There is a kind of checking I see students do that looks responsible.
They finish the question.
Then they do the same calculation again.
Sometimes more slowly.
Sometimes on a fresh line.
Sometimes with the calculator a second time.
If both attempts produce the same answer, they feel reassured.
I understand why.
Agreement feels like confirmation.
But there is a weakness in this kind of checking.
If the same mistaken idea drives both attempts, the second calculation can reproduce the first error perfectly.
The student has not really challenged the answer.
She has asked the same machinery to vote twice.
After many years of teaching Mathematics, I have become increasingly interested in a different kind of checking.
I want the check to have some independence from the original route.
If the student solved an equation algebraically, perhaps substitute the answer back into the original equation.
If she calculated an exact length, perhaps compare it with the geometry of the diagram and the bounds imposed by the other sides.
If she used a calculator, perhaps ask whether the order of magnitude makes sense.
If she differentiated to find a stationary point, perhaps inspect what the graph or derivative sign should do around it.
If she found a probability, perhaps use the complement or check that the result lies between 0 and 1.
The point is not to perform Mathematics twice.
It is to ask the answer a different question.
A useful check should be capable of saying:
“No. Something in your first route cannot be right.”
The direct answer
Checking is strongest when the checking method can fail differently from the original method.
Repeating the same working can catch a transcription slip or an accidental calculator entry.
That has value.
But it is poor protection against a conceptual mistake, a wrong model, a misread condition or a consistently misapplied formula.
The more independent the check is, the more kinds of failure it can expose.
This distinction matters especially in Secondary Mathematics and A-Math because students often have enough technical skill to produce long, internally consistent wrong solutions.
They differentiate correctly after choosing the wrong function.
They substitute accurately into an equation that was formed incorrectly.
They press the calculator flawlessly after selecting the wrong trigonometric ratio.
They perform ten lines of algebra without once violating the rules—on a mathematical object that was wrong from Line 1.
A second performance of the same route may look equally convincing.
A different check has a better chance of noticing.
Substitution is a powerful example
Suppose a student solves:
3x + 7 = 25.
She subtracts 7:
3x = 18
then divides by 3:
x = 6.
To check, she could repeat:
25 − 7 = 18,
18 ÷ 3 = 6.
Fine.
But a stronger check is:
put x = 6 into the original equation.
3(6) + 7 = 18 + 7 = 25.
The original relationship is satisfied.
That matters.
The check approached the answer from the other direction.
The solution route asked:
“What x makes the equation true?”
The check asks:
“Does the x I found actually make it true?”
These are related questions, but they are not identical calculations.
This becomes much more important when the algebra is longer
Consider:
2x² − 7x + 3 = 0.
Suppose the student factorises:
(2x − 1)(x − 3) = 0.
So:
x = 1/2 or x = 3.
Instead of factorising again, substitute.
For x = 1/2:
2(1/4) − 7(1/2) + 3
= 1/2 − 7/2 + 3
= −3 + 3
= 0.
For x = 3:
2(9) − 21 + 3 = 0.
Both work.
Now imagine the student had factorised incorrectly but consistently.
Repeating the same factorisation logic might reproduce the same pair.
Substitution returns authority to the original equation.
That is why I like it.
The check is anchored to what the question originally required.
A student can check the arithmetic while missing the Mathematics
Suppose a right-angled triangle has hypotenuse 10 cm and one leg 6 cm.
The student is asked for the other leg.
She accidentally writes:
x² = 10² + 6²
instead of:
x² + 6² = 10².
Then:
x² = 136
so:
x ≈ 11.66 cm.
The square root is correct.
The calculator entry is correct.
The answer is neatly rounded.
Now she checks by typing:
√136
again.
11.66.
Same answer.
The check has confirmed only the final arithmetic.
It has not questioned the model.
But the geometry itself gives a powerful independent check.
The unknown leg of a right triangle cannot be longer than the hypotenuse.
The answer 11.66 cm should be rejected immediately.
We did not need to know the correct answer yet.
We needed only one structural fact:
the hypotenuse is the longest side.
This is why conceptual understanding is part of checking.
It gives the student constraints that calculation must obey.
Bounds are an underrated checking tool
Suppose:
3.9 × 21.2
appears in a calculation.
Before multiplying exactly, we know:
roughly:
4 × 20 = 80.
So an answer around 80 is plausible.
If the calculator returns:
826.8,
something is wrong.
Perhaps a decimal point was entered incorrectly.
Perhaps the expression was typed as:
39 × 21.2.
The estimate does not tell us the exact answer.
It tells us where the exact answer is allowed to live.
This kind of check is powerful because it is extremely cheap.
