A graph can tempt a learner to extend a line long after the evidence has ended. This workshop trains the distinction between reading within the measured range and predicting beyond it—without turning every estimate into false precision.
It follows Vol 0119 and the G3 SEC learner route. For 2027 school candidates, use the official K326/K327/K328 combined Science syllabus and SEAB G3 directory for current requirements. All datasets below are original teaching examples.
Interpolation stays within the measured range
Interpolation estimates a value between measured or plotted values. It uses the local pattern inside the observed range. The learner should still inspect whether the graph is approximately linear, curved or irregular before choosing how to estimate.
Extrapolation goes beyond the measured range
Extrapolation extends a trend outside the observed data. This can be useful, but it carries more risk because the relationship may change beyond the range tested. The farther the extension, the more cautious the conclusion should become unless the model is strongly justified.
The first question is whether the relationship is suitable
Do not interpolate or extrapolate automatically from any two points. Ask whether the data suggest a smooth relationship, whether the independent variable changes continuously, and whether the same scientific conditions remain valid across the range.
Worked case 1: interpolation on a straight trend
A fictional experiment records a measured quantity of 12 units at 20°C and 20 units at 40°C, with a straight-line trend between those points. A simple linear interpolation at 30°C gives 16 units. This estimate depends on the stated linearity between the measured values.
Show the interpolation fraction
30°C is halfway from 20°C to 40°C, so the estimated value is halfway from 12 to 20. Writing the fraction or using the graph scale makes the reasoning visible. The learner should not guess from visual position alone when the axes can be read precisely.
Worked case 2: interpolation with unequal spacing
Suppose a graph gives 10 units at x = 2 and 22 units at x = 8, with a straight segment between them. To estimate at x = 5, note that 5 is 3/6 of the way from 2 to 8, so the estimate is 10 + 3/6 × 12 = 16.
Do not interpolate linearly across a curve without thought
If the graph curves strongly, a straight proportional estimate between distant points may be poor. Use the drawn curve or the model specified by the question. The source representation controls the estimation method.
Worked case 3: a curved response
Imagine enzyme activity rises from 20°C to 40°C and falls by 60°C. Estimating at 50°C requires the actual curve or local data; a straight line from 20°C to 60°C would ignore the peak. The learner should read local structure, not average across a changing mechanism.
Interpolation does not create new precision
If the graph scale supports reading to the nearest unit, do not report an estimate to four decimal places. The precision of the answer should reflect the information source and the task requirement.
Extrapolation needs a model, not optimism
A line that rises through the measured range does not guarantee indefinite increase. Physical limits, saturation, depletion, structural change or another mechanism may alter the trend. Extrapolate only as far as the question and model reasonably support.
Worked case 4: short extrapolation from a linear calibration
A calibration graph is stated to be linear over a validated range from 0 to 10 units. Estimating at 9.5 is interpolation; estimating at 10.5 is extrapolation. Even though the extension is small, it is beyond the validated range and should be treated differently.
Validated range matters
If a calibration curve is only established within a stated interval, values outside it should not be assumed to obey the same relationship. The learner may need to report that the unknown is beyond the calibration range rather than inventing a precise value.
Worked case 5: reaction graph plateau
A gas-volume graph rises and then reaches a plateau. Extrapolating the early steep gradient as a straight line would predict volumes beyond the plateau and violate the observed mechanism. Use the part of the graph relevant to the stage being analysed.
Stage changes break naive extrapolation
Many Science graphs have phases: heating, boiling, cooling; acceleration, constant speed, deceleration; growth, saturation, decline. A trend from one phase should not be extended automatically into another.
Use the axes before the trend
Interpolation on a graph requires accurate reading of axis quantity, unit, interval and scale. A non-zero axis or uneven scale can make visual estimates misleading. The first task is to understand the coordinate system.
Worked case 6: concentration calibration
A fictional absorbance-like reading is 0.20 at concentration 2 units and 0.50 at 5 units, with a stated linear calibration. The gradient is 0.10 reading units per concentration unit. A reading of 0.35 corresponds to concentration 3.5 units.
