Discover why log–log and semilog plots can turn power and exponential models into straight lines, and how to recover their original constants.
Explore the data-to-graph guide
- Begin with a curved relationship
- Why the raw graph curves
- The power-law model
- Take logarithms on both sides
- Read the slope and intercept correctly
- Transform the science table
- Calculate the slope without premature rounding
- Recover the coefficient a
- A quick verification with a fresh input
- What does the power exponent mean?
- A negative exponent produces a downward line
- A straight line does not always mean a power law
- A second table: exponential growth
- Recover the exponential model from the line
- The key comparison: log–log versus semilog
- Why the model’s domain matters
- Why a model cannot be selected by a perfect graph alone
- The danger of using only two points
- What rounding can do to the gradient
- A third example: recovering n from two points
- Why the intercept is easiest to misread
- A counterexample: swapping the axes changes the gradient
- A counterexample: base changes do not invalidate the method
- A step-by-step graph-reading checklist
- A short independent practice set
- What parents can ask in ordinary language
- What careful A-Math tuition should diagnose
- Where linearisation sits in the 2027 G3 syllabus
- Frequently asked questions
- Related Additional Mathematics guides
- Official source and evidence limits
What is Additional Mathematics when a curved data pattern can be studied with a straight-line graph? An invented science-measurement table makes the idea visible.
A student may plot a dependent quantity y against an input x and see a curve. The shape may suggest a power law or exponential relationship. Rather than guessing parameters by eye, we can transform the original variables so that a suitable model becomes linear. The slope and intercept of that new straight line then reveal the constants in the original nonlinear equation.
This is called linearisation. It appears in the 2027 Singapore G3 Additional Mathematics syllabus as a transformation of relationships into linear form.
The measurements in this article are deliberately constructed for teaching. They are not observations from a real laboratory or evidence that a particular physical process follows a power law.
Begin with a curved relationship
Suppose a science table contains pairs of positive input and output values.
For an illustrative experiment, let the exact mathematical relationship be:
y = 3x²,
where x is a positive numerical input and y is a positive numerical output.
We can imagine that x is a measured setting and y is an associated instrument reading, but we have not specified a real device or attached physical units. The model is purely educational.
Here are five data pairs.
| x | y |
|---|---|
| 1 | 3 |
| 2 | 12 |
| 3 | 27 |
| 4 | 48 |
| 5 | 75 |
The output does not grow by a constant amount when x increases by one.
From x = 1 to x = 2, y increases by nine.
From x = 2 to x = 3, it increases by fifteen.
From x = 3 to x = 4, it increases by twenty-one.
These increasing differences show that a direct plot of y against x is not a straight line.
Why the raw graph curves
A straight line of the form:
y = mx + c
has a constant gradient m.
By contrast, y = 3x² is quadratic. Differentiate it:
dy/dx = 6x.
The gradient depends on x, so it changes across the plotted interval.
At x = 1, the gradient is six.
At x = 3, the gradient is eighteen.
The curve becomes steeper as x increases.
We can still work with that curve, but a transformation may allow us to read the mathematical constants from a straight line, using familiar gradient and intercept methods.
That is the purpose of linearisation.
The power-law model
A more general relationship is:
y = axⁿ,
where a and n are constants.
For the logarithmic transformation used here, assume a > 0, x > 0 and y > 0. These conditions allow us to take real logarithms.
The constant a is the multiplicative coefficient. The exponent n describes how output scales with input.
In our invented table:
a = 3 and n = 2.
But imagine the table were given without telling the student those constants. Could we recover them using a straight-line graph?
Yes, if the power-law model is appropriate. The logarithm laws provide the transformation.
Take logarithms on both sides
Start with:
y = axⁿ.
Take base-ten logarithms:
log₁₀ y = log₁₀(axⁿ).
Use the law for a product:
log₁₀(ab) = log₁₀ a + log₁₀ b.
Then use the power rule:
log₁₀(xⁿ) = n log₁₀ x.
Therefore:
log₁₀ y = log₁₀ a + n log₁₀ x.
This now has the same structure as a straight-line equation.
Write:
Y = log₁₀ y.
X = log₁₀ x.
Then:
Y = nX + log₁₀ a.
