A quantity increases by five percent every month. Does that mean it is increasing every second? Or does the change occur only when each month ends?
Those are different mathematical descriptions.
In one model, the quantity changes in separate steps. In another, it changes smoothly over time. The formulas may agree at the end of each month while giving different answers between those dates.
This is a useful reason to study Additional Mathematics: students learn to connect repeated changes with continuous rates and to recognise what a model assumes about time.
The distinction appears in compound growth, graphs, derivatives, integration and mathematical modelling. It also explains why a threshold may be reached at a fractional time in one model but only after a whole number of completed periods in another.
This article develops those ideas using familiar A-Math functions and calculus, with some discrete-versus-continuous comparisons included as conceptual enrichment rather than claims about separately examinable topics.
What does discrete change mean?
A discrete model describes values at separated inputs or events.
For example, consider a quantity recorded only at the end of each completed month.
Let A_n represent its value after n months, where n is a non-negative integer.
Suppose the initial value is:
A_0 = 100.
If it increases by five percent at the end of each month, the values are:
A_1 = 105.
A_2 = 110.25.
A_3 = 115.7625.
Each new value is the previous value multiplied by 1.05.
We can write the repeated rule as:
A_(n+1) = 1.05 A_n.
This is a discrete recurrence. Its input n counts complete periods.
It does not by itself specify a value halfway through a period.
Turn the repeated rule into a formula
Starting from A_0 = 100:
After one period:
A_1 = 100(1.05).
After two periods:
A_2 = 100(1.05)².
After three periods:
A_3 = 100(1.05)³.
The general formula is:
A_n = 100(1.05)^n,
where n = 0, 1, 2, 3, …
This is a geometric progression in discrete time.
The formula looks like an exponential function, and it can be extended mathematically to real-number inputs.
But the original discrete model assigns meaning specifically to completed periods.
A student should not assume that substituting n = 1/2 automatically describes what physically happens halfway through the first month.
That interpretation requires an additional assumption.
What does continuous change mean?
A continuous mathematical model assigns a value across a continuous range of time inputs, rather than only at separate observation points.
Consider:
A(t) = 100e^(kt),
where t is a real-valued time variable measured in months and k is a constant with units of inverse months.
This function is smooth and differentiable.
At every t, it has a well-defined instantaneous rate of change.
Differentiate:
A'(t) = k100e^(kt).
Therefore:
A'(t) = kA(t).
The rate is proportional to the current amount.
That is a characteristic of continuous exponential growth.
Unlike the discrete recurrence, the continuous model describes intermediate times such as t = 0.5 months.
Match a continuous curve to a discrete growth factor
Suppose we want our continuous model to agree with the discrete five-percent model after each complete month.
We require:
A(1) = 105.
Using A(t) = 100e^(kt):
105 = 100e^k.
Divide by 100:
e^k = 1.05.
Take natural logarithms:
k = ln 1.05.
Numerically:
k ≈ 0.04879 per month.
The matching continuous model is:
A(t) = 100e^((ln 1.05)t).
Because e^(ln 1.05) = 1.05, we can write the same mathematical function as:
A(t) = 100(1.05)^t.
For integer t = n, it agrees exactly with the discrete formula.
Between integers, it describes a smooth interpolation that the original discrete rule did not necessarily specify.
Why five percent is not the same as k = 0.05
This distinction is easy to miss.
A five-percent increase over one complete period means multiplication by 1.05.
In a continuous exponential model written A(t) = A_0e^(kt), choosing k = 0.05 gives a one-period multiplication factor:
e^0.05.
That factor is approximately 1.05127, an increase of about 5.127 percent.
It is not exactly 1.05.
To match a five-percent effective increase per period, use:
k = ln 1.05,
approximately 0.04879.
The difference is small for this example, but mathematically real.
A-Math helps students avoid replacing one rate convention with another without checking the formula.
Three possible values halfway through a month
Let the starting value be 100 and the end-of-month value be 105.
Different models give different possible intermediate descriptions.
Step-change interpretation: If the quantity changes only at the month-end event, it remains 100 before the event. At half a month it is still 100.
Linear interpolation: If we assume a straight increase from 100 to 105 across the month, the halfway value is 102.5.
Exponential interpolation: If we assume smooth proportional growth matching the endpoint factor 1.05, the halfway value is:
100(1.05)^(1/2),
approximately 102.47.
All three interpretations agree at the start and may agree at the end.
But they describe different behaviour between those times.
The original monthly growth statement alone does not necessarily choose among them.
Why an equation needs a time interpretation
A student might write A(t) = 100(1.05)^t for every real t without explanation.