Reverse interpolation is still translation
In the calibration example, the measured reading is used to infer the independent quantity. The learner should know which axis is being read and which value is being estimated. Reversing axes changes the numerical gradient and interpretation.
A best-fit line differs from joining every point
Experimental data often scatter. A best-fit line or curve represents the overall pattern rather than passing through every observation. Interpolation from that line estimates the trend, not a guarantee of a future exact reading.
An anomalous point should not define the trend alone
If one point lies far from the others, investigate whether it is an anomaly or meaningful change. Do not force a best-fit line through it simply because it exists. Equally, do not delete it automatically without justification.
Worked case 7: repeated measurements around a line
Suppose repeated readings at x = 4 are 7.9, 8.1 and 8.0, while the best-fit line predicts 8.0. The line captures the central trend. A future reading can still vary around it. Interpolation estimates the modelled relationship, not random measurement noise.
Interpolation and causation are separate
A smooth graph can support an estimated value between measurements, but it does not by itself prove the cause of the relationship. Description, estimation and explanation are different tasks.
Extrapolation and prediction are separate from certainty
A prediction beyond the data should be framed as model-dependent. “If the linear trend continues, the value would be…” is more accurate than claiming the value will definitely occur when the question leaves the mechanism uncertain.
Worked case 8: light intensity and distance
Suppose a simplified model or graph is provided for light intensity over measured distances. If the relationship is not stated as linear, do not interpolate with a straight-line rule automatically. Use the graph or inverse-square model only if the question establishes it.
Worked case 9: cooling data
Temperature is 75°C at 0 min, 65°C at 5 min and 58°C at 10 min. The drops are not equal, so assuming a constant linear rate may be poor. Interpolation between 5 and 10 minutes can use the local graph if drawn; extrapolation far beyond 10 minutes needs a cooling model or caution.
A decreasing gradient can signal a changing process
If a curve becomes flatter, the rate of change is decreasing. Extending the original steep gradient ignores that information. The graph shape itself is evidence about whether a simple extrapolation is defensible.
Worked case 10: population-style growth curve
A fictional microbial population rises rapidly and then levels off. Interpolating within the measured phase can be reasonable; extrapolating exponential growth indefinitely beyond the plateau is not. The observed limit changes the model.
The farther the extrapolation, the larger the model risk
A prediction just beyond the range may be more defensible than one many times farther away, but distance alone does not decide validity. The governing mechanism and model conditions matter. State the assumption when the extension is part of the reasoning.
Do not hide extrapolation inside a calculation
A learner may fit a line and substitute x = 100 even though data stop at x = 10. The arithmetic can be correct while the scientific use of the model is weak. Always compare the requested input with the observed range.
Range checking should happen before substitution
Before using a calibration equation, state the validated or measured range. This quick check prevents false precision and tells the learner whether the answer is interpolation, extrapolation or out-of-range.
Worked case 11: interpolation from a table
A table records 4.0 units at t = 2 min and 7.0 units at t = 5 min, with a stated constant rate between. At t = 4 min, the estimate is 6.0 units. The phrase constant rate justifies linear interpolation.
Worked case 12: no interpolation for categories
If the independent variable is a category such as material A, B and C, there is no meaningful halfway material between A and B unless the question defines one. Interpolation requires an ordered or continuous quantity, not arbitrary labels.
Interpolation on a bar chart can be meaningless
Bars often represent discrete categories. Drawing a line between bar tops and estimating an intermediate category may invent a quantity the graph never represents. Match the method to the data type.
Extrapolation on a histogram needs care
A histogram describes distribution across intervals, not a functional relationship to extend beyond the range. Do not treat every graphical display as a curve to be projected.
Graph-reading task versus model-building task
If the question says “use the graph to estimate”, follow the graph. If it asks for a model or relationship, you may need to form an equation. The response form is controlled by the task.
Worked case 13: choosing an estimate from a curve
A curve passes through about y = 30 at x = 4 and y = 42 at x = 6, bending upward. At x = 5, a straight midpoint of 36 may not match the drawn curve exactly. Read the curve where it crosses x = 5 instead of assuming linearity.
