Suppose a measurement changes by only a small amount. Can its calculated result change by a much larger amount?
Sometimes yes. Sometimes no. The answer depends on the mathematical relationship.
This is one of the more subtle reasons to study Additional Mathematics: it teaches students to understand how sensitive a result is to a change in its input.
The idea appears naturally in functions, graphs, derivatives and mathematical modelling. A small change in radius affects area differently from volume. A small input change near a steep graph can have a much greater effect than the same input change near a flat part. A measurement that looks precise may still produce an uncertain result if the relationship amplifies small differences.
The subject gives learners a way to investigate these effects rather than simply say that a number “went up a bit”.
This is useful preparation for later study in Mathematics, science, engineering, economics and computing. Practical applications need suitable models and evidence, but the mathematical foundation can begin with familiar secondary-school functions.
What does sensitivity mean in a function?
A function connects an input with an output.
For example:
f(x) = x².
At x = 2, the output is 4.
At x = 3, the output is 9.
At x = 10, the output is 100.
Now ask how much the output changes when the input increases by a small amount h.
The new output is:
f(x + h) = (x + h)².
Expand:
x² + 2xh + h².
Subtract the original output x².
The exact change is:
Δy = 2xh + h².
This equation reveals something important. The same h may produce a different output change depending on the starting value x.
A small movement in x has a larger effect when x itself is large.
That is sensitivity: how the output responds when an input is varied.
Compare the same change at two starting points
Let h = 0.1.
First start at x = 2.
The output changes from 4 to 2.1² = 4.41.
The increase is 0.41.
Now start at x = 10.
The output changes from 100 to 10.1² = 102.01.
The increase is 2.01.
The input increased by exactly 0.1 in both cases.
But the resulting output increases are not equal.
The second increase is much larger.
This is because the quadratic function becomes steeper as x increases in the positive direction.
A student who knows only how to substitute may calculate the results separately. A learner who understands the structure can predict which case will change more even before the arithmetic.
That is a significant conceptual step.
The derivative expresses local sensitivity
Differentiate:
f(x) = x².
We obtain:
f'(x) = 2x.
The derivative describes the instantaneous rate of change of the function with respect to x.
At x = 2, the gradient is 4.
At x = 10, the gradient is 20.
For a sufficiently small input change h, the output change can often be estimated by:
Δy ≈ f'(x)h.
This is a local approximation based on the tangent line.
For h = 0.1 at x = 2, it estimates:
Δy ≈ 4(0.1) = 0.4.
The exact change is 0.41.
At x = 10, it estimates:
Δy ≈ 20(0.1) = 2.
The exact change is 2.01.
Both estimates are close.
This is a useful extension of differentiation: the derivative can help estimate the effect of a small change without recomputing the full expression every time.
Why the approximation is not exactly equal
Return to the exact difference:
Δy = 2xh + h².
The derivative estimate is:
Δy ≈ 2xh.
The missing part is h².
When h is small, h² is often much smaller than h.
For h = 0.1, h² = 0.01.
For h = 0.01, h² = 0.0001.
This explains why the tangent approximation becomes more accurate as h approaches zero for our smooth quadratic example.
But the approximation is not a magical equality for every finite change.
The difference between exact and approximate output depends on the function and the size of the input change.
Students should learn to label an estimate as an estimate and check whether its precision is suitable for the question.
A negative input change has a different exact correction
Take f(x) = x² at x = 10.
Now decrease x by 0.1 instead of increasing it.
The new input is 9.9.
The output changes from 100 to 98.01.
The exact change is -1.99.
The derivative estimate gives:
Δy ≈ 20(-0.1) = -2.
Again the estimate is close but not exact.
Compare the two directions:
An increase of 0.1 gives an exact output increase of 2.01.
A decrease of 0.1 gives an exact output decrease of 1.99.
The small asymmetry comes from the h² term.
That detail helps students see why local linear approximation is useful without claiming that a curved graph behaves exactly like a straight line.
The graphical meaning of a sensitive point
Imagine a smooth graph with a steep slope at one point and a gentle slope at another.