Mathematically, that is a valid positive exponential function.
But if the original scenario describes a payment, review or adjustment made only at month end, continuous growth between months may not be what actually occurs.
Conversely, some physical processes do change continuously and can be modelled approximately by exponential functions.
The appropriate interpretation depends on the real conditions.
Additional Mathematics provides the mathematical language. The modelling task requires deciding which language fits the question.
A correct equation can still be the wrong description of a process if its assumptions are misunderstood.
The threshold problem in discrete time
Suppose the amount begins at 100 and grows by five percent at the end of each completed period.
How many completed periods are needed before the amount reaches at least 150?
Use:
100(1.05)^n ≥ 150.
Divide by 100:
(1.05)^n ≥ 1.5.
Take natural logarithms:
n ln 1.05 ≥ ln 1.5.
Since ln 1.05 is positive, dividing preserves the inequality direction:
n ≥ ln 1.5 / ln 1.05.
The right-hand side is approximately 8.31.
But n counts completed periods and must be a non-negative integer.
Therefore the first qualifying period is n = 9.
The mathematical threshold and the discrete decision must be distinguished.
Check the threshold with nearby periods
At n = 8:
A_8 = 100(1.05)^8.
This is approximately 147.75, still below 150.
At n = 9:
A_9 = 100(1.05)^9.
This is approximately 155.13, above 150.
Therefore period 9 is the first completed period meeting the target.
This check is useful because students sometimes round 8.31 to the nearest whole number, obtaining 8.
But a threshold requiring at least a target cannot be met by a period that still falls below it.
The required rounding direction follows from the meaning of the question, not a generic decimal-rounding rule.
The continuous threshold is different
In the matching smooth exponential model:
A(t) = 100(1.05)^t,
the equality A(t) = 150 occurs at:
t = ln 1.5 / ln 1.05,
approximately 8.31 months.
This is a valid answer if the question describes a continuous model and asks for the mathematical crossing time.
In the step-change model, the quantity changes only at completed periods, so it first reaches the target at period 9.
The two models agree at every integer period, yet their threshold descriptions are different.
That is not an algebraic contradiction.
The distinction comes from what happens between the completed periods.
Students learn to interpret a formula rather than report a decimal without context.
Compare fixed increments with fixed percentages
A discrete model need not be exponential.
Suppose a quantity begins at 100 and increases by five units at the end of every period.
Then:
B_n = 100 + 5n.
This is a discrete linear model.
After eight completed periods, B_8 = 140.
After ten, B_10 = 150.
The first qualifying period for reaching 150 is n = 10.
By contrast, the compound five-percent model reaches at least 150 after nine completed periods.
The models differ because one adds a fixed number and the other multiplies by a fixed factor.
Both may be called “growth” in ordinary language, but mathematically they are different.
Additional Mathematics trains students to identify the actual relationship from the description.
The same early observations can mislead
A fixed increase of five units and a five-percent increase on an initial amount of 100 produce the same first update.
Both give 105 after one period.
After two periods, however:
The additive model gives 110.
The percentage model gives 110.25.
The difference is small.
After many periods, it can become much larger.
A learner who sees only the first two values may not know which model is intended.
The correct interpretation requires information about whether the amount added stays fixed or the percentage factor stays fixed.
This is a useful caution about trying to infer a long-term model from too few observations.
A discrete difference is not the same as a derivative
For a sequence A_n, a one-period change is:
A_(n+1) – A_n.
For our five-percent model:
A_(n+1) – A_n = 0.05A_n.
That is an exact difference between successive completed-period values.
For the continuous exponential interpolation A(t) = 100e^(kt), the instantaneous derivative is:
A'(t) = kA(t),
with k = ln 1.05.
These are related but different statements.
One compares two separated period values.
The other describes an instantaneous rate in a smooth model.
Their numerical coefficients, 0.05 and ln 1.05, are not identical.
This is a central distinction between discrete differences and continuous derivatives.
A quadratic model makes the distinction visible
Consider the continuous function:
s(t) = t²,
where s is displacement in metres and t is time in seconds.
Its instantaneous velocity is:
v(t) = 2t metres per second.
Now compare consecutive whole-second positions.
The displacement during the interval from n to n + 1 seconds is:
s(n + 1) – s(n).
Calculate:
(n + 1)² – n².
Expand:
n² + 2n + 1 – n² = 2n + 1.
So the average velocity over each one-second interval is:
2n + 1 metres per second.
The instantaneous velocity at the interval’s beginning is 2n.
At its end it is 2n + 2.
