.
Now I change 10% to 20%.
I do not initially ask for the answer.
I ask:
“What should happen?”
The discount should become larger.
The final price should become lower.
Then:
0.20 × 200 = 40
so the new price is:
$160.
Fine.
Now change the original price from $200 to $400 while keeping the discount at 10%.
What should happen to the discount amount?
It should double.
What should happen to the final price?
It should also double:
$360.
These are simple examples of proportional structure.
The student is not only applying percentage procedures.
She is learning how the output responds when an input changes.
That is an important habit far beyond percentage questions.
But not every relationship responds linearly
This is where the exercise becomes more interesting.
Take the area of a circle:
A = πr².
Suppose r = 3.
Area:
9π.
Now double the radius to 6.
A student who has developed an overly broad “double input, double output” instinct may expect:
18π.
But:
A = π(6²) = 36π.
The radius doubled.
The area quadrupled.
This is precisely the kind of small change that reveals the governing structure.
The student sees that area depends on the square of the radius.
Now triple the radius.
Area becomes nine times as large.
The question has moved from procedure into behaviour.
That is where formulas become more meaningful.
Similarity gives the same lesson in geometry
Suppose a square has side length 4 cm.
Area:
16 cm².
Enlarge the square by scale factor 2.
Side length:
8 cm.
Area:
64 cm².
Again:
length ×2,
area ×4.
If the scale factor becomes 3:
length ×3,
area ×9.
Now take a cube.
Scale every length by 2.
Volume becomes:
2³ = 8
times as large.
This is why I like changing one number rather than immediately teaching another formula.
The student can see the consequence of the structure she already knows.
The exponent is no longer decorative notation.
It tells us how sensitive the output is to changes in the input.
A-Math contains many richer versions of this idea
Consider:
y = x².
At x = 2:
y = 4.
At x = 3:
y = 9.
Increasing x by 1 did not increase y by a constant 1.
The change was 5.
From x = 3 to x = 4:
y goes from 9 to 16.
Now the increase is 7.
A student looking only at points may notice this.
A student beginning calculus can interpret something deeper.
The rate of change itself is changing.
For:
y = x²
we have:
dy/dx = 2x.
The gradient at x = 2 is 4.
At x = 3 it is 6.
At x = 4 it is 8.
Small movement along the curve changes the gradient predictably.
This is a much stronger way to understand differentiation than simply remembering:
“Bring the power down and subtract one.”
Sometimes I change a sign instead of a number
Suppose:
y = x² + 4x + 3.
Factorise:
y = (x + 1)(x + 3).
Roots:
x = −1, −3.
Now change only one sign:
y = x² − 4x + 3.
Factorise:
y = (x − 1)(x − 3).
Roots:
x = 1, 3.
The graph’s relationship to the y-axis has changed.
The roots reflected from negative to positive.
The x-coordinate of the turning point also changes sign.
For the first quadratic:
axis of symmetry:
x = −2.
For the second:
x = 2.
A tiny symbolic alteration produces a highly structured change.
The student can begin to see coefficients as controlling features of the graph rather than merely supplying numbers for algebra.
A sign change can also change the method’s conditions
Consider:
x² + 4 = 0
over the real numbers.
There are no real solutions.
Change one sign:
x² − 4 = 0.
Now:
(x − 2)(x + 2) = 0
so:
x = ±2.
The questions look almost identical.
But the real-solution landscape has changed completely.
This is where a student who simply recognises “quadratic” and begins factorising mechanically may miss what the sign is doing.
A student who understands squares knows:
x² ≥ 0
for real x.
Therefore:
x² = −4
has no real solution.
One symbol changed.
The allowable world changed.
Conditions are often more important than numbers
Suppose a quadratic equation is said to have two distinct real roots.
The student knows:
discriminant > 0.
Now I alter just two words:
two equal real roots.
The condition becomes:
discriminant = 0.
Then:
no real roots.
Now:
discriminant < 0.
The algebraic expression for the discriminant has not changed.
The relationship between its value and the root structure has.
This is a very useful test.
Students who memorise:
“b² − 4ac”
may still not understand what information it carries.
Change the condition and ask what should happen to the discriminant.
If the answer comes quickly, the concept is becoming organised.
Trigonometry can reveal method dependence beautifully
Take a right-angled triangle.