It is in the form Y = mX + c.
That is the linearised equation.
Read the slope and intercept correctly
Compare:
Y = nX + log₁₀ a
with:
Y = mX + c.
The slope is:
m = n.
The vertical intercept is:
c = log₁₀ a.
Thus, for a power law, the slope of a log–log graph gives the exponent n.
The intercept is not a itself. It is the logarithm of a.
To recover a:
a = 10ᶜ.
This is a frequent source of error. A student may read a vertical intercept of approximately 0.477 and report a = 0.477. The actual coefficient is close to three because ten raised to 0.477 is close to three.
The transformed graph’s labels matter.
Transform the science table
We will plot log₁₀ y against log₁₀ x, not y against x.
For our exact model y = 3x², calculate the transformed pairs.
| Original x | Original y | log₁₀ x | log₁₀ y |
|---|---|---|---|
| 1 | 3 | 0.0000 | 0.4771 |
| 2 | 12 | 0.3010 | 1.0792 |
| 3 | 27 | 0.4771 | 1.4314 |
| 4 | 48 | 0.6021 | 1.6812 |
| 5 | 75 | 0.6990 | 1.8751 |
The logarithmic values are rounded for display. The underlying algebraic relationship is exact.
Notice that when x = 1:
log₁₀ 1 = 0.
The transformed point has horizontal coordinate zero and vertical coordinate log₁₀ 3.
That point reveals the vertical intercept without any extrapolation in this example.
The remaining points align along a straight line in the transformed coordinates, apart from insignificant rounding in the displayed table.
Calculate the slope without premature rounding
Choose the exact original data pairs x = 1, y = 3 and x = 4, y = 48.
In transformed coordinates, the gradient is:
m = [log₁₀ 48 − log₁₀ 3]/[log₁₀ 4 − log₁₀ 1].
Use logarithm laws:
Numerator = log₁₀(48/3) = log₁₀ 16.
Denominator = log₁₀ 4.
Since 16 = 4²:
log₁₀ 16 = 2 log₁₀ 4.
Therefore:
m = 2.
This is an exact result obtained without relying on rounded table entries.
So the power-law exponent is n = 2.
This check is valuable because rounding the plotted logarithms too early can make the gradient appear to be 1.99 or 2.01 instead of exactly two.
Recover the coefficient a
The transformed vertical intercept is:
c = log₁₀ 3.
Because X = 0 at x = 1, we can identify the intercept directly from the first point:
c ≈ 0.4771.
To obtain a, reverse the logarithm:
a = 10ᶜ.
Thus:
a = 3.
We have recovered the original power-law model:
y = 3x².
The straight-line graph did not change the underlying relationship. It made the parameters easier to identify.
A student should always transform the answer back into the original variables before reporting the model.
A quick verification with a fresh input
Use x = 6, which was not in the original table.
The recovered model predicts:
y = 3(6²) = 3(36) = 108.
In logarithmic form:
log₁₀ y = log₁₀ 3 + 2 log₁₀ 6.
By logarithm laws:
log₁₀ y = log₁₀(3 × 36) = log₁₀ 108.
Therefore y = 108.
Both forms agree.
This is a useful mathematical check on the transformation. But because the table and model were invented, the number 108 is a model prediction, not a verified laboratory outcome.
Real experimental data require independent measurement and assessment of whether a model is appropriate.
What does the power exponent mean?
For:
y = axⁿ,
multiply x by a positive factor k. The output changes to:
y_new = a(kx)ⁿ = kⁿ axⁿ.
Therefore:
y_new = kⁿ y_old.
In the model y = 3x², doubling x multiplies y by 2² = 4.
For example, at x = 2, y = 12.
At x = 4, y = 48.
The output is four times as large.
The exponent n = 2 describes that scaling law.
The straight-line gradient on the log–log graph is therefore more than a numerical statistic. It encodes how the output responds to multiplicative changes in the input.
A negative exponent produces a downward line
Consider another constructed model:
y = 6x⁻³ = 6/x³,
for x > 0.
Take logarithms:
log₁₀ y = log₁₀ 6 − 3 log₁₀ x.
So a graph of log₁₀ y against log₁₀ x has slope −3.