These values are different.
The midpoint connection
For the quadratic displacement model s(t) = t², the midpoint of the interval from n to n + 1 is:
t = n + 1/2.
The instantaneous velocity there is:
v(n + 1/2) = 2(n + 1/2) = 2n + 1.
This equals the interval’s average velocity.
That agreement occurs because the velocity function v(t) = 2t is linear.
It should not be assumed for every displacement model.
The example helps students see how a discrete interval difference and a continuous derivative can be connected without being interchangeable at every instant.
A cubic shows why the midpoint shortcut is not universal
Let:
s(t) = t³.
Consider t from 1 to 3.
The average velocity is:
[s(3) – s(1)]/(3 – 1).
This equals:
(27 – 1)/2 = 13.
The instantaneous velocity is:
v(t) = 3t².
At the midpoint t = 2, it equals:
v(2) = 12.
The midpoint velocity is not the same as the interval’s average velocity.
The earlier result was a property of the quadratic displacement function, not a universal rule.
This is a useful mathematical habit: when a numerical pattern appears, investigate whether it follows from the structure or happens only for particular examples.
Riemann-style estimates from discrete samples
Suppose a continuous velocity function is:
v(t) = 2t,
with t measured in seconds and velocity in metres per second.
We want the net displacement from t = 0 to t = 2.
Integration gives the exact model result:
∫ from 0 to 2 of 2t dt = 4 metres.
Now suppose we only use velocity samples at whole seconds.
Over the first second, the start velocity is v(0) = 0.
Over the second, the start velocity is v(1) = 2.
Using each start velocity as though it were constant over its one-second interval gives:
0(1) + 2(1) = 2 metres.
That is a left-endpoint estimate.
It underestimates the exact displacement for this increasing velocity function.
A right-endpoint estimate
Use the velocities at the ends of the one-second intervals instead.
The first interval ends at t = 1, with velocity 2.
The second ends at t = 2, with velocity 4.
A right-endpoint estimate is:
2(1) + 4(1) = 6 metres.
This overestimates the exact displacement of 4 metres.
The interval estimates are different because the continuous velocity changes within each one-second interval.
Both use discrete samples, but neither exactly represents the changing rate across the interval.
This illustrates why integration is useful.
It accounts for the rate’s continuous variation rather than assuming each sampled value describes an entire interval.
A midpoint estimate
Now use the velocity at the midpoint of each one-second interval.
The midpoints are t = 0.5 and t = 1.5.
The velocities are:
v(0.5) = 1.
v(1.5) = 3.
The midpoint estimate is:
1(1) + 3(1) = 4 metres.
In this particular linear-velocity example, the midpoint estimate matches the exact integral.
That is a property of integrating a linear function over equal intervals with midpoint sampling.
For more complicated functions, a midpoint estimate may still have an error.
The educational benefit is understanding how a finite sum of samples can approximate a continuous accumulation.
It helps explain the conceptual connection between summation and definite integration.
More intervals can improve an approximation
For a smooth increasing function such as v(t) = 2t, dividing the interval into smaller equal pieces reduces the width of each approximation rectangle.
Left- and right-endpoint estimates move closer to the exact integral as the interval widths shrink.
The mathematical limit of appropriate sums is the definite integral.
This is part of the deeper background of calculus.
Secondary students need not master every formal definition of a Riemann integral to understand the picture.
They can see why a continuous total requires more information than one average sample and how refined sampling can improve an estimate.
The exact school assessment scope remains determined by the official syllabus.
A changing radius offers another comparison
Suppose a circular region has radius:
r(t) = 2 + 0.1t,
where r is in metres and t in seconds.
Its area is:
A(t) = π(2 + 0.1t)².
The instantaneous area-changing rate is:
A'(t) = 0.2π(2 + 0.1t).
At t = 0, the rate is:
0.4π square metres per second.
Now calculate the exact area change over the first complete second.
At t = 0, area is 4π.
At t = 1, radius is 2.1, so area is 4.41π.
The increase is:
0.41π square metres.
The rate at the beginning was 0.4π square metres per second.
It does not equal the exact one-second change, although it is close.
Explain the difference with algebra
Let the starting radius be r and the radial increase over a period be h.
The exact area difference is:
π[(r + h)² – r²].
Expand:
π(2rh + h²).
For our example r = 2 and h = 0.1:
ΔA = π[2(2)(0.1) + (0.1)²].
So:
ΔA = π(0.4 + 0.01) = 0.41π.
The derivative estimate based on the initial radius corresponds to the first term:
2πrh = 0.4π.