Move the same small horizontal distance from each point.
Near the steep section, the vertical change is typically larger.
Near the gentle section, the vertical change is typically smaller.
The derivative captures this local behaviour.
A positive derivative suggests the output increases as x increases nearby.
A negative derivative suggests it decreases.
A derivative with larger absolute magnitude indicates a steeper local response.
However, this comparison uses the same input and output units and is local to the point of interest.
A derivative is not a universal measure of importance. A large numerical gradient may partly reflect the units chosen.
Mathematical interpretation must keep the model’s scales in view.
Radius and area: why the starting size matters
Consider a circle whose area is:
A = πr².
The derivative of area with respect to radius is:
dA/dr = 2πr.
This tells us how rapidly area changes as radius changes.
At r = 1 centimetre, the derivative is 2π centimetres of area-change per centimetre of radius-change.
At r = 5 centimetres, it is 10π in the corresponding rate units.
At r = 10 centimetres, it is 20π.
Therefore the same small increase in radius produces a larger area increase for a larger circle.
This is not because the geometry has become inconsistent.
It is because a larger circle has a longer boundary, so an additional thin outer layer covers more area.
The derivative and the geometric explanation reinforce each other.
A complete circular-area example
Suppose a circular disc has radius 5 centimetres.
Its area is:
A = 25π square centimetres.
Now suppose the radius increases by 0.02 centimetres.
The new radius is 5.02 centimetres.
The exact increase in area is:
π(5.02² – 5²).
Compute:
5.02² = 25.2004.
Therefore:
ΔA = 0.2004π square centimetres.
The derivative estimate at r = 5 gives:
ΔA ≈ (2π × 5)(0.02) = 0.2π square centimetres.
The difference is:
0.0004π square centimetres.
That is small relative to the area change in this example.
The estimate comes from the slope at the starting radius; the exact result includes the curvature effect.
Why the radius has more influence on volume
Now consider a sphere with volume:
V = (4/3)πr³.
Differentiate with respect to r:
dV/dr = 4πr².
At radius 5 centimetres:
dV/dr = 100π square centimetres.
An increase of 0.02 centimetres gives the local estimate:
ΔV ≈ 100π(0.02) = 2π cubic centimetres.
Compare this with the circle’s estimated area increase of 0.2π square centimetres.
The numerical changes and units are different.
We cannot simply declare one ten times “more sensitive” in an absolute physical sense without specifying what quantities and units we compare.
But we can see an important structural difference: sphere volume depends on the cube of radius, while circle area depends on the square.
That changes how each quantity responds to the same radial adjustment.
Percentage change can tell a clearer story
Sometimes absolute changes are not the most useful comparison.
For a circle, area is proportional to r².
If radius increases from r to 1.01r, an increase of 1 percent, area changes by the factor:
(1.01)² = 1.0201.
Therefore area increases by 2.01 percent.
For a sphere, volume is proportional to r³.
The same 1 percent increase in radius changes volume by:
(1.01)³ = 1.030301.
Therefore volume increases by 3.0301 percent.
These are exact finite-change calculations.
For very small percentage changes, the area percentage change is approximately twice the radius percentage change, and the volume percentage change approximately three times the radius percentage change.
The factor-two and factor-three patterns come from the exponents.
This is one reason algebraic structure matters when evaluating sensitivity.
Understand relative sensitivity through a power function
Let:
y = Cx^n,
where C is a non-zero constant and x is positive.
Differentiate:
dy/dx = Cn x^(n-1).
We can relate a small relative output change to a small relative input change.
Using the local approximation:
Δy ≈ Cn x^(n-1) Δx.
Divide by y = Cx^n:
Δy/y ≈ n Δx/x.
Thus the exponent n tells us approximately how many times a small percentage input change is reflected in the percentage output change.
For n = 2, the factor is approximately 2.
For n = 3, it is approximately 3.
For n = -1, the approximate relative change has the opposite sign.
This is an enrichment connection. It grows naturally from the derivative and should be understood as a local approximation, not an exact rule for every large change.
