How to Make a PSLE Science Prediction at the Right Precision Without Guessing an Exact Number

Wait, What? A Prediction Can Be Scientifically Stronger by Saying Less

A learner sees a graph rising and predicts, “The value will be 47.6.” The graph never gave enough information to support 47.6. Another learner writes, “The value will increase.” That second answer may look less impressive, but it can be scientifically stronger because it matches the evidence that is actually available.

Prediction is not a competition to produce the most exact-looking number. It is a reasoning job: use the given pattern, scientific relationship, conditions and evidence to state what is most defensible about a future or unobserved outcome.

THE PRECISION OF A PREDICTION MUST COME FROM THE EVIDENCE, NOT FROM YOUR CONFIDENCE.

Quick Answer

Before making a PSLE Science prediction, decide what level of precision the evidence supports:

READ THE GIVEN INFORMATION → IDENTIFY THE SCIENTIFIC RELATIONSHIP → CHECK THE CONDITIONS → ASK WHAT IS ACTUALLY DETERMINED → CHOOSE DIRECTION, RELATIVE ORDER, BOUNDED RANGE OR EXACT VALUE → EXPLAIN WHY → CHECK THAT YOU DID NOT INVENT EXTRA PRECISION.

Direction: increase, decrease, remain unchanged, become more/less.
Relative comparison: A will be greater than B, Q will take longer than P.
Bounded range: the result should lie between two justified limits.
Exact value: only when the supplied relationship, rule or calculation determines it.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: choosing the correct precision for a PSLE Science prediction so the claim is exactly as specific as the evidence and scientific relationship allow—no weaker than necessary, and no more exact than justified.

It does not replace the general prediction-and-hypothesis guide, the separate guide on evidence showing direction but not magnitude, the guide on extrapolating beyond a tested range, or the guide on approximate and rounded values. Those remain separate owners. This page is about the claim strength of the prediction itself.

Why This Matters in the Current PSLE Science Frame

For examination from 2026, PSLE Science assesses the 2023 Primary Science syllabus. The official assessment objectives include making predictions and formulating hypotheses, interpreting and analysing information, evaluating observations and information, and communicating explanations and reasoning. Those objectives do not require every prediction to be numerical. The learner must use the kind of prediction the evidence can support.

The 2023 syllabus also treats inquiry as connected reasoning rather than isolated memorisation. A good prediction therefore binds together the object, relevant condition, scientific relationship and expected outcome.

Prediction Is Not the Same as Guessing

A guess can be a number chosen without a defensible chain. A scientific prediction should have a route:

GIVEN CONDITION → RELEVANT RELATIONSHIP OR MECHANISM → EXPECTED DIRECTION / ORDER / RANGE / VALUE.

The precision belongs at the final step. You first decide what the evidence constrains. Only then decide how exact the outcome can be.

The Four Levels of Prediction Precision

Prediction level	What the evidence must support	Example form
Directional	Only the direction of change or effect	“The temperature will increase.”
Relative	An ordering or comparison between cases	“Set-Up B will have a greater value than Set-Up A.”
Bounded	A justified lower/upper limit or interval	“The value should be between 20 and 30 units.”
Exact	A rule, relationship or calculation uniquely determines the value	“The value will be 24 units.”

These four levels are a learning scaffold, not an official examination template. A real question may require wording that fits its context. The principle is durable: match precision to evidence.

Worked Example 1 — The Pattern Supports Direction, Not an Exact Number

A table shows that as the distance from a source decreases across the tested values, a measured response becomes larger. A new distance is given that is closer than one tested condition, but no mathematical rule is supplied.

A learner predicts an exact response of 18.7 units. Where did 18.7 come from? If the table does not determine it, that precision is invented. A defensible prediction may be: “The measured response is expected to be greater than at the farther tested distance,” provided the scientific relationship and range justify that direction.

The prediction is useful because it says what the evidence supports, not because it uses many digits.