Use local slope to think about shape
On a curve, gradient changes with x. A short-range estimate can follow the local trend more closely than a long-range straight-line projection. This is a conceptual tool, not a requirement to use calculus.
Worked case 14: threshold from interpolation
A graph shows a concentration threshold at reading 0.60. If nearby calibration points bracket 0.60, interpolation can estimate the concentration where the threshold occurs. State that it is an estimate from the graph, not an exact measured value.
Threshold conclusions should respect uncertainty
If the threshold estimate lies very close to a decision boundary and the graph has scatter, avoid presenting the classification as more certain than the data allow. The learner should use the precision and variability shown in the source.
Interpolation can be checked algebraically
If a straight-line equation is known, substitute the intermediate x-value and compare with the graph estimate. Agreement between graph and equation is useful error control.
Extrapolation can be checked for physical plausibility
Before accepting a projected value, ask whether the sign, range, maximum, minimum or conservation constraints make sense. A model predicting negative mass or probability above 1 is warning evidence.
Worked case 15: impossible extrapolation
A straight trend predicts concentration -2 units at a later time. Concentration cannot be negative in this context. The model has been extended beyond a physically meaningful region, so the negative result signals model failure rather than a real quantity.
Independent task A
A straight-line graph has y = 10 at x = 2 and y = 22 at x = 8. Estimate y at x = 5 and explain why this is interpolation.
Independent task B
The same straight trend is observed only from x = 2 to x = 8. A learner predicts y at x = 30 using the same line. Explain what additional assumption is being made.
Independent task C
A graph rises steeply, then levels off. Explain why extrapolating the early steep section beyond the plateau is inappropriate.
Independent task D
A calibration graph is valid from 0 to 100 mg/L. An unknown gives a reading corresponding to 130 mg/L if the line is extended. What should the learner say about the estimate and range?
Independent task E
Three repeated measurements scatter around a best-fit line. Explain what the line represents and why an interpolated value is not guaranteed to equal a future single measurement.
Independent task F
A bar chart shows values for three materials. Explain why estimating a value halfway between Material A and Material B may be meaningless.
Worked feedback A
x = 5 is halfway between 2 and 8, so y is halfway between 10 and 22: y = 16. The estimate lies inside the observed interval, so it is interpolation.
Worked feedback B
The learner assumes the same linear relationship continues far beyond the measured range. The arithmetic follows the line, but the scientific prediction is only as valid as that assumption.
Worked feedback C
The system changes behaviour as the graph levels off. The early gradient no longer represents the later stage, so extending it ignores evidence already shown by the curve.
Worked feedback D
The inferred value is outside the validated calibration range. Report that the reading is beyond the calibrated interval or that extrapolation would be required, rather than presenting 130 mg/L as an equally reliable interpolated value.
Worked feedback E
The best-fit line represents the overall trend through variable measurements. An interpolated point estimates the trend at that x-value. A future observation can differ because of random variation and measurement uncertainty.
Worked feedback F
The bars represent categories, not a continuous numerical x-axis. Without a defined intermediate material or ordered variable, there is no meaningful halfway x-value to interpolate.
How to review a graph estimate
Check axes, units, scale, observed range, shape and whether the target lies inside or outside that range. Then decide whether the task calls for interpolation, extrapolation or simply reading a plotted value.
How to review an extrapolated conclusion
State the model being extended, the condition that must continue, and any physical or biological limits visible in the context. Use cautious wording if the evidence beyond the observed range is not established.
Repair route
Begin with straight-line interpolation between measured points. Make the learner identify inside versus outside range before calculating. Only then add curved graphs and model limits.
Stabilisation route
Mix graphs where interpolation is valid, where the curve must be read directly, and where categories make interpolation meaningless. The learner should classify the representation first.
Extension route
Use calibration limits, changing gradients, thresholds and physical constraints. Ask learners to explain not just the estimate, but why the chosen method is defensible over that range.
What progress should look like
Progress is visible when the learner checks the observed range before estimating, distinguishes interpolation from extrapolation, uses graph shape rather than automatic straight lines, and qualifies predictions beyond the data.
