Wait, What? A Scientific Prediction Is Not “What I Think Will Happen”

In Science, a prediction is not a guess with a confident tone. It is an expected outcome tied to evidence, a known relationship or a scientific model.

A hypothesis is not just a longer prediction either. It proposes a testable relationship between factors and gives an investigation something specific to examine.

The real PSLE Science job is not to sound certain. It is to make an expectation that is traceable to the question conditions and scientifically defensible.

Quick Answer

For a prediction question, use this reasoning chain:

READ THE CONDITION → IDENTIFY WHAT CHANGES → IDENTIFY THE OUTCOME → FIND THE PATTERN OR SCIENTIFIC RELATIONSHIP → STATE THE EXPECTED OUTCOME → EXPLAIN WHY → CHECK THE BOUNDARY.

For a hypothesis question, identify the factor that may affect another factor, state the expected relationship in a testable way, and make sure the proposed relationship can actually be investigated using observable or measurable evidence.

The Exact PSLE Science Learning Job This Guide Owns

This guide teaches a Primary 5 or Primary 6 learner how to answer PSLE Science prediction and hypothesis questions.

It does not replace the existing Primary Science concept guide on making predictions with reasons. That broader scientific-inquiry owner remains canonical. This page owns the PSLE learner job: reading the specific question conditions, deciding whether the task is prediction or hypothesis, using evidence without overclaiming, and communicating an answer that survives a changed surface context.

The Official Inquiry Frame

The revised 2026 PSLE Standard Science paper assesses the 2023 Primary Science syllabus. SEAB identifies scientific inquiry as part of the application assessment objective, including making predictions and formulating hypotheses, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.

That wording matters. Prediction is not separated from evidence and reasoning. The learner has to connect what is expected to the scientific relationship represented by the question.

First Separate Four Different Jobs

Observation

An observation reports what was directly seen, measured or recorded.

Example: The water level decreased from 50 mL to 42 mL after two days.

Prediction

A prediction states an expected future or unmeasured outcome based on evidence, a pattern or a scientific relationship.

Example: If the same trend continues under similar conditions, the amount of water remaining after another interval is expected to decrease further.

Hypothesis

A hypothesis proposes a testable relationship between variables or conditions.

Example: Increasing the exposed surface area of water increases the amount of water that evaporates over a fixed time under the same surrounding conditions.

Conclusion

A conclusion states what the collected evidence supports after the investigation.

A prediction comes before the result. A conclusion comes after the result. Confusing them weakens the logic of the whole investigation.

Where Can a Prediction Come From?

A strong prediction is usually supported by one or more of four sources.

1. A Pattern in Given Data

If the question gives a sequence or trend, the learner may extend it cautiously.

Example data:

20 cm² exposed surface area → 92 g water remaining
40 cm² → 85 g remaining
60 cm² → 78 g remaining

Across those tested values, larger exposed surface area is associated with less water remaining after the same time. If asked to predict for 80 cm² under the same conditions, the learner may expect still less water to remain.

But do not pretend the exact numerical relationship is known unless the pattern justifies it. Predicting “71 g” assumes the observed decrease continues in the same way. That is a model, not a guaranteed fact.

2. A Known Scientific Concept

If no data trend is provided, the learner may use an established concept.

Example: if an identical hot drink is placed in two containers and one has better thermal insulation, a learner can predict that the better-insulated container will show a smaller temperature decrease over the same period, because heat is transferred to the surroundings more slowly.

3. A Mechanism

A mechanism tells how one condition can produce an outcome.

For example, if a circuit is opened, the conducting path is incomplete, so current does not flow through the bulb. The learner can predict that the bulb will not light.

4. A Stated Relationship in the Question

Sometimes the question itself gives a relationship discovered earlier and asks what should happen under a new condition. The learner should use that relationship rather than importing an unrelated memorised fact.

The Prediction Ladder

Build a prediction in layers.

Outcome: What do I expect to happen?
Condition: Under exactly which changed condition?
Comparison: More, less, faster, slower, higher, lower, same or different compared with what?
Reason: Which pattern, concept or mechanism supports the prediction?
Boundary: Is the prediction only justified within similar conditions or the tested range?

Not every answer needs five sentences. The ladder is a reasoning check. A concise answer can still contain all the essential logic.

Worked Example 1: Prediction From a Pattern

An original practice table shows that as the number of hours of light received by identical seedlings increases from 2 to 4 to 6 hours per day, their mean mass gain over a fixed period increases.

The question asks what you predict for seedlings receiving 5 hours of light.

A careful learner first identifies that 5 hours lies within the tested range. That is interpolation.

A suitable prediction is that the mean mass gain is likely to fall between the values observed at 4 and 6 hours, assuming the relationship between those values continues under similar conditions.