Worked Example 2 — Relative Prediction Without Magnitude

Two similar systems differ in one relevant condition. Scientific knowledge supports that the condition should make the process occur faster in B than in A. The question does not provide enough information to calculate how many seconds faster.

The correct prediction job is comparative: B will reach the stated outcome sooner than A. Writing “B will be 12 seconds faster” overclaims. Writing only “something changes” underclaims. The relative comparison is the right resolution.

Worked Example 3 — A Bounded Prediction

A results table shows a value of 30 units at one tested condition and 42 units at the next tested condition. An untested condition lies between those two conditions, and the question explicitly states that the relationship changes smoothly between them.

The evidence may support a bounded prediction: the new value is expected to lie between 30 and 42 units. Unless a rule is provided, the exact midpoint is not automatically justified. “Between” is not a weak answer here; it captures the actual information boundary.

Worked Example 4 — When an Exact Number Is Justified

Suppose the question supplies a rule: for every 1-minute interval under the stated conditions, the quantity increases by 3 units. It starts at 10 units and the conditions remain unchanged for four minutes.

Now the rule determines the value: 10 + 4 × 3 = 22 units. The exact numerical prediction comes from the supplied relationship. It is not guessed from a graph shape or familiar topic.

Worked Example 5 — Extrapolation Beyond the Tested Range

A graph rises over the tested range. A new condition lies far beyond every measured point. A learner extends the line and announces an exact future value.

This is where prediction needs a model-limit check. A trend inside the tested range does not guarantee the same numerical relationship far outside it. A cautious directional prediction may still be possible if the scientific mechanism supports it, but the exact extension may be unjustified. Use the separate extrapolation guide when the new condition sits beyond the tested range.

Worked Example 6 — When the Direction Itself Is Not Certain

Sometimes the evidence does not even support a clear increase or decrease. Imagine repeated data that fluctuate with no consistent relationship across the tested conditions. The learner should not force a trend because a prediction question appears.

A scientifically disciplined response recognises the limitation. If the question asks for a prediction from a supplied rule, use the rule. If no relationship is supported, state only what can be justified and avoid manufacturing a pattern.

Prediction Versus Hypothesis

A prediction states an expected outcome under stated conditions. A hypothesis is a testable proposed relationship or explanation. In school practice the words can appear together, but their jobs should not be merged automatically.

For example:

Hypothesis: Increasing the tested condition will increase the measured outcome, under the stated conditions.
Prediction: Therefore, Set-Up B, which has the higher tested condition, should produce a greater measured outcome than Set-Up A.

The precision of the prediction depends on what the hypothesis, data and conditions actually determine.

The Prediction-Precision Reasoning Chain

READ GIVEN INFORMATION → IDENTIFY THE OBJECT / RELATIONSHIP → DISTINGUISH OBSERVATION FROM INFERENCE → SELECT THE SCIENTIFIC CONCEPT → CHECK THE CONDITION → ASK WHAT THE EVIDENCE DETERMINES → CHOOSE DIRECTION / RELATIVE / RANGE / EXACT → STATE THE OUTCOME → CHECK THAT THE PRECISION IS JUSTIFIED.

A Claim-Strength Ladder

If you know…	You may be able to predict…
Only direction of a relationship	Increase/decrease or more/less
Direction plus two comparable cases	Which case is greater/lower/faster/slower
Valid bounds around an unobserved case	A range or interval
A deterministic supplied rule and required inputs	An exact numerical value

The ladder prevents two opposite errors: being too vague when stronger evidence exists, and being too exact when it does not.

Observable Failure Signatures

Failure signature	Likely weak link
An exact number appears with no rule or calculation	Precision invented
The learner predicts “increase” when a clear relative comparison is required	Prediction too vague
The learner uses the midpoint automatically between two data points	Linear relationship assumed without support
The learner extends a trend far beyond the tested range as if certain	Model limit ignored
A prediction is rewritten after seeing the result	Prediction and outcome evidence mixed
A prediction copies a familiar textbook fact but ignores the question condition	Condition tracking failed

Find the Earliest Weak Link

What exactly is being predicted?
Which condition changes?
What scientific relationship connects that condition to the outcome?
Is the evidence directional, comparative, bounded or exact?
Does the question provide a calculation rule?
Is the new condition inside or outside the tested range?
Does the measuring method support the requested precision?
Have I added a number only because numbers look scientific?