Why a reciprocal can react strongly near zero
Consider:
f(x) = 1/x,
for x ≠ 0.
Differentiate:
f'(x) = -1/x².
At x = 4, the derivative is -1/16.
At x = 1, it is -1.
At x = 0.1, it is -100.
The derivative’s magnitude becomes very large as x approaches zero.
The function is increasingly sensitive to small absolute changes in x near the excluded input.
For example, move from x = 0.1 to x = 0.11.
The output changes from 10 to approximately 9.0909.
That is a decrease of about 0.9091 for an input increase of only 0.01.
A student can understand the behaviour from both the derivative and the original reciprocal expression.
The domain restriction x ≠ 0 remains essential.
Sensitivity and invalid extrapolation
A local derivative estimate works best when the input change is small enough for the tangent to represent the nearby curve well.
It cannot automatically be extended over a very large interval.
For f(x) = x² at x = 10, the derivative is 20.
Using a linear estimate for a change h = 5 would predict an output increase of 100.
But the exact output increase is:
15² – 10² = 225 – 100 = 125.
The difference is 25.
That is much larger than the error for h = 0.1.
The tangent line approximates the curve near the chosen point; it does not replace the curve everywhere.
This is an important lesson in judging model limits.
A mathematically correct derivative can lead to a poor numerical estimate if used outside the range where the approximation is appropriate.
A square-root example
Let:
f(x) = √x,
for x ≥ 0.
At x = 9, the function value is 3.
For positive x, the derivative is:
f'(x) = 1/(2√x).
At x = 9, f'(9) = 1/6.
Suppose we want to estimate √9.12.
The input change from 9 is h = 0.12.
The derivative approximation gives:
√9.12 ≈ 3 + (1/6)(0.12).
That equals:
3 + 0.02 = 3.02.
The exact numerical value is approximately 3.01993.
The estimate is very close.
It gives students a way to interpret a square-root change using local slope rather than treating the decimal as an unexplained calculator result.
What if the derivative is zero?
Consider:
f(x) = x²
at x = 0.
The derivative is:
f'(0) = 0.
A first-order local approximation predicts:
Δy ≈ 0 × h = 0.
But if h = 0.1, the actual change is:
f(0.1) – f(0) = 0.01.
The change is small, but not zero.
This is a useful warning: at a stationary point, the first-order term may vanish, and a higher-order effect becomes important.
In our quadratic example, the exact difference is h².
The derivative accurately describes the instantaneous gradient at zero.
It does not imply that the function stays constant in a neighbourhood around zero.
This connects sensitivity with the distinction between stationary behaviour and broader change.
A negative derivative can still describe a valuable effect
Let:
f(x) = 100 – 2x,
for an appropriate modelled interval.
The derivative is -2.
If x increases by 0.5, the output decreases by exactly 1.
The negative sign shows the direction of response.
It does not mean the function’s output is necessarily negative.
For example, at x = 10 the output is 80.
A negative derivative and a positive output can coexist.
Students should distinguish a function value from its rate of change.
This seems elementary, but it becomes increasingly important when functions describe physical quantities or economic models.
A-Math teaches students to name the relevant quantity before interpreting its sign.
Comparing sensitivities requires compatible units
Imagine one function measures output in metres and another measures output in centimetres.
The numerical derivatives will differ by a factor of 100 even if they describe the same physical relationship.
A gradient of 2 metres per second is numerically 200 centimetres per second.
It would be wrong to conclude that the second function describes a process a hundred times more sensitive.
The numbers use different units.
Similarly, changing the horizontal axis from seconds to minutes changes the numerical rate.
When comparing slopes, students must ask what each axis measures.
In more advanced settings, dimensionless percentage or relative comparisons can sometimes be more informative, but only when their meanings are clear.
Sensitivity can help us check a predicted result
Suppose a learner evaluates:
f(x) = 3x² + 2
near x = 5.
The derivative is:
f'(x) = 6x.
At x = 5, the gradient is 30.
If x increases by 0.01, the output should increase by approximately 0.3.
Now calculate exactly.
f(5) = 77.
f(5.01) = 3(25.1001) + 2 = 77.3003.