Suppose we know:
opposite side = 6,
hypotenuse = 10.
Find angle θ.
Use:
sin θ = 6/10.
Now I change one fact:
The side of length 10 is no longer the hypotenuse; it is adjacent to θ.
The numerical values are unchanged.
The method changes.
Now:
tan θ = 6/10.
The student who was reacting to the numbers may continue using sine.
The student who was reading the relationships changes method.
This is why very small perturbations are powerful.
The question still looks familiar.
Only the mathematical role of one quantity has changed.
That is enough to test whether the student is seeing the diagram or merely recalling a pattern.
This is particularly useful with rotated diagrams
Students often learn trigonometry on triangles that all face roughly the same way.
Then a triangle is rotated.
Nothing mathematical has changed.
But “opposite”, “adjacent” and “hypotenuse” are no longer located in familiar visual positions.
Now rotate again.
Then change which acute angle is labelled θ.
The physical side that was previously “opposite” may become “adjacent”.
The triangle has not changed.
The reference angle has.
This is an excellent example of why mathematical labels are relational.
A side is not permanently “opposite”.
It is opposite to a particular angle.
Changing one labelled angle tests whether the student understands that dependence.
Sometimes changing one number should change almost nothing
Suppose a student has learned to find the gradient through points:
A(2, 3),
B(6, 11).
Gradient:
(11 − 3)/(6 − 2) = 8/4 = 2.
Now change B to:
B(7, 13).
Gradient:
(13 − 3)/(7 − 2) = 10/5 = 2.
The point changed.
The gradient did not.
Why?
Both points still lie on a line with gradient 2.
This creates a different kind of expectation.
A student’s job is not always to assume that changing an input must change the requested output.
Sometimes a quantity is invariant under the particular change made.
That is mathematically interesting too.
This is where I ask a second question: “What did I change that mattered?”
Suppose I change a coordinate but keep the point on the same line.
The gradient remains unchanged.
Now move the point off the line.
The gradient changes.
The student can compare.
This begins to teach sensitivity.
Some features of a problem strongly control the result.
Some do not.
Some affect only one part.
Knowing the difference is an important form of mathematical judgement.
Probability gives a useful caution
Suppose a fair coin is tossed 10 times.
The expected number of heads is 5.
Now change 10 tosses to 100.
Expected heads:
50.
The expectation scales neatly.
But does that mean exactly 50 heads becomes guaranteed?
No.
The number of trials changed.
The expected frequency changed.
The random nature of the experiment did not disappear.
This is why changing one number is useful only if the student also understands what the model promises.
A larger sample may make proportions more stable in many contexts.
It does not convert probability into certainty.
Again, the perturbation reveals which conclusion is allowed to change and which is not.
Small changes can expose overgeneralisation
Suppose a student believes:
(a + b)² = a² + b².
Choose:
a = 1,
b = 1.
Then:
left side = 4,
right side = 2.
The claim fails.
Now choose:
a = 0.
Both sides become:
b².
The false rule appears to work.
Change one value.
The illusion disappears.
This is why examples have to be chosen carefully.
A student can sometimes carry a false idea for a long time because familiar examples happen not to expose it.
Small controlled changes are a good way to test whether the rule is genuinely general.
I sometimes change the question after the student gives a correct answer
Suppose:
“Find the value of x.”
The student solves correctly.
Then I ask:
“Now suppose x represents a length. Does anything change?”
Perhaps one of the algebraic roots is negative.
It is no longer admissible.
Or:
“Suppose x must be an integer.”
A decimal solution is now impossible.
Or:
“Suppose the two roots must be distinct.”
Now a repeated solution becomes unacceptable.
The calculation may remain completely correct.
The interpretation changes.
This is important because examination Mathematics often includes conditions that apply after the algebra.
A student who regards solving as finished the moment a number appears can miss this.
Small perturbations are also useful for checking modelling
Suppose a student models taxi fare as:
C = 4 + 0.50d
where d is distance.
If d increases by 1, what happens to C?
It increases by $0.50.
Fine.
Now suppose the base fare changes from $4 to $5.
What happens to the graph?
It shifts upward by 1.
Does the gradient change?
No.
Now change the per-distance rate from $0.50 to $0.70.
The gradient changes.
The intercept does not.
This is much more informative than asking the student merely to identify m and c in:
y = mx + c.
The student can see what each parameter controls.