The original output decreases as x increases.
A negative slope on the transformed power-law graph reveals a negative exponent.
For instance, doubling x changes y by a factor:
2⁻³ = 1/8.
Thus the output becomes one-eighth of its former value.
The sign of the linearised gradient has a meaningful interpretation, provided the assumed model and positive-value domain are appropriate.
A straight line does not always mean a power law
Different original relationships can become straight lines after different transformations.
Consider an exponential model:
y = kbˣ,
where k > 0 and b > 0, with b ≠ 1 for a nonconstant exponential relationship.
Take base-ten logarithms:
log₁₀ y = log₁₀ k + x log₁₀ b.
Define:
Y = log₁₀ y.
Keep X = x, without logging x.
Then:
Y = (log₁₀ b)X + log₁₀ k.
This is a straight line too.
But its slope is log₁₀ b, not b itself. Its vertical intercept is log₁₀ k.
The correct axes depend on which mathematical model is being tested.
A second table: exponential growth
Suppose a second invented data set follows:
y = 2.5(3ˣ).
Its values are:
| x | y |
|---|---|
| 0 | 2.5 |
| 1 | 7.5 |
| 2 | 22.5 |
| 3 | 67.5 |
| 4 | 202.5 |
This time, equal increases of one in x multiply y by three.
That differs from a power law, where multiplying x by the same factor causes a fixed multiplicative change in y.
For the exponential model:
log₁₀ y = log₁₀ 2.5 + x log₁₀ 3.
The transformed graph uses x on the horizontal axis and log₁₀ y on the vertical axis.
It is called a semilog transformation because only one plotted variable is logged.
Recover the exponential model from the line
The slope is:
m = log₁₀ 3 ≈ 0.4771.
The vertical intercept is:
c = log₁₀ 2.5 ≈ 0.39794.
Recover b by raising ten to the slope:
b = 10ᵐ = 3.
Recover k by raising ten to the intercept:
k = 10ᶜ = 2.5.
Therefore the original model is:
y = 2.5(3ˣ).
A student who reports b = 0.4771 has confused the transformed slope with the base of the exponential model.
This is exactly the type of distinction linearisation is intended to clarify.
The key comparison: log–log versus semilog
| Original relationship | Horizontal axis | Vertical axis | Gradient |
|---|---|---|---|
| y = axⁿ | log₁₀ x | log₁₀ y | n |
| y = kbˣ | x | log₁₀ y | log₁₀ b |
For the power law, both variables are transformed by logarithms.
For the exponential model, only the output is transformed.
The intercept has a different parameter meaning in each case, but both require reversing the logarithm to recover the original multiplicative coefficient.
Reading the axis labels is the first step before using any gradient or intercept.
Why the model’s domain matters
Real base-ten logarithms are defined only for positive arguments.
For the power-law transformation, log₁₀ x and log₁₀ y require:
x > 0 and y > 0.
For the exponential transformation, log₁₀ y requires y > 0. The x-values may be zero, negative or positive where the exponential model itself is defined.
A student must not take log₁₀ 0 or the real logarithm of a negative number.
If experimental data include zero or negative readings, the student cannot simply apply the same log plot without considering what those readings mean and whether another model is needed.
The domain is part of the transformation’s validity, not a small technical footnote.
Why a model cannot be selected by a perfect graph alone
Our constructed tables lie exactly on their assumed relationships because we generated the numbers from those equations.
Real experimental data include measurement uncertainty and other effects.
A transformed graph that appears nearly straight may be consistent with a model, but appearance alone does not prove that the model is physically correct or valid outside the observed range.
The scale, data range, uncertainty and residual deviations all matter.
In a school question, the problem may instruct the student to assume a power or exponential relationship and use a line to determine constants.
That is a clear mathematical task.
Outside such a question, claiming a scientific law from a neat-looking graph requires stronger evidence than just two endpoints.
The danger of using only two points
Two distinct points determine one straight line. So almost any pair of transformed data points can produce some gradient.
If a third, fourth or fifth point lies far away from the proposed line, the simple model may not fit the full data set.
Our example includes multiple points so the student can check whether they follow the same transformed relationship.
Even then, because the data are invented, the conclusion is about the internal consistency of the model rather than external evidence.