A few seconds of approximate reasoning can protect several minutes of accurate but mis-entered calculator work.
But estimation is not enough for every error
There is an important boundary.
Suppose the true answer is 82.68 and the student obtains 80.68.
Both are of the same order of magnitude.
An estimate near 80 will not catch the error.
So different checks have different resolution.
Order-of-magnitude checking catches gross numerical errors.
Substitution catches invalid equation solutions.
Units catch dimensional inconsistencies.
Domain restrictions catch inadmissible values.
An alternative method may catch a wrong route.
A graph may catch impossible behaviour.
No single check catches everything.
Good checking is therefore selective.
The student should ask:
“What kind of mistake is most plausible here, and what check would expose it?”
Units can disagree with otherwise beautiful algebra
Suppose a student calculates speed.
Distance = 120 km.
Time = 2 hours.
She somehow writes:
speed = time ÷ distance
and obtains:
2/120 = 1/60.
The arithmetic is flawless.
But inspect the units.
Hours divided by kilometres gives:
h/km.
That is not the requested unit for speed.
Speed should have dimensions:
km/h.
The unit has caught a reversed relationship.
This is more useful than recalculating 2 ÷ 120.
The numerical division was never the problem.
The formula was.
This is why I do not treat units as decoration at the end
The unit can participate in the reasoning.
Area should produce square units.
Volume should produce cubic units.
Speed should be distance per time.
Density should be mass per volume.
A gradient in a real-world graph may carry a meaningful rate unit.
If the units emerging from a formula make no sense, that is evidence.
Students sometimes think checking means asking:
“Did I press the calculator correctly?”
Sometimes the better question is:
“What kind of quantity did I just calculate?”
Probability has its own natural checks
Suppose a student obtains:
P(A) = 1.2.
No additional calculation is required.
For an ordinary probability:
0 ≤ P(A) ≤ 1.
So 1.2 cannot be correct.
Or suppose a question asks for probabilities of four mutually exclusive exhaustive outcomes and the student obtains:
0.20,
0.35,
0.15,
0.25.
Add them:
0.95.
Something is missing or miscalculated.
Again, the checking relationship differs from the individual probability calculations.
We are using a structural property of the whole probability model.
Complements can provide especially elegant checks
A bag contains 7 red and 3 blue counters.
Probability of red:
7/10.
Probability of blue:
3/10.
Since these are the only possibilities:
7/10 + 3/10 = 1.
Good.
Now suppose a more complicated problem asks for:
P(at least one success).
A direct calculation may involve several cases.
If we can independently find:
P(no successes),
then:
P(at least one success) = 1 − P(no successes).
Two approaches may meet at the same result.
That is stronger evidence than repeating one case-by-case sum twice.
Simultaneous equations give students a built-in check
Suppose:
2x + y = 11
x − y = 1.
Solving gives:
x = 4,
y = 3.
Now put those values into both original equations.
First:
2(4) + 3 = 11.
True.
Second:
4 − 3 = 1.
True.
This is worth teaching because simultaneous equations create multiple constraints.
A candidate pair must satisfy all of them.
A student can make an algebraic error and still accidentally satisfy one equation.
Satisfying both is stronger evidence.
Graphs can check algebra at a different level
Suppose:
x² − 5x + 6 = 0
is solved algebraically.
Roots:
2 and 3.
Now think about:
y = x² − 5x + 6.
The graph should cross the x-axis at:
x = 2,
x = 3.
If a sketch or graphing tool shows intercepts around 2 and 3, the representations agree.
This is not necessary for every routine quadratic.
But as a teaching exercise it is valuable because the check comes from another representation.
Equation solving and graph intersections converge.
That strengthens both.
A graph can also challenge an algebraic answer before exact checking
Suppose a student claims a quadratic has roots:
x = 2 and x = 7,
but the graph clearly crosses near:
x = −2 and x = 7.
The disagreement is useful.
Now we investigate.
The graph may be wrong.
The algebra may be wrong.
But one representation has challenged the other.
This is what I mean when I say a good check should be allowed to disagree.
If every checking method is designed merely to confirm the first answer, it is not much of a check.
A-Math differentiation offers several different checks
Suppose:
y = x³ − 3x² + 2.
Then:
dy/dx = 3x² − 6x.
A student may check the differentiation term by term.
Fine.
But we can also reason structurally.
Derivative of a cubic should ordinarily be quadratic.
The derivative of the constant 2 should be zero.
The derivative should factor:
3x(x − 2).
So stationary points occur at:
x = 0 and x = 2.
If the student’s derivative is:
3x³ − 6x,
the degree is immediately suspicious.
The original function was cubic.
Its derivative should not still be cubic.
That is a conceptual check.
