How to Predict Beyond the Tested Range in PSLE Science Without Pretending the Trend Must Continue

Wait, What? A Graph Does Not Promise to Continue Forever

A graph rises from 20°C to 30°C to 40°C. The measured outcome also rises. What happens at 50°C?

A learner may be tempted to extend the line and say, “It will keep increasing.” Sometimes that is a reasonable prediction. Sometimes it is not.

The tested evidence tells you what happened inside the measured range. A prediction outside that range is an extrapolation: it asks the current pattern and scientific model to travel beyond the evidence already collected.

A trend can support a prediction beyond the tested range, but the farther you travel from the evidence, the more important the scientific mechanism and its limits become.

This guide teaches how to make that move carefully in PSLE Science without turning one short trend into a universal rule.

Quick Answer

When asked to predict beyond measured data:

READ THE TESTED RANGE → IDENTIFY THE PATTERN → CHECK WHETHER THE SCIENTIFIC MECHANISM SUPPORTS THE SAME DIRECTION → LOOK FOR THRESHOLDS, LIMITS OR COMPETING PROCESSES → PREDICT CAUTIOUSLY → STATE THE CONDITION → CHECK HOW FAR THE PREDICTION LIES BEYOND THE EVIDENCE.

Do not assume a line must stay straight, a rate must stay constant, or a process must keep increasing forever.

The Exact PSLE Science Learning Job This Guide Owns

This page owns one learner job: using PSLE Science evidence and mechanism to make a cautious prediction outside the tested range, while knowing where the prediction becomes uncertain.

It does not replace prediction and hypothesis reasoning. It does not replace the guides on thresholds, plateaus, turning points or gaps between measured points. It owns the particular boundary between what was measured and what is being predicted beyond what was measured.

Why This Matters in the 2026 PSLE Science Frame

For examination from 2026, Standard PSLE Science assesses the 2023 Primary Science syllabus. The official assessment objectives include making predictions and formulating hypotheses, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.

A good prediction therefore is not a guess. It is a claim supported by the pattern, the conditions and the relevant scientific concept.

Interpolation and Extrapolation Are Different

Job	Example	Risk
Interpolation	Predicting at 35°C when 30°C and 40°C were measured	You are filling a gap inside the tested range.
Extrapolation	Predicting at 60°C when only 20–40°C were tested	You are extending beyond the evidence.

Both require reasoning. Extrapolation usually deserves more caution because a new limit, threshold or competing effect may appear outside the measured range.

Worked Example 1 — Drying Time Across Temperature

Original practice situation: Equal wet cloths are tested under otherwise comparable conditions.

Temperature	Time to reach the stated dry condition
20°C	80 min
25°C	68 min
30°C	57 min

A learner is asked to predict the drying time at 35°C.

The data show shorter drying times across the tested range as temperature increases. If the other conditions remain comparable and the same mechanism continues to dominate, it is reasonable to predict a shorter drying time at 35°C than at 30°C.

But the data do not justify an exact value unless a relationship is defined strongly enough to support it. Nor do they prove drying time would keep falling at the same rate at every higher temperature.

Worked Example 2 — More Light Does Not Mean Unlimited Increase

A plant-related investigation shows a measured response increasing across three light levels.

A weak prediction says, “At double the highest light level, the response will double.”

That assumes proportionality and ignores possible limits. A stronger prediction says the response may continue increasing initially if light remains a limiting condition, but the data do not establish indefinite increase or doubling because another condition may become limiting.

Worked Example 3 — The Plateau Warning

Suppose measured values are:

Condition	Response
1	4
2	7
3	9
4	9

Predicting 12 at Condition 5 simply because earlier values rose would ignore the plateau already visible in the tested range.

The evidence now suggests that increasing the condition further may produce little or no further change under the current setup, though another test would be needed to confirm what happens beyond Condition 4.

Worked Example 4 — A Turning Point Can Break a Trend

A measured response rises across the first three conditions and then falls at the fourth.

The learner should not extend only the first rising section. The full evidence shows that the relationship changes direction.

Prediction beyond the range should therefore consider the reason for the turning point and whether the same process is likely to continue.

Worked Example 5 — A Threshold Can Appear Outside the First Measurements

Three low test conditions show no visible response. A learner predicts “there will never be a response at any higher value.”

That is too strong. The tested range may simply lie below a threshold. The correct statement is that no visible response was detected within the tested low range.

The Mechanism Test

A trend is more trustworthy for prediction when the learner can explain why the relationship should continue.

Ask:

What scientific process links the changed condition to the measured outcome?
Is that process still expected to operate beyond the measured values?
Could another process become important?
Could a required resource run out?
Could the measuring method stop detecting additional change?
Could the system reach a physical or biological limit?

The Distance-from-Evidence Rule

A prediction just beyond the tested range is generally less adventurous than one very far beyond it.

If measurements stop at 40°C, predicting at 45°C may be more defensible than predicting at 200°C, because the latter may introduce entirely different processes or states.

The farther the prediction travels, the more carefully the model limits must be stated.

Do Not Extend a Straight Line by Habit

A line drawn through several data points is a representation of the measured relationship. It does not promise that nature continues in a straight line outside the graph.

Ask whether the underlying concept supports linearity. In many Primary Science contexts, you only need to predict the direction of change, not invent a precise numerical continuation.

