Did you know that a smooth line on a graph can contain values nobody actually measured? A pupil plots a plant’s height every four days, draws a line and confidently reads the height on another day. That reading may be a sensible estimate—but it is not the same thing as an observation, and extending the line beyond the observations is a different kind of guess.
Interpolation means estimating a value between known data points. Extrapolation means estimating outside the observed range. Interpolation uses nearby evidence but still needs a defensible model; extrapolation reaches beyond the measured range and usually needs more caution. To tell them apart, find the observed smallest and largest x-values before reading or extending a line.
First, diagnose what the learner is reading
A graph question can go wrong in at least four ways. The pupil might misread the axis scale, mistake a drawn trend line for a series of direct measurements, use a straight line when the pattern curves, or extend the pattern outside the measured range without naming the assumption. These are different errors and deserve different feedback.
Before providing a formula, ask: “Which x-values did the experiment measure? Is the number you are looking for between them or outside them? How is the line between points being chosen?” If the learner cannot answer, work on the evidence boundary first rather than on arithmetic speed.
A worked example: how tall was the classroom plant?
Consider a fictional student experiment recorded in a table. Height was measured as follows: day 0, 2 cm; day 4, 6 cm; day 8, 10 cm; and day 12, 12 cm. The data are deliberately invented to demonstrate graph reasoning and should not be treated as a normal growth rate or real eduKate pupil results.
Suppose the question asks for the plant’s estimated height on day 6. Day 6 lies between measured days 4 and 8. If we choose straight-line interpolation for this interval, the change is 10 − 6 = 4 cm over 8 − 4 = 4 days, giving a modelled increase of 1 cm per day. Two days after day 4, the estimate is 6 + (2 × 1) = 8 cm.
Eight centimetres is a valid calculation under that linear assumption. It is not a height actually observed on day 6. If the plant experienced irregular growth, the true value could differ from the straight-line estimate. A sensible answer says “about 8 cm, assuming roughly linear change between the day 4 and day 8 readings.”
What changes when the question asks about day 20?
Day 20 sits outside the observed range of day 0 to day 12. That makes it an extrapolation. A student might take the final measured interval, day 8 to day 12, where the height increased from 10 cm to 12 cm in four days. Continuing that last interval’s straight-line rate gives 0.5 cm per day, so day 20 would be modelled as 12 + (8 × 0.5) = 16 cm.
But another student might extend the earlier day 4–8 slope of 1 cm per day from day 8 all the way to day 20, giving 10 + (12 × 1) = 22 cm. Both operations are arithmetically correct for their chosen straight-line rules, yet they give different predictions. Neither figure is established by the measurements. The plant might slow, plateau, suffer damage or grow differently after day 12.
That contrast is the lesson. A graph is evidence about recorded data, while a model is a set of assumptions about what happens at other points. The further we move outside the observed range, the more we rely on assumptions that need independent checking. An examination question may specify a line of best fit or a model; use the one actually given instead of inventing a new rule.
A four-step graph boundary check
- Identify the measured domain: underline the first and last x-values actually observed, including their units.
- Classify the target: inside the observed domain means interpolation; outside means extrapolation; at a plotted measured point means direct reading.
- Name the model: straight-line segment, fitted curve, explicitly given equation or another justified relationship.
- Label the result honestly: measured value, estimated value or model prediction, with a sensible unit and precision.
These questions are useful in Mathematics, Science, Geography and school project data. In some graphs, the line itself is a prescribed idealisation—such as a distance–time exercise with constant speed—while in others it is merely a visual guide between sparse observations. The instructions and subject context determine what interpolation can reasonably mean.
Progressive practice that teaches the distinction
1. Recognise, without calculating
Offer a table measured at 1, 3 and 5 minutes, then ask students to classify requests for 3 minutes, 4 minutes and 7 minutes. Correct categories: direct reading, interpolation and extrapolation. The goal is to locate the evidence boundary, not to guess a number.