That is the beginning of understanding a model as an object that responds systematically to parameter changes.
Parameters become much more meaningful this way
Consider:
y = ax².
If a = 1:
y = x².
If a = 2:
y = 2x².
If a = 1/2:
y = 0.5x².
If a = −1:
y = −x².
Rather than memorising separate statements about “narrower”, “wider” and “reflected”, the student can compare nearby cases.
What happens as |a| increases?
The graph becomes steeper/narrower relative to the standard parabola.
What happens when the sign changes?
It reflects vertically.
Now a is not simply a letter in a formula.
It controls a family of graphs.
This is a powerful conceptual transition in Secondary and Additional Mathematics.
But this exercise has a boundary
I do not want students endlessly changing numbers after every problem.
That would make lessons slow and artificial.
Sometimes practice needs straightforward repetition.
Execution needs fluency.
A student learning algebraic fractions may simply need several clean examples.
A student learning differentiation may need enough repetition for the power rule to become automatic.
The perturbation question is most useful when I want to test ownership.
The student has already demonstrated one successful execution.
Now one carefully chosen alteration can reveal more than five near-identical questions.
It is a diagnostic instrument, not a ritual.
The change also has to be meaningful
Suppose a student solves:
2x + 3 = 11.
I change 11 to 12.
Fine.
But if I repeatedly change only arbitrary numbers without asking the student to compare relationships, I have merely created another worksheet.
The power comes from choosing changes that probe something.
Change a coefficient to see sensitivity.
Change a sign to expose structure.
Change a condition to test admissibility.
Rotate a diagram to test relational understanding.
Change one piece of data to see whether the method still applies.
Change the target to see whether different information becomes relevant.
The perturbation should have a reason.
This can help distinguish conceptual difficulty from execution difficulty
Suppose the student correctly solves:
x² − 5x + 6 = 0.
Now I change it to:
x² − 5x + 5 = 0.
She immediately begins searching for integer factors of 5 whose sum is −5.
She searches.
And searches.
Nothing works.
What should happen next?
The original factorisation method no longer produces neat integer factors.
The student needs another route.
Perhaps the quadratic formula.
The important question is not whether she can perform the quadratic formula once prompted.
It is whether she recognises that the changed coefficients have altered the usefulness of factorisation.
This is method selection.
One number changed.
The method landscape changed.
Another small change can test execution without changing the idea
Original:
x² − 5x + 6 = 0.
Changed:
x² + 5x + 6 = 0.
The same factorisation structure still works:
(x + 2)(x + 3) = 0.
The method did not change.
The signs did.
If the student knows what to do but obtains:
x = 2, 3
instead of:
x = −2, −3,
the issue is execution or interpretation of factors.
Because the conceptual route remained stable, the diagnostic signal is cleaner.
Small changes are especially good at exposing transfer without changing the whole world
Transfer is sometimes tested too dramatically.
Teach a routine algebra question.
Then immediately give a long unfamiliar word problem.
The student fails.
What failed?
Possibly everything.
Recognition.
Language.
Representation.
Method choice.
Arithmetic.
Working memory.
The jump was too large to locate the weakness.
A small perturbation is gentler.
First change the coefficient.
Then the sign.
Then the form.
Then the context.
The student crosses the distance gradually.
We can see where the understanding stops travelling.
That is much more useful for teaching.
This gives parents a better way to review homework
If your child has just completed a question correctly, you do not necessarily need another page of practice.
Choose one question.
Change one feature.
Then ask for a prediction before a full calculation.
For example:
Original:
A circle has radius 4 cm.
Find its area.
Child obtains:
16π cm².
Change radius 4 to 8.
Ask:
“Before calculating, what happens to the area?”
If she says:
“It doubles,”
there is something worth discussing.
If she says:
“The radius doubled, but area depends on r², so it becomes four times as large,”
that is strong.
Or:
Original gradient = 3.
Change both coordinate differences by the same factor.
Should gradient change?
No.
Or:
Original triangle is right-angled.
Remove the right angle.
Can we still use Pythagoras?
Not automatically.
One careful variation can tell you whether the child owns the relationship underneath the answer.
A useful four-question sequence
I sometimes think of it like this.
First:
Solve it.
Second:
Predict what happens if I change this.
Third:
Explain why.
Fourth:
Now calculate and check whether your prediction survived.