Final operating rule
Inside the measured range, estimate from the local relationship the source actually shows. Beyond the measured range, state the model you are extending and check whether the same conditions can reasonably continue. A neat line is not permission to predict forever.
Science interpolation and extrapolation checklist
- read axes, units and scale
- identify the measured range
- decide whether the target lies inside or outside it
- inspect line versus curve shape
- avoid false precision
- state assumptions behind extrapolation
- check physical plausibility and limits
- separate estimate from causal explanation
Advanced interpolation, calibration and model-limit laboratory
Advanced case: interpolation across a discontinuity
If a graph jumps suddenly because the system changes state or the rule changes, estimating across the jump with a smooth line is invalid. The learner should identify the discontinuity and use the appropriate branch or stage rather than averaging across two different regimes.
Advanced case: phase-change plateaus
During a phase change, temperature may remain nearly constant while energy continues to transfer. A learner who extrapolates the earlier rising temperature trend through the plateau ignores a change in the process. The graph shape signals a new regime.
Advanced case: saturation limits
Sensor response, enzyme activity, population growth and other systems can approach a maximum. A straight-line extrapolation from low values may predict impossible outputs beyond the saturation limit. The learner should ask whether the model has a known upper bound.
Advanced case: detection limits
A calibration graph can be linear over a central range but unreliable near very small signals. If a reading falls below the stated detection or reliable measurement range, interpolation may be numerically possible yet scientifically weak. Range validity is more than axis coverage.
Advanced case: overloaded sensors
At the high end, a measuring system may saturate and produce nearly constant readings even while the true quantity continues increasing. A plateau caused by instrument limit differs from a plateau caused by the underlying process. Context matters.
Advanced case: hysteresis or path dependence
Some systems do not follow the same path when a variable increases and decreases. If the question shows separate curves, do not interpolate between them as though one universal relationship exists. The system history may matter.
Advanced case: logarithmic or nonlinear axes
If an axis is logarithmic, equal visual spacing does not represent equal numerical increments. Read labels carefully before interpolating. A midpoint on the page may correspond to a multiplicative midpoint rather than an arithmetic average.
Advanced case: broken axes
A graph with a broken or truncated axis can exaggerate visual change. Interpolation still uses the actual numerical scale, not the apparent height difference. The learner should read labels before judging magnitude.
Advanced case: uncertainty bands
If a graph includes error bars or uncertainty bands, an interpolated central value should not be treated as exact. The uncertainty information shows a range of plausible values around the trend. Use it when the task asks about confidence or overlap.
Advanced case: overlapping error bars
Overlapping uncertainty intervals do not automatically prove two conditions are identical, and non-overlap does not by itself answer every statistical question. For school-level interpretation, use the evidence and instructions provided rather than importing unsupported significance claims.
Advanced case: threshold within a curve
Suppose a safe-operating threshold is y = 50 and a best-fit curve crosses it near x = 7.4. That x-value is an estimate from the graph. If the curve is thick or the scale coarse, reporting x = 7.4000 would create false precision.
Advanced case: threshold beyond the data
If the threshold lies outside the measured x-range, the learner should recognise that locating the crossing requires extrapolation. The answer may be model-dependent or not justified, depending on what the question supplies.
Advanced case: local versus global fit
A dataset can be approximately linear over a small interval but curved over the full range. Interpolation within the local interval may be reasonable, while extrapolating the local line across the whole domain is not.
Advanced case: calibration by two standards
Using only two standards can define a line, but it does not test whether the relationship is actually linear between or beyond them. More calibration points can reveal curvature or saturation. The learner should distinguish defining a line from validating a model.
Advanced case: repeated calibration points
If repeated standard measurements scatter around the fitted line, the learner should treat the line as an estimate of the underlying relationship. Interpolation still carries measurement variability even though the line itself is smooth.
Advanced case: zero intercept assumption
A calibration line should not be forced through the origin unless the method or question justifies it. A background signal can produce a non-zero intercept. Forcing the wrong intercept changes every interpolated value.
Advanced case: baseline correction before interpolation
If a known background reading must be subtracted, perform that correction consistently before using the calibration model. Mixing corrected and uncorrected values can create a false trend that no interpolation method can repair.