2. Explain an estimate before computing it
Return to the plant data. The student should first state that day 6 sits inside a measured interval and that a straight-line assumption is being used. Only then calculate 8 cm. This order makes the method visible and prevents a copied procedure from masquerading as understanding.
3. Compare two plausible models
Ask the learner to estimate day 20 using the last segment and then the earlier segment. The differing answers, 16 cm and 22 cm, are not grounds to pick whichever looks prettier. They show that the forecast is model-dependent. What further measurements would help? Measurements after day 12, preferably with consistent methods and conditions.
4. Transfer to another subject independently
Use a fictional cooling experiment at 0, 5 and 10 minutes. Ask whether 7 minutes or 30 minutes requires extrapolation. The learner should classify first and explain why predictions at 30 minutes are more sensitive to a model’s behaviour beyond the observed data. A correct answer without a borrowed diagram is an encouraging check.
Common errors, and what to do instead
- Calling a plotted line a measurement: mark actual data points differently from the fitted or connecting line.
- Reading the wrong x-interval: inspect uneven tick marks and units before calculating any slope.
- Assuming every graph is linear: look for bends, plateaus or a subject-specific model.
- Reporting excessive precision: an estimated height of 8.0000 cm is misleading when the original readings are approximate centimetres.
- Using an extrapolated number as a fact: label it as a prediction and explain what evidence would test it.
- Ignoring a model supplied by the question: use the stated conditions rather than choosing an unrelated trend.
What parents and teachers should look for
A useful five-minute home conversation begins with “Show me which points were measured.” Invite the child to draw circles around recorded points, a light line between them, and a dashed continuation beyond them. Ask which part rests most heavily on an assumption. The child does not need sophisticated statistics to understand the difference between seeing and projecting.
Teachers can keep an initial answer and a revised one. In a stronger response the student does three things independently: identifies the observed range, chooses or states the model appropriate to the question, and distinguishes estimates from observations. A graphing-tool screenshot alone cannot demonstrate those judgements.
If a pupil consistently confuses extrapolation with interpolation, give classification practice before numerical problems. If they classify correctly but calculate the wrong slope, return to change in y divided by change in x and the units. If the slope is right but the conclusion overstates certainty, practise wording and model limitations rather than assigning more arithmetic.
Limits: even interpolation can be unsafe
Interpolation is not automatically accurate. It can fail across abrupt changes, missing categories, nonlinear relationships or a discontinuity. Extrapolation is not automatically forbidden either: a sound physical model with checked conditions can support useful forecasts. The distinction is about evidence and assumptions, not a rule that one type of estimate is always good and the other always bad.
This article explains general graph reasoning and does not prescribe the marking requirements of a particular MOE or SEAB examination paper. If a question specifies rounding, graph-reading conventions or the model to use, follow the current paper’s instructions.
Continue learning with connected resources
For another common graph trap, see the Core Aim article about two graph axes with different scales. The guide to measurement resolution and real change helps when differences are smaller than the tool can reliably show. Continue through the eduKate Education Hub, or read the Clementi Mathematics teaching reference for explanation-first and checking-first tuition practice.
Frequently asked questions
Is drawing a straight line between points always interpolation?
Only if it is being used to estimate between observed points—and a straight line is an appropriate assumption. Connecting points for visibility does not prove the real process changed linearly.
Can extrapolation be correct?
A prediction can later match reality. What makes it well supported beforehand is a tested model and relevant evidence, not the confident appearance of an extended line. An accurate guess by chance is not a proof of method.
What if the graph already has a line of best fit?
Use the fitted line according to the task, noting whether your target is inside or beyond the data range. A fitted line is still a model, and unusually far extrapolation may be unreliable even when arithmetic is neat.
Learn where the evidence ends
The central habit is to ask one small question before reading a value: “Was this point observed, estimated between observations, or projected beyond them?” That distinction helps students become capable graph readers rather than simply confident line followers.