The answer does not need to invent an exact value if the graph or table does not justify one.

Interpolation Is Usually Safer Than Extrapolation

Interpolation predicts within a tested range. Extrapolation predicts beyond it.

If a plant grows more from 2 to 6 hours of light, it does not follow that 24 hours of light must produce four times the growth. Other limits can appear. Biology, materials and physical systems often stop following a simple trend outside the observed range.

A graph shows what happened in the measured range. It does not automatically grant permission to extend the same rule forever.

Worked Example 2: Prediction From a Mechanism

Two identical cups contain equal amounts of hot water at the same starting temperature. Cup A is wrapped in a thicker layer of the same insulating material than Cup B.

A question asks which cup is predicted to have a higher water temperature after ten minutes.

A strong reasoning chain is:

Thicker insulation reduces the rate of heat transfer from the hot water to the surroundings. Therefore the water in Cup A is predicted to lose less thermal energy and remain at a higher temperature after the same time.

The prediction is not “Cup A because thicker is better”. It names the changed condition, mechanism and measured outcome.

Worked Example 3: The Dangerous Slogan

A learner sees two circuit diagrams and remembers “more cells make bulbs brighter”.

That slogan cannot be used before checking the circuit.

Are the cells connected in the same direction?
Is the conducting path complete?
Are the bulbs arranged in the same way?
Are the cells identical?
What exactly is the question asking to compare?

A prediction must fit the actual system shown. Familiar chapter phrases are not substitutes for reading the condition.

What Makes a Hypothesis Testable?

A useful hypothesis identifies a relationship that can be examined with evidence.

Weak: Plants like sunlight.

Stronger: Increasing the daily duration of light received by similar seedlings increases their mean mass gain over a fixed growth period, when other relevant growing conditions are kept comparable.

The stronger version identifies:

a changed factor;
a measurable response;
a direction of expected relationship;
a comparison that could be investigated.

A Hypothesis Is Not a Guarantee

A hypothesis can be reasonable and still not be supported by the results.

That is not failure in Science. It is the point of testing.

After results are collected, the learner should compare evidence with the hypothesis rather than rewriting the original expectation to match what happened.

Worked Example 4: A Prediction Meets Unexpected Data

Suppose three groups receive low, medium and high amounts of a condition. Their measured growth is:

low → 2.0 cm
medium → 4.5 cm
high → 4.4 cm

A learner expected growth to keep increasing.

The correct response is not to pretend 4.4 is larger than 4.5. Nor should the learner immediately declare the whole concept false.

The evidence shows an increase from low to medium, but little or no further increase from medium to high in this dataset. Possible explanations might include a plateau, natural variation, measurement limits or another condition becoming limiting. More evidence may be needed.

This is scientific maturity: the data can change what you conclude without changing what was honestly predicted beforehand.

The PSLE Prediction Protocol

Find the target: What outcome are you asked to predict?
Find the condition: What is changing or being extended?
Find the basis: Pattern, concept, mechanism or stated relationship?
State direction: More/less, increase/decrease, faster/slower, on/off, same/different.
Explain why: Connect the scientific mechanism or observed pattern.
Check scope: Are you predicting only within what the evidence can reasonably support?

The PSLE Hypothesis Protocol

State the relationship you want to test.
Name the factor that changes.
Name the observable or measurable outcome.
State the expected direction if justified.
Make sure the variables can be investigated safely and meaningfully.
Do not include a result that has already been observed and call it a hypothesis.

Prediction and Fair Tests Are Connected

A prediction can be scientifically reasonable, yet the investigation may be unable to test it fairly.

Example: you predict that rougher surfaces increase friction and reduce the distance a toy car travels. If the cars are released from different ramp heights, the design introduces another cause of distance difference. The prediction may be sensible; the experiment is weak.

This is why prediction, variable control and evidence evaluation should be learned as one inquiry system rather than isolated vocabulary boxes.

Prediction and Graphs Are Connected

When a prediction is based on a graph, inspect:

axes and units;
the tested range;
whether the trend is steady or changing;
any unusual points;
whether the question asks for a qualitative direction or a numerical estimate;
whether extrapolation goes far beyond the evidence.

Prediction and Explanations Are Connected

A high-quality prediction often becomes stronger when the learner can explain the mechanism.

Weak: The water will be hotter.

Stronger: The water is predicted to remain hotter because the insulating material reduces the rate at which heat is transferred from the water to the surroundings.

The second answer shows why the outcome follows from the condition.

Scientific Vocabulary Without Keyword Dumping

Words such as “evaporation”, “conduction”, “friction”, “photosynthesis”, “current” and “insulation” are useful only when they carry the correct relationship.