Misconception Repair — “A More Exact Number Is a Better Prediction”

No. An exact number is better only when it is justified. Unsupported digits make the prediction weaker because they pretend to know more than the evidence reveals.

Misconception Repair — “Prediction Means Extrapolate the Graph”

Not always. Prediction can come from scientific knowledge, a supplied relationship, a comparison or a pattern. Extending a graph is only one possible route and must respect the tested range and mechanism.

Misconception Repair — “If I Cannot Give a Number, I Am Guessing”

A directional or relative prediction can be rigorous when the evidence supports direction but not magnitude. Scientific precision means matching the claim to the evidence, not forcing arithmetic.

Question-Reading Protocol

Underline the object or measured outcome.
Circle the condition that changes.
Identify the relevant relationship or rule.
Ask: “What is the strongest thing this evidence determines?”
Choose the precision level.
Write the prediction.
Add the scientific reason when the task requires it.
Check that every number in the prediction has a source.

Retrieval and Practice Sequence

Direction practice: predict only increase/decrease from five simple relationships.
Relative practice: compare paired set-ups without adding exact numbers.
Boundary practice: identify when an interval is safer than a point value.
Exact practice: use a supplied rule to compute a unique prediction.
Mixed practice: hide the requested answer form and decide which precision level each case supports.
Transfer: change the Science topic but keep the same prediction job.

Unfamiliar Transfer Challenge

Create four original Science prediction prompts. Make one support only a direction, one support a relative comparison, one support a bounded range and one support an exact value through a supplied rule. Give the prompts to someone else without labelling the precision level. Their job is to decide which type of prediction each prompt supports and explain why.

If two prompts accidentally support the same level, improve the information until the boundaries are clear. Designing the examples makes the distinction deeper than memorising a list.

Delayed Independent Return Test

Three to five days later, take four unfamiliar prediction questions. Before answering, write only one label beside each: DIRECTION / RELATIVE / RANGE / EXACT. Then answer without notes. If you can choose the correct claim strength and explain why, the skill is becoming independent.

Prediction-Precision Receipt

I identified what is being predicted.
I kept the prediction attached to the stated condition.
I used a scientific relationship rather than a familiar-looking word.
I know whether the evidence supports direction, relative order, a range or an exact value.
I did not invent an exact number.
I checked the tested range when using a trend.
I can explain why my prediction is no more precise than the evidence.

Common Traps

adding decimal places to sound scientific;
assuming the midpoint between two measurements must be correct;
treating a graph line as a universal formula;
forgetting that a new condition may be outside the tested range;
using an exact answer from memory instead of the question evidence;
predicting an amount when only direction is known;
being so cautious that you refuse a clear relative comparison the evidence supports.

Parent and Tutor Teaching Guide

When a learner gives an exact prediction, ask one question: “Where did every digit come from?” If the answer is “I estimated it,” ask whether the question supports an estimate, a range or only a direction.

Use contrast pairs. Show two questions with almost the same Science context, but give one a supplied numerical rule and the other only a directional pattern. Ask why one supports an exact value and the other does not. This teaches evidence control rather than topic memorisation.

Also practise the opposite problem: learners who answer everything with “increase” even when the evidence can support a more informative comparison. The aim is not always less precision. It is appropriate precision.

Useful Internal Routes

Authoritative References

Evidence and Boundary Note

The four-level prediction ladder in this guide is an eduKate learning scaffold, not an official SEAB marking formula. Real questions vary. The official frame supports prediction and inquiry; the learner’s job is to match the prediction to the evidence and conditions given in that particular question.

The Quiet Return

A good prediction is not the one that sounds most certain.

It is the one that says exactly what the evidence gives you permission to say—and stops there.

Small Group Tutorials