It is cheap.
And it catches a type of error that repeating the same power-rule mistake may not.
This is one reason I like students to know what an answer should look like
Not the exact answer.
The shape of it.
A derivative of a quadratic should be linear.
An area cannot be negative in an ordinary geometry problem.
A probability cannot exceed 1.
A length in a diagram has physical bounds.
A mean should lie between the minimum and maximum values of a non-empty ordinary dataset.
A scale enlargement with factor greater than 1 should not produce a smaller corresponding length.
A percentage discount below 100% should not ordinarily create a negative selling price.
These expectations create a mathematical environment around the answer.
The answer has to fit into it.
Checking can therefore begin before calculation
This is important.
Students often think checking happens after the final line.
But the best protection sometimes begins before the first calculation.
Before solving, ask:
Should the answer be positive?
Roughly how large?
Should there be one solution or two?
Should this graph be increasing or decreasing here?
If I double this input, should the output double?
Which unit should appear?
Then solve.
Now the final answer has something to report back to.
A student who predicts:
“the probability should be less than one-half”
and obtains 0.82 has a reason to stop.
Without the prediction, 0.82 may simply look like a respectable decimal.
This does not mean every prediction must be right
A student’s expectation can be wrong.
That is useful too.
Suppose she predicts that doubling a circle’s radius doubles its area.
Then calculates:
radius 3 → area 9π.
radius 6 → area 36π.
The calculation contradicts the prediction.
Good.
Now there is something worth understanding.
Area depends on the square of the radius.
The check has not merely protected the calculation.
It has exposed a conceptual assumption.
This is why disagreement can be educational.
Students sometimes “check” only after they already know the answer
This creates another weakness.
They look at the answer key.
Then return to the working and say:
“Yes, I see.”
The answer has now become a cue.
Many mistakes become hard to notice because the student is reading toward a known destination.
A stronger checking habit happens before external confirmation.
Finish.
Pause.
Interrogate the answer.
Then look at the key.
This keeps the student’s own judgement alive.
Parents can help without re-teaching the whole Mathematics question
If your child finishes a problem, you can ask:
“How could you check that without doing exactly the same thing again?”
You do not necessarily need to know the solution.
The child might suggest:
substitute it back,
estimate,
check the unit,
use another formula,
draw a quick graph,
use the complement,
look at whether the value is physically possible,
try the reverse operation.
If she cannot think of a check, that itself is useful.
Perhaps the answer is currently supported by only one route.
That is not automatically bad.
Some questions genuinely have no cheap independent check.
But thinking about what would count as independent evidence is a valuable habit.
Reverse operations are often useful
Suppose a student calculates:
15% of 240 = 36.
To check:
240 − 36 = 204,
so after a 15% discount the price is 204.
Could we reverse the reduction?
204 is 85% of the original.
So:
204 ÷ 0.85 = 240.
The reverse calculation returns us to the starting value.
That gives stronger evidence than pressing:
0.15 × 240
again.
But reverse checking also has limits
If the original model was wrong, reversing within the same model may still confirm it.
Suppose a student misunderstands a 20% increase as adding 20 rather than multiplying by 1.2.
Starting with 240, she obtains 260.
Then reverses by subtracting 20.
Back to 240.
The check “works”.
The misconception survives.
This is an excellent reminder:
a check can only test assumptions that it does not itself share.
The reverse operation was independent of arithmetic error.
It was not independent of the student’s incorrect interpretation of percentage increase.
To catch that, we need a conceptual check:
20% of 240 is 48, not 20.
Or compare with a proportional expectation.
That is why checking quality matters more than the mere presence of a check.
Repetition is still useful for some mistakes
I do not want to dismiss it.
If a student suspects a calculator-key error, enter the expression again carefully.
If a long addition may contain a transcription mistake, recompute it.
If a sign may have been copied incorrectly, reread the line.
Simple repetition is cheap and appropriate for simple execution risks.
The mistake is believing it provides universal assurance.
It does not.
Different failure modes need different checking methods.
There is also such a thing as over-checking
This matters during examinations.
A student can spend so long validating secure answers that she leaves unattempted marks later in the paper.
The objective is not maximum certainty on every line.
It is sensible confidence across the whole assessment.
I generally prefer checks that are:
cheap,
targeted,
and capable of catching a plausible error.
Substitute one root.
Inspect units.
Estimate the magnitude.
Check the sign.
Review the domain.
See whether probabilities sum correctly.
These can take seconds.
Solving the entire five-mark problem again from scratch may not be a good use of examination time.
Strong students sometimes over-check because they do not trust success
This is a different issue.
They solve correctly.
Then solve again.
Then calculate a third time.
Then compare with a friend.