The exact increase is 0.3003.
That is consistent with the derivative estimate.
If a student obtained an increase of 30.03 or a decrease of 0.3, the estimate would alert them to a likely arithmetic or sign problem.
Thus sensitivity is not only a topic to study. It can serve as a quick numerical reasonableness check.
A sensitivity question is not always a rate-of-time question
Many calculus problems use time t as the independent variable.
But sensitivity can be measured with respect to any appropriate input.
For A = πr², dA/dr measures how area changes with radius.
For s(t), ds/dt measures displacement change with time.
For a model C(q), dC/dq measures the output’s change with quantity q.
The derivative symbol identifies the input used for the comparison.
A student should not automatically attach “per second” to every derivative.
The correct unit comes from the output unit divided by the input unit.
Reading the symbols accurately is essential.
Four mistakes that distort sensitivity conclusions
Mistake one: treating a local approximation as an exact identity
The student writes Δy = f'(x)Δx for a nonlinear function without recognising that it is generally approximate for a finite input change.
The repair is to compare the tangent estimate with an exact calculation and observe how the error depends on the size of Δx.
Mistake two: ignoring the starting value
The student assumes that a fixed increase in x always creates the same increase in x².
The repair is to compare x = 2 and x = 10 using the same h.
Mistake three: comparing incompatible gradient units
The student decides one process changes faster based only on larger numerical values measured in different units.
The repair is to standardise the units and name both axes.
Mistake four: interpreting a zero derivative as zero total change
The student treats a stationary point as an interval of constant output.
The repair is to compare f(x) = x² near x = 0 and distinguish instantaneous from finite change.
These errors need explanation, not simply additional substitution drills.
A five-session practice plan
Session one: calculate exact input-output differences
Use y = x².
Compare the effect of h = 0.1 at x = 2 and x = 10.
Derive Δy = 2xh + h² and test it numerically.
Session two: connect slopes with approximations
Differentiate x² and estimate nearby changes.
Compare those estimates with exact differences for h = 0.1, h = 0.01 and h = 1.
Describe when the approximation improves and why.
Session three: explore area and volume
Differentiate A = πr² and V = (4/3)πr³.
Compare how a small change in radius affects each quantity.
Check the unit of every derivative.
Session four: examine sensitive regions
Study y = 1/x near x = 4, 1 and 0.1.
Plot or calculate nearby values.
Explain why the domain restriction at zero cannot be ignored.
Session five: attempt independent transfer
Estimate √9.12 using a derivative and then check it.
Try an unfamiliar function and identify the input, output and rate units.
Finish by explaining whether the result is exact or approximate.
The aim is to build judgement rather than merely memorise the phrase “multiply by the derivative”.
Parents can help by asking about meaning
A parent does not need to teach calculus to help a child reason about changes.
Ask what happens when the input grows slightly.
Ask whether the output changes by the same amount everywhere.
Ask what the derivative measures at a particular point.
Ask whether the student has produced an exact value or an approximation.
Ask how a result could be checked using substitution.
These questions encourage attention to the relationship behind the answer.
A student who can explain why the same radius change affects large and small circles differently is developing more than calculator skill.
They are beginning to recognise mathematical structure.
Singapore’s SEC Additional Mathematics context
The Singapore Examinations and Assessment Board lists 2027 G3 Additional Mathematics as K341.
The syllabus includes differentiation, gradients and rates of change, functions and mathematical reasoning.
The local-sensitivity and approximate-change examples here are explanatory extensions of those ideas. Students should not assume that every approximation technique presented is itself an independently examined requirement.
The official SEAB 2027 G3 syllabus directory remains the authority for assessed content.
The examples are simplified mathematical models. Practical measurements require appropriate units, uncertainty information and subject-specific assumptions.
How three-student teaching helps with this idea
One learner may understand derivatives but not why a numerical estimate is only local.
Another may be accurate with algebra but unable to explain why a larger radius produces a larger area change.
A third may confuse a function’s numerical value with the magnitude of its derivative.
These weaknesses require different explanations.