The extra πh² is the finite-change correction.
This example connects discrete differences, derivatives and binomial expansion.
It explains why the results are close but not identical.
When discrete counts are essential
Some quantities are inherently counted in whole units.
A classroom cannot ordinarily contain 12.4 physical chairs.
The number of pupils present is an integer.
The number of completed inspection cycles is also a count.
Mathematical models may use smooth formulas to estimate trends or choose approximate targets, but final decisions involving indivisible items may need whole-number restrictions.
For example, if a continuous optimisation produces 12.4 as an ideal quantity, a real choice may require comparing 12 and 13 using the original objective.
Ordinary rounding to the nearest integer is not always sufficient.
The optimum among allowed integers must be evaluated.
Why a continuous optimum can still be useful
Suppose a modelled quantity is:
R(q) = 100q – q²,
where q is a non-negative number of items.
As a continuous quadratic, its maximum occurs at q = 50.
That is also an allowed integer count, so the continuous and discrete optima agree.
Now modify the model:
R(q) = 101q – q².
The continuous maximum is at q = 50.5.
If only whole items are allowed, compare q = 50 and q = 51.
Evaluate:
R(50) = 5050 – 2500 = 2550.
R(51) = 5151 – 2601 = 2550.
Both allowed counts give the same maximum modelled value.
The continuous answer 50.5 is informative, but it is not a feasible integer decision.
This illustrates why the domain matters in optimisation.
A model may be discrete in input and continuous in output
Discrete versus continuous is not always an all-or-nothing description of a whole problem.
A formula can accept only integer counts while producing real-valued outputs.
For example:
C(n) = 2.5n,
with n a non-negative integer, gives a monetary quantity that can include cents.
Likewise, an integer period n can determine a compound growth amount 100(1.05)^n, which need not be an integer.
The input domain and output range are separate features.
Students should identify both instead of assuming that discrete inputs require whole-number outputs.
This distinction becomes useful when modelling money, measurements and counts in the same problem.
What a graph reveals about discrete and continuous models
A discrete sequence can be shown as separate points at integer inputs.
A step-change model may be represented by horizontal segments with jumps at event times.
A continuous exponential model appears as a smooth curve.
These graphs might share some points.
But connecting discrete observations with a smooth curve adds a modelling assumption about what happens between them.
A chart that displays a smooth line through monthly totals does not necessarily prove the underlying process changed smoothly every moment.
Mathematical graph literacy requires asking what was observed and what was interpolated.
That distinction is valuable beyond examination questions.
The time unit changes the model
Suppose a growth factor of 1.05 applies per month.
After 12 completed months, the factor is:
(1.05)^12.
It would be incorrect to treat 1.05 as an annual factor without conversion.
Similarly, a continuous model parameter k measured per month cannot be substituted into a formula with t measured in years unless the unit conversion is handled.
The exponent must represent the appropriate number of periods or a consistent dimensionless time-rate product.
Students should read the unit attached to the period before calculating.
Many modelling errors arise from using a correct formula with an incorrectly interpreted time variable.
Compare two time descriptions consistently
Let n count months.
The discrete model is:
A_n = 100(1.05)^n.
Now let T measure years.
Twelve months correspond to T = 1.
A matching continuous expression in years is:
A(T) = 100(1.05)^(12T).
Equivalently:
A(T) = 100e^[(12 ln 1.05)T].
At T = 1, both formulas give the same end-of-year factor (1.05)^12.
The annual continuous exponent coefficient differs from the monthly coefficient because the time unit changed.
This demonstrates how precise parameter definitions prevent confusion.
The equations are consistent when their units and periods are handled carefully.
Four common modelling mistakes
Mistake one: using a fractional period in a discrete event model without explanation
The student substitutes n = 0.5 into an end-of-month recurrence and treats the result as a guaranteed real value halfway through the month.
The repair is to distinguish the discrete definition from an optional continuous interpolation.
Mistake two: confusing five-percent growth with k = 0.05
The student uses e^(0.05t) and assumes it gives exactly five-percent growth over one unit of t.
The repair is to compare e^0.05 with 1.05 and derive k = ln 1.05 where matching is required.
Mistake three: rounding a threshold down
The calculation gives 8.31 completed periods, and the student reports eight periods even though the target is not yet reached.
The repair is to check the model value at the neighbouring whole periods.
Mistake four: treating an instantaneous rate as an exact whole-period increase
The student uses a derivative at the interval start as though it were constant throughout the interval.
The repair is to compare the derivative estimate with the exact finite difference.
Each mistake reflects a misunderstanding of how time and change are represented.