A useful habit is to check all available transformed points, not merely calculate a line from the first two and stop.
What rounding can do to the gradient
Logarithms are often displayed to four decimal places in a classroom table.
If the student uses heavily rounded values to calculate the slope, the result may be slightly different from an exact integer or an exact logarithm.
For the power-law example, using the exact log ratio shows the gradient is exactly two.
In real measurements, the slope may not be a simple whole number. Students should not force it to become an integer just because the sample resembles a squared law.
The correct level of rounding follows the question’s instructions.
Under the 2027 G3 syllabus’s general assessment notes, non-exact numerical answers normally use three significant figures unless a different accuracy is required. But exact symbolic values should be retained where appropriate.
A third example: recovering n from two points
Suppose a power-law model y = axⁿ is assumed and two exact positive data points are given:
(x, y) = (2, 20) and (4, 80).
Because x doubles and y quadruples:
80/20 = (4/2)ⁿ.
So:
4 = 2ⁿ.
Therefore n = 2.
Now substitute x = 2, y = 20:
20 = a(2²) = 4a.
Thus a = 5.
The model is:
y = 5x².
The logarithmic gradient method would give the same result:
n = log₁₀(80/20) / log₁₀(4/2) = log₁₀ 4 / log₁₀ 2 = 2.
This shows that linearisation is not a different mathematical truth. It is a representation that makes the same relationship useful for plotted data and slope calculations.
Why the intercept is easiest to misread
For the power law:
Y = nX + log₁₀ a.
The vertical intercept is log₁₀ a.
For the exponential law:
Y = (log₁₀ b)X + log₁₀ k.
The vertical intercept is log₁₀ k.
In both cases, an intercept is a logarithm of a coefficient.
The common error is to treat c as the original coefficient instead of reversing the transformation.
A good written solution should show both steps:
Intercept c = …
Original coefficient = 10ᶜ.
This ensures the final equation is in the original units and original variables.
A counterexample: swapping the axes changes the gradient
Suppose the transformed relationship is:
Y = 2X + 0.4771.
If a student mistakenly plots X vertically against Y horizontally, the resulting gradient is not two.
Rearrange:
X = (Y − 0.4771)/2.
The gradient of X against Y is one-half.
That graph describes the same points but answers a different slope question.
Therefore the instruction “plot log y against log x” has an exact meaning: log y is the vertical variable and log x the horizontal variable.
The order cannot be silently reversed.
A counterexample: base changes do not invalidate the method
We chose base-ten logarithms.
Natural logarithms also work.
For:
y = axⁿ,
taking ln gives:
ln y = ln a + n ln x.
If both axes use natural logarithms, the slope remains n.
The intercept becomes ln a, and the original coefficient is recovered using a = eᶜ rather than 10ᶜ.
For an exponential law y = kbˣ:
ln y = ln k + x ln b.
Its slope is ln b and its intercept is ln k.
The logarithm base may change the transformed numerical slope for an exponential model, but the recovered b remains the same when the inverse exponential operation uses the matching base.
Students should not mix a base-ten intercept with eᶜ without an appropriate conversion.
A step-by-step graph-reading checklist
A good solution starts by identifying the original assumed relationship.
Then:
- Decide which variables should appear on the transformed axes.
- State the positivity conditions needed for logarithms.
- Transform the original equation algebraically.
- Match it to the form Y = mX + c.
- Determine the gradient and intercept from the supplied line or data.
- Convert those transformed parameters back into the original constants.
- Write the final equation using the original x and y.
- Substitute at least one known data point to check it.
This workflow prevents an otherwise correct straight-line calculation from becoming a wrong nonlinear model.
A short independent practice set
1. For y = 4x³, what is the gradient of a graph of log₁₀ y against log₁₀ x? 2. What is the vertical intercept of that transformed graph? 3. For y = 7(2ˣ), which horizontal variable produces a straight line when the vertical variable is log₁₀ y? 4. If the semilog gradient is log₁₀ 5, what is the original exponential base? 5. Why can you not place x = 0 on a real log₁₀ x horizontal axis? 6. Why does a straight line through two transformed points not prove the model fits all possible data?