Advanced case: inverse calibration
Sometimes the measured signal is on the horizontal axis and the unknown concentration is on the vertical axis, or vice versa. Read the graph orientation before estimating. The physical meaning of gradient and interpolation direction depends on which quantity occupies each axis.
Advanced case: categorical endpoints
If an outcome is pass/fail, colour A/colour B, or another discrete category, ordinary numerical interpolation may not make sense. The learner should respect the measurement scale rather than force continuous methods onto categorical data.
Advanced case: count data
Counts can be discrete even when plotted against a continuous variable. An interpolated expected count may be fractional as a model value, but an actual observed count must be whole. Distinguish trend estimate from single outcome.
Advanced case: monotonicity check
Before interpolating, check whether the local curve moves consistently in one direction. If it oscillates or turns around, a single y-value may correspond to more than one x-value. Reverse interpolation can then be ambiguous.
Advanced case: two possible x-values
A bell-shaped response can cross the same y-level on the rising and falling sides. If asked to infer x from y, there may be two possible answers. The graph shape determines whether the inverse reading is unique.
Advanced case: sparse data
With only a few widely spaced points, the shape between them is less certain. Connecting them with a smooth line is a modelling choice. The learner should not treat the visual connection as directly measured evidence.
Advanced case: dense local data
More points close to the target can make interpolation more defensible because the local shape is better observed. Data density near the estimate matters, not only the total number of points on the graph.
Advanced case: mechanism supports, but does not guarantee, extrapolation
A well-understood mechanism can strengthen confidence that a trend continues over some range. But physical limits and changing conditions still matter. Scientific prediction should combine data pattern and model conditions.
Advanced case: conservation can reject extrapolation
If an extrapolated value violates conservation of mass, energy or another stated constraint, the model has been pushed beyond its valid range or formed incorrectly. Constraints are powerful plausibility checks.
Advanced case: biological limits
A model predicting negative population, activity above a defined maximum or concentration below zero is physically or biologically impossible. Such outputs reveal model breakdown rather than unusual real values.
Workshop drill: classify before estimating
For each graph, label the task as direct read, interpolation, extrapolation, threshold estimate or not meaningful. This prevents automatic calculation and builds model awareness.
Workshop drill: state the range
Before any graph estimate, write the observed x-range. Then mark whether the target lies inside or outside. This one habit sharply reduces accidental extrapolation.
Workshop drill: describe the shape
Use words such as linear, curved, plateauing, increasing at a decreasing rate, or changing direction. Shape description should come before choosing an interpolation method.
Workshop drill: check units and precision
After estimating, state the unit and choose a sensible number of significant figures based on graph resolution. The graph cannot support more precision than its scale and data justify.
Workshop drill: compare two methods
Estimate from the graph, then use a supplied equation if available. If the answers disagree beyond expected reading uncertainty, inspect the scale, equation or substitution rather than averaging the two blindly.
Workshop drill: qualify a prediction
Rewrite “At x = 20, y will be 80” as a model-aware sentence: “If the observed linear trend continues to x = 20, the model predicts y ≈ 80.” The extra phrase identifies the assumption rather than weakening the Mathematics.
Workshop drill: identify impossible outputs
Extend several models until one produces a physically impossible result. State which constraint fails. This helps learners recognise that a mathematically continued line can outlive its scientific meaning.
Workshop drill: find the nearest evidence
When asked to estimate at a target x, identify the nearest measured points that bracket it. Use local evidence first. This is especially useful when the global graph has several stages.
Workshop drill: explain why an estimate is uncertain
Name the source of uncertainty: coarse graph scale, scatter, sparse data, extrapolation distance, changing mechanism or instrument limit. “The graph is not exact” is too vague.
Interpolation under examination pressure
The fastest safe sequence is axes, range, shape, target, estimate, unit. Extrapolation adds one more step: state or test the assumption that the same relationship continues.
Final science standard
A strong G3 Science answer knows where the evidence ends. It can estimate within a supported range, use a model beyond that range when justified, and say clearly when the graph does not support a precise or unique prediction.