Then worry that all three are wrong.
Accuracy has stopped increasing meaningfully.
Anxiety has taken control of the checking process.
For these students, the repair may be to establish a stopping rule.
For example:
one valid solution,
one meaningful independent check,
then move.
Confidence should not require infinite confirmation.
Mathematical checking should reduce uncertainty enough for a decision.
It cannot remove uncertainty from life entirely.
Weaker students sometimes under-check because the answer feels like relief
The moment a number appears, they want to leave.
Especially after a difficult question.
This is understandable.
But the questions most likely to contain errors are often the ones where a ten-second check has the highest value.
A complicated percentage answer.
A negative length.
A probability.
A trigonometric angle.
A root from a radical equation.
A stationary point.
A calculation involving many calculator entries.
The harder the path, the more useful a cheap independent check can become.
This is why I sometimes teach a “check menu” rather than one checking rule
For equations:
substitute into the original.
For numerical calculations:
estimate order of magnitude.
For geometry:
inspect bounds and known relationships.
For measurement:
check units and dimensions.
For probability:
check 0 to 1, complements, total probability.
For graphs:
compare intercepts, direction and general behaviour.
For differentiation:
check degree, constants, sign behaviour where appropriate.
For modelling:
return to the assumptions and ask whether the answer belongs to the real situation described.
The student does not need to perform every check.
She needs to choose one that is relevant.
That choice itself is mathematical judgement.
Transfer appears when the student starts inventing checks
This is one of my favourite developments.
At first, I say:
“Substitute your answer back.”
Later, the student does it without being asked.
Then she notices:
“This length can’t be 14 because the hypotenuse is only 10.”
Or:
“This percentage answer is impossible because the discounted price became larger.”
Or:
“My derivative can’t be right. I differentiated a quadratic and still have x².”
Now checking has stopped being a teacher-imposed final step.
It has become part of the student’s internal Mathematics.
That is much more valuable.
Measurement can be simple
Take five completed questions.
For each one, ask the student to name a check without performing it.
Then ask:
“What type of mistake would that check catch?”
For example:
Substitution into an equation:
catches an invalid candidate solution.
Estimate:
catches a gross numerical error.
Units:
catches a relationship with the wrong dimensional structure.
Probability range:
catches impossible probabilities.
Alternative method:
may catch a route-specific conceptual or execution error.
Then ask:
“What could this check fail to catch?”
That second question is powerful.
It teaches students that evidence has limits.
No check is magical.
The useful next route for parents
Choose one completed Mathematics question that your child believes is correct.
Do not ask her to solve it again.
Ask instead:
“What is the cheapest different way of trying to prove yourself wrong?”
The wording is deliberate.
We are not looking for reassurance.
We are looking for a test.
If it is an equation, substitute.
If it is geometry, ask whether the size and relationship are possible.
If it is numerical, estimate.
If it is a rate, inspect units.
If it is probability, test bounds or the complement.
If another representation exists, use it briefly.
Then compare.
If the check agrees, confidence improves.
If it disagrees, do not immediately decide which route is wrong.
Investigate the disagreement.
That is where useful learning often begins.
Over several weeks, notice whether the student begins selecting her own checks.
That is a good sign of growing independence.
What long teaching has made me notice
When students first learn to check Mathematics, the instruction sounds simple:
“Check your work.”
But that phrase hides a surprisingly difficult question.
Check it against what?
The same calculation?
The original equation?
An estimate?
A definition?
A physical constraint?
A graph?
A second method?
A unit?
A theorem?
A domain?
The quality of the check depends on the answer.
I think this matters because modern students have very powerful calculation tools.
A calculator can repeat an incorrect model with extraordinary precision.
A computer can execute a wrong formula perfectly.
Speed and accuracy inside the chosen procedure do not guarantee that the chosen procedure belongs to the problem.
So the student’s own mathematical judgement still has a job.
She has to create friction.
Something that can resist the first answer.
Something sufficiently independent to say:
“Look again.”
That is why I do not always feel reassured when a student tells me:
“I checked it twice.”
I sometimes ask:
“Did you check it twice, or did you do the same thing twice?”
There is a difference.
One produces repetition.
The other produces evidence.
And perhaps this is one of the quieter intellectual habits Mathematics can give an adolescent.
Do not trust a conclusion merely because the same route keeps returning it.
Look for another constraint.
Another representation.
Another consequence.
Another way the world would have to behave if the answer were true.
Then let those things meet.
If they agree, we have earned more confidence.
If they disagree, we have learned where to look.
Either way, the Mathematics has become more than a procedure that produces answers.
It has become a system in which answers can be challenged.
That is a much stronger kind of knowledge.