Take:
A = πr², r = 5.
Solve:
A = 25π.
Change r to 10.
Predict:
Area becomes four times as large.
Explain:
Radius doubled, and area depends on r².
Calculate:
A = 100π.
Compare:
100π / 25π = 4.
This closes the loop.
The prediction was not a guess.
The calculation tests the reasoning.
Measurement becomes more interesting too
Instead of recording only whether a student gets the altered question right, I look at several stages.
Did she notice which relationship changed?
Could she predict direction?
Did she know whether the original method remained valid?
Could she execute the revised method?
Could she explain the difference between the two questions afterwards?
Those are useful observations.
A student may calculate both correctly but fail to explain the connection.
Another may make an arithmetic slip yet give an excellent structural prediction.
Their teaching needs differ.
A single mark would hide that.
This also changes how I think about strong students
High-performing students can become exceptionally good at recognising familiar forms.
That skill helps them move quickly.
But speed creates its own risk.
The student sees enough familiar cues and launches the standard method before noticing the one detail that changed.
A minus sign.
A domain restriction.
A non-right triangle.
“With replacement” becomes “without replacement”.
“Maximum” becomes “minimum”.
“Two equal roots” becomes “two distinct roots”.
The question is almost familiar.
That is precisely why it can be dangerous.
Small perturbations train a valuable pause:
What changed, and does that change my route?
For strong students, that pause can be worth more than another twenty routine questions.
Adolescents are also learning something larger here
School can sometimes make knowledge feel binary.
Right or wrong.
Know or do not know.
Finished or unfinished.
Mathematics is often subtler.
A student may know a method but not its conditions.
Understand a relationship but not its sensitivity.
Execute correctly but fail to predict direction.
Solve one case but overgeneralise to another.
Changing one piece of a problem makes these gradations visible.
It shows the student that understanding is not merely possession of an answer.
It is knowledge of the dependencies that make the answer true.
That is a more mature intellectual standard.
The useful next route
Take one Mathematics question your child recently solved correctly.
Do not choose a failed question.
Start from success.
Then alter one feature.
If it is algebra, change a sign or coefficient.
If it is geometry, change an angle condition or scale factor.
If it is trigonometry, change which side is known.
If it is a graph, change one parameter.
If it is probability, change the number of trials or whether replacement occurs.
Before solving the altered question, ask:
What do you expect to change?
Then:
What should stay the same?
Then:
Can you still use the same method? Why?
Only after that should the student calculate.
Compare the two answers.
If the prediction was wrong, that is useful.
Find out which relationship was misunderstood.
If the prediction was right but the calculation failed, the concept may be stronger than the execution.
If the student cannot decide whether the old method still applies, method selection needs attention.
If she predicts, chooses and calculates correctly, change the surface slightly more next time.
The exercise can take five minutes.
That is enough.
What long teaching has made me notice
A correct answer is satisfying.
It closes a question.
But teaching sometimes needs to reopen it.
Not because the answer was insufficient.
Because the answer gives us a stable place from which to ask something deeper.
What if this number were larger?
What if this sign changed?
What if the triangle were not right-angled?
What if the root had to be positive?
What if the scale factor doubled?
What if the constant changed but the gradient did not?
What if the same information were used to answer a different question?
A student who can respond to those changes begins to see Mathematics less as a collection of frozen examples.
She sees relationships with behaviour.
Inputs affect outputs.
Conditions open and close methods.
Some quantities scale.
Some remain invariant.
Some changes are gentle.
Some small changes alter the entire mathematical world.
That is one reason I sometimes resist moving immediately to the next worksheet after a student gets something right.
There may still be a great deal of Mathematics sitting inside the question she has already solved.
Changing one carefully chosen piece can expose it.
And over time, I want the student to develop a habit that becomes increasingly independent of me.
When a problem looks familiar but not quite identical, she should not merely think:
“I have seen this before.”
She should look more carefully and ask:
“What changed—and what does that change?”
That is a very good question.
It protects against careless pattern matching.
It supports estimation and checking.
It strengthens method selection.
It makes formulas behave like relationships rather than instructions.
And perhaps most importantly, it teaches a young person that understanding something means more than knowing what happens in one case.
It means having some idea of how the thing will behave when the world around it moves.
That is when Mathematics begins to feel less like completing examples and more like understanding a system.