Answers: gradient three; intercept log₁₀ 4; ordinary x; exponential base five; because log₁₀ 0 is undefined; and because two points do not establish the behaviour of other observations or the model’s suitability.
Ask the student to say whether each answer concerns an original variable or a transformed variable.
What parents can ask in ordinary language
Ask your child why the original graph can curve while the transformed graph forms a straight line.
Ask why the log–log slope represents a power exponent.
Ask what the intercept gives directly and what still has to be reversed.
Then ask how a power law differs from exponential growth: for a power law, multiplicative changes in x matter; for an exponential relationship, additive changes in x correspond to multiplicative changes in y.
These questions reveal whether the learner is understanding the relationship rather than merely drawing lines with a ruler.
What careful A-Math tuition should diagnose
A student can make many different mistakes in the same graph question.
One may misunderstand which variable belongs on each axis. Another may forget a logarithm law. Another may calculate a correct gradient but fail to reverse the logarithm for the original coefficient.
A fourth may ignore nonpositive values, while a fifth may round too early and lose accuracy.
A useful small-group lesson should identify the precise missing step, repair it with a compact example and then change the numerical data to see whether the student can repeat the reasoning independently.
That is more reliable than telling every student to memorise an unlabelled “slope equals exponent” rule.
Where linearisation sits in the 2027 G3 syllabus
The official 2027 Singapore–Cambridge Secondary Education Certificate G3 Additional Mathematics K341 syllabus lists transformations of relationships, including y = axⁿ and y = kbˣ, to linear form in order to determine unknown constants from a straight-line graph. This appears under Geometry and Trigonometry G2, coordinate geometry in two dimensions.
The required skill is not just plotting. It is connecting the algebraic transformation, graph axes, gradient, intercept and original parameters.
The present tables are original exact mathematical examples illustrating that syllabus content. They do not claim to be real experimental measurements.
SEAB lists K341 as the 2027 G3 Additional Mathematics code, with the familiar 4049 reference for 2026 and earlier. The G2 Additional Mathematics K232 specification should be consulted separately.
Frequently asked questions
What does linearisation mean?
It means transforming a mathematical relationship so that the relevant variables satisfy a straight-line equation, allowing the slope and intercept to reveal useful information.
Is a curved graph always an exponential graph?
No. Quadratic, power-law, exponential and other relationships can produce curves. The model must be established or assessed, not guessed from curvature alone.
Is the slope always the exponent?
For the power-law transformation log y against log x, the slope is n. For the exponential transformation log y against x, the slope is log b.
Why do we take logarithms?
Logarithms convert multiplication into addition and powers into products, allowing certain nonlinear relationships to be rewritten in straight-line form.
Why is the vertical intercept not a directly?
In the transformed power-law equation, the intercept is log a. The original a must be recovered by the matching inverse exponential operation.
Can I use ln instead of log₁₀?
Yes, provided the transformations and inverse conversions use the same base consistently.
Does a straight line prove the scientific law is true?
No. It can be consistent with an assumed model over observed data. Model validity requires appropriate evidence and attention to uncertainty and range.
Is this in G3 Additional Mathematics?
Yes. The 2027 K341 syllabus explicitly lists transformations of y = axⁿ and y = kbˣ to linear form to determine unknown constants from a straight-line graph.
Related Additional Mathematics guides
The savings target that explains logarithms introduces the inverse relationship between exponential powers and logarithms.
The basketball and quadratic functions demonstrates how a curved graph can convey useful information before any transformation.
For the overall subject, start with What Is Additional Mathematics? and the How Additional Mathematics Works learning hub.
Parents can also read the immutable Clementi small-group Mathematics tutor reference for the foundational reasoning and algebra accuracy behind more advanced work.
Official source and evidence limits
The primary curriculum reference is SEAB’s 2027 G3 Additional Mathematics K341 syllabus (PDF), Geometry and Trigonometry G2, together with the official G3 subject listing.
All table entries were generated from the stated equations. No real instrument, experiment or scientific law was verified. The purpose is to teach valid transformations, not to turn invented observations into evidence.
A student who understands linearisation learns a transferable lesson: the same relationship can become easier to analyse when the variables are represented in a mathematically useful way.