In a three-student tutorial, the tutor can compare the student’s predicted change with the exact calculation, identify the first misunderstanding and choose a corrective example.
At eduKateSG, foundational algebra and reasoning lead into guided methods, independent practice and transfer.
The immutable Secondary 1 Mathematics tutor Clementi small-groups tutorials reference illustrates the importance of early symbolic fluency and method explanation.
Students making steady independent progress through school teaching may not need tuition. Extra support should address a diagnosed need.
Frequently asked questions
Does a larger derivative always mean a more important change?
No. The interpretation depends on the units, the input, the output and the problem. Numerical gradients should not be compared without considering these factors.
Why can a small input change create a relatively large output change?
Some functions have steep local gradients. A small input movement can then create a noticeable change in the output.
Is Δy = f'(x)Δx always exact?
No. It is generally a local first-order approximation for differentiable functions. Exact finite changes may include additional terms.
Why do we calculate both exact and estimated differences?
Their comparison helps explain the usefulness and limitations of the tangent approximation.
Can a zero derivative mean the function never changes?
No. A zero derivative at one point does not imply the function is constant over an interval.
How can I check a sensitivity estimate?
Substitute the perturbed input into the original function and compare the exact difference with the derivative-based estimate.
Is this relevant outside Mathematics?
Yes. It is foundational to later work with measurements, modelling and changing quantities, although practical applications also need attention to uncertainty and the suitability of assumptions.
Why studying sensitivity matters
Additional Mathematics does more than describe a relationship at one input.
It helps students understand how that relationship responds to change.
A learner can use a derivative to anticipate the direction and approximate size of a nearby output change, compare the result with an exact calculation and decide whether a local estimate is appropriate.
They also learn to check units and recognise when an assumption becomes unreliable.
That combination of calculation and judgement is worth learning.
For connected reading, see Why Study Additional Mathematics | Read Parameters to Predict a Family of Graphs, Why Study Additional Mathematics | Connect Changing Measurements with Related Rates and the eduKateSG Additional Mathematics project library.
A worked log-sensitivity example
Consider f(x) = ln x for x > 0.
Its derivative is f'(x) = 1/x.
At x = 10, a small increase of 0.1 gives the tangent estimate:
Δf ≈ (1/10)(0.1) = 0.01.
The exact change is ln 10.1 – ln 10 = ln 1.01, approximately 0.0099503.
Now compare the same absolute input change near x = 1.
The derivative is 1, giving an estimated increase of 0.1.
The exact change is ln 1.1 – ln 1 = ln 1.1, approximately 0.09531.
The same absolute input increment has a very different output effect because the logarithm’s slope depends on its starting input.
This also shows why a logarithm cannot simply be extended through x = 0. The derivative becomes increasingly steep near the positive side of the boundary.
A model can be sensitive without being unreliable
High sensitivity is not automatically a flaw. It is a property of the relationship being studied.
For example, the volume of a sphere naturally depends on the cube of its radius. It would be wrong to treat a large volume response to radial changes as a mathematical defect.
The practical question is whether our input is measured accurately enough and whether the model describes the real object appropriately.
If a radius is measured with limited precision, a volume calculated from it inherits uncertainty. The exact uncertainty depends on the measurement process and any additional sources of error.
A-Math can teach the foundational derivative relationship, while experimental science introduces fuller uncertainty analysis.
Students should learn to separate the sensitivity of a mathematical rule from the quality of the evidence supplied to that rule.
Why relative changes sometimes matter more than absolute changes
Suppose the radius of one circle grows from 1 to 1.1 centimetres, while another grows from 10 to 10.1 centimetres.
Both radii increase by 0.1 centimetres.
But the first radius increases by 10 percent, while the second increases by only 1 percent.
The corresponding circle areas change by factors 1.1² and 1.01².
These are increases of 21 percent and 2.01 percent respectively.
So the same absolute input change can represent very different relative changes, and the percentage effect on area is different too.
This is why good mathematical communication states whether a comparison is absolute or relative.
Neither measure is always the better one. The right comparison depends on what the problem asks and what decision is being made.