Do Not Assume Equal Input Steps Give Equal Output Steps

If temperature rises by 5°C each time, the outcome does not have to rise by the same amount each time. Equal steps in one variable do not establish proportional response.

Do Not Predict Outside a Model’s Conditions

A prediction depends on conditions. If the original investigation kept surface area, material, light, mass or other factors constant, extending the pattern assumes those conditions still hold.

Change the conditions and you may be using a different model.

The Earliest-Weak-Link Diagnostic

Failure signature	Earliest weak link	Repair
“The graph goes up, so it will always go up.”	Tested trend became a universal law.	State the tested range and check mechanism limits.
“It doubled once, so it will double again.”	Proportionality was invented.	Check multiple intervals and scientific basis.
“No response at low values means no response ever.”	A possible threshold was ignored.	Limit the conclusion to the tested range.
“The next point must lie exactly on the line.”	Representation was treated as certainty.	Predict direction or range only as strongly as evidence allows.
“I can predict 200°C from data between 20°C and 30°C.”	Prediction travelled far beyond the model’s safe conditions.	Check whether state changes or new processes would appear.
“I used the trend but ignored the plateau.”	Only part of the evidence was selected.	Use the whole tested pattern.

Misconception Repair — Prediction Is Not Guessing

A scientific prediction should have a reason. It may come from a measured pattern, a scientific concept, or both.

Misconception Repair — Prediction Is Not Proof

A sensible prediction can still turn out to be wrong. That does not make the reasoning useless. It means the new evidence has tested the model.

Misconception Repair — Interpolation Is Not the Same Risk

Estimating between measured values may still be uncertain, but it stays within the tested range. Extrapolation asks the model to cross an evidence boundary.

Question-Reading Protocol

Mark the highest and lowest tested values.
Locate where the requested prediction sits.
Decide: inside range or beyond range?
Describe the observed pattern without overclaiming.
Select the relevant concept.
Check for limits, thresholds, plateaus or competing effects.
Make the prediction at the correct strength.
State the condition under which it is expected.
Check what new evidence would test the prediction.

How This Appears in MCQ

Identify whether the option extends beyond measured data.
Reject options that claim exact numerical continuation without support.
Reject options that ignore a threshold or plateau already shown.
Check whether the direction agrees with the relevant Science concept.
Prefer the option that stays within the conditions and strength of evidence.

How This Appears in Open-Ended Answers

A useful reasoning shape is:

Within the tested range, as ______ increased, ______. If the same conditions and mechanism continue, at the next higher/lower value ______ is expected to ______ because ______. However, the exact value is not established by the current data.

This is a scaffold, not a required marking phrase.

Practice Sequence

Separate interpolation from extrapolation.
Describe only the measured trend.
Predict just beyond the range.
Predict much farther beyond the range and explain why confidence should fall.
Add a plateau and revise the prediction.
Add a threshold and revise again.
Change the scientific topic while keeping the same reasoning job.
Return after several days to a fresh graph without the checklist.

Unfamiliar Transfer Challenge

An unknown process is measured at Conditions 1, 2, 3 and 4. The response is 5, 9, 12 and 13 units.

What is safer at Condition 5?

The response may continue to increase, but the shrinking increments suggest the relationship may be approaching a limit. Predicting exactly 17 because earlier values increased would ignore the changing pattern.

Delayed Independent Return

Three to five days later, use a new table or graph and answer:

What range was actually tested?
Is the requested value inside or outside it?
What pattern is genuinely supported?
What mechanism could continue the pattern?
What could stop or reverse it?
What is the least overconfident prediction?
What extra measurement would test the prediction?

The Answer-Checking Receipt

Did I identify the tested range?
Did I distinguish interpolation from extrapolation?
Did I use the whole pattern?
Did I avoid assuming proportionality?
Did I check the mechanism?
Did I consider thresholds, limits and competing processes?
Did I keep the original conditions visible?
Did I avoid inventing exact values?
Did I state uncertainty when the evidence becomes weaker?

Evidence and Model Limits

Real scientific prediction can use mathematical models, uncertainty estimates and much larger datasets. Primary Science needs a simpler but important habit: a pattern is evidence within its tested range, and a prediction beyond that range is a model-based extension that should remain open to correction.

Useful Internal Routes

Parent and Tutor Teaching Guide

When a learner extends a line automatically, ask:

“Where does the evidence stop?”

Then ask, “What scientific mechanism makes you think the pattern continues?” and “What could make it stop?”

Use paired questions: one prediction inside the measured range and one just outside it. Later, move the prediction farther away. The learner should become more cautious without becoming paralysed.

A good learner does not refuse to predict. A good learner predicts with a visible reason and a visible boundary.

Authoritative and Research References

SEAB — PSLE Formats Examined in 2026.
SEAB — PSLE Science syllabus, for examination from 2026.
MOE — Science Teaching and Learning Syllabus, Primary, 2023.
Peterman and colleagues, International Journal of Science Education — primary students’ graph interpretation skills.
Boote, Journal of Science Teacher Education — evidence about the demands of interpreting line graphs in science.

The Quiet Ending

Evidence gives a trend a place to stand.

Prediction asks it to take one more step.

Take that step—but remember where the ground ended.

Small Group Tutorials