A five-session learning plan
Session one: build a discrete recurrence
Start at 100 and multiply by 1.05 at the end of each period.
Write the first four values.
Derive A_n = 100(1.05)^n for integer n.
Session two: create a continuous interpolation
Find k = ln 1.05.
Compare the step-change, linear and exponential values at half a period.
Explain why the original recurrence did not determine that intermediate value.
Session three: compare target times
Solve 100(1.05)^t = 150 with logarithms.
Explain the smooth crossing time and the first whole completed period that meets the target.
Session four: compare derivatives and differences
Use s(t) = t².
Find the one-second change from n to n + 1 and the instantaneous derivative.
Explain why the midpoint relation holds for this model.
Session five: connect sums and integration
Use v(t) = 2t over 0 ≤ t ≤ 2.
Calculate left, right and midpoint sample estimates.
Then integrate to obtain the exact displacement.
The goal is to understand how separated observations and continuous rates relate.
How parents can support the reasoning
Parents do not need to teach formal calculus to ask the key questions.
“Does this model change only at completed periods, or all the time?”
“What does n count?”
“What happens at half a period?”
“Why did we round a threshold time upward?”
“Does this derivative describe a moment or an entire interval?”
“Is the quantity something we can count only in whole numbers?”
These questions help students interpret the model before manipulating the formula.
A learner who can explain the time assumption is much less likely to produce a plausible-looking but conceptually incorrect answer.
Singapore Additional Mathematics context
The Singapore-Cambridge Secondary Education Certificate G3 Additional Mathematics syllabus K341 includes functions, exponential and logarithmic relationships, differentiation and integration.
The discrete-recurrence and numerical-sampling comparisons in this article are conceptual extensions linking those skills. They should not be interpreted as claims that every method shown is separately assessed in G3 Additional Mathematics.
Students should use the official SEAB 2027 G3 syllabus directory for precise current requirements and follow their school’s topic sequence.
The numerical situations here are illustrative models, not forecasts of actual investments, populations or physical systems.
How small-group tuition can help
One learner may know how to calculate compound growth but use an unsuitable time unit.
Another may solve a logarithmic threshold correctly and choose the wrong whole-number period.
A third may understand derivatives yet confuse instantaneous rate with average change.
They need different explanations.
In a three-student Additional Mathematics tutorial, a tutor can ask what the input variable represents, inspect the model’s assumptions and then guide a focused corrective example.
The immutable Secondary 1 Mathematics small-group tutorial in Clementi illustrates why early number sense, equations and mathematical communication matter before advanced modelling.
Students progressing well independently may not require tuition. Additional support is useful when an identifiable misconception repeatedly interrupts their work.
Frequently asked questions
Is a discrete exponential model the same as a continuous exponential model?
They can agree at chosen integer time points, but the continuous model defines additional intermediate values and instantaneous rates that the discrete rule alone may not specify.
Why is ln 1.05 used for matching continuous growth?
Because e^(ln 1.05) = 1.05, so the smooth exponential matches the five-percent one-period multiplication factor.
Does a continuous crossing time of 8.31 mean eight whole periods are enough?
No. If the change occurs only after complete periods, the first qualifying period may be nine. Check neighbouring completed-period values.
Is an average change the same as a derivative?
No. An average compares two separated inputs, while a derivative describes the local instantaneous rate for a differentiable function.
Can a continuous optimisation give an impossible practical answer?
Yes. If the decision involves whole items, the continuous optimum may be fractional. Evaluate suitable allowed integers.
Does a smooth graph through monthly points prove smooth real growth?
No. The connecting curve may be a chosen interpolation rather than an observation of the process between measurements.
Why learn both descriptions?
Because different questions require different assumptions. Recognising whether a process is discrete or continuous helps students select valid equations and interpret results responsibly.
The lasting value of knowing how change occurs
Additional Mathematics teaches students to reason about change in more than one way.
A recurrence compares successive completed periods.
An exponential function can interpolate smoothly between them under an added assumption.
A finite difference describes what changed over an interval.
A derivative describes the instantaneous rate at a point.
A definite integral can accumulate a continuous rate over an interval.
These ideas are connected, but they are not interchangeable.
A student who understands the distinctions can choose the right model, interpret a threshold accurately and recognise the limits of a calculation.
That is a strong reason to study Additional Mathematics.
For connected reading, explore Why Study Additional Mathematics | Use Logarithms to Find How Long Growth Takes, Why Study Additional Mathematics | Understand Sensitivity to Small Changes, and the eduKateSG Additional Mathematics project library.