Do not write five Science words around a guess. Use one or two precise terms to make the causal route clearer.

Observable Failure Signatures

Failure 1: The Learner Predicts From Personal Experience

“I think this happens because I have seen it before.”

Earliest weak link: evidence source is uncontrolled. Repair: ask, “Which part of the question or which scientific concept supports that expectation?”

Failure 2: The Prediction Is Not Bound to the Condition

Earliest weak link: learner states an outcome but does not identify what changed. Repair: require the sentence “When ______ changes, I predict ______ because ______.”

Failure 3: The Learner Treats a Trend as a Law

Earliest weak link: model limits are invisible. Repair: distinguish “within the observed range” from “always”.

Failure 4: The Hypothesis Cannot Be Measured

Earliest weak link: vague words such as “better”, “likes” or “healthier” are not linked to observable outcomes. Repair: ask, “What result would we actually record?”

Failure 5: The Learner Changes the Prediction After Seeing the Result

Earliest weak link: prediction and conclusion are confused. Repair: preserve the original prediction and then evaluate whether the evidence supports it.

Common Traps

Writing “I think” without evidence.
Copying the last data point instead of predicting a relation.
Predicting an exact number when only direction is justified.
Extrapolating far outside the graph.
Ignoring a plateau or unusual result.
Using a chapter slogan before reading the setup.
Calling a conclusion a hypothesis.
Writing a hypothesis with no measurable outcome.
Changing two variables in the proposed investigation.
Using a correct concept but attaching it to the wrong condition.

A Six-Stage Practice Sequence

Direction only: practise predicting increase/decrease/same from clear relationships.
Add condition: force every prediction to name the changed condition.
Add reason: connect one concept or pattern.
Add boundaries: separate interpolation from extrapolation and tested from untested ranges.
Add hypotheses: turn relationships into testable changed-factor → measured-outcome statements.
Delayed transfer: several days later, solve mixed prediction questions from unfamiliar topics without a sentence frame.

Transfer Check

Situation A: A table shows that a shadow becomes shorter as a light source moves higher above an object across three tested positions. If asked about an intermediate position, use the observed relation cautiously. Do not claim the relationship continues indefinitely beyond the tested setup.

Situation B: An investigation shows that a toy car travels less distance over rougher surfaces when released from the same height. A reasonable hypothesis for a follow-up test is that increasing surface roughness decreases travel distance under comparable release conditions.

Situation C: A plant-growth graph rises and then levels off. Predicting a dramatic further rise just because “more input means more growth” ignores the actual evidence pattern.

Situation D: A circuit prediction changes when the switch state changes. Before predicting the bulb, inspect whether the conducting path becomes complete.

The Prediction Checking Receipt

Target: Did I predict the correct measured outcome?
Condition: Did I use the exact changed condition?
Basis: Is the prediction supported by data, concept or mechanism?
Direction: Is more/less/higher/lower/same clear?
Reason: Does the explanation show why?
Scope: Am I staying inside the evidence range?
Language: Am I distinguishing prediction from observation and conclusion?

Parent and Tutor Teaching Guide

When a child predicts incorrectly, do not immediately supply the “right” prediction. Ask three questions first:

“What condition changed?”
“What result are you predicting?”
“What evidence or Science idea makes you expect that?”

If the learner cannot answer question 1, the problem is question reading. If question 2 is vague, the measured outcome is unclear. If question 3 is missing, the learner is guessing or recalling a slogan.

For hypotheses, ask: “What will you change?” and “What will you measure?” If either cannot be named precisely, the hypothesis is not yet operational.

Fade the prompts over time. The goal is not for the child to depend on a sentence template, but to internalise condition → relationship → expected outcome → reason.

How We Know This Skill Can Be Taught

Research on scientific inquiry and control-of-variables reasoning shows that these skills improve with explicit instruction and practice; they should not be treated as automatic by-products of content study. Broader research on scientific explanation likewise shows that students benefit when claims, evidence and reasoning are made explicit rather than left implicit.

For PSLE preparation, that supports a practical principle: do not practise prediction only as a one-line answer. Practise the reasoning route that makes the one line scientifically justified.

Canonical Boundary

For the general Primary Science concept of prediction, use the existing guide: Making a Prediction with a Reason. This article specifically owns how a PSLE Science learner interprets and answers prediction/hypothesis questions.

Useful Internal Routes

Authoritative References and Further Learning

SEAB: 2026 PSLE Science syllabus and examination format
MOE: 2023 Primary Science Teaching and Learning Syllabus
Science-education research on scientific inquiry, explanation and control-of-variables reasoning.

The Quiet Ending

The beginner asks: “What do I think will happen?”

The developing learner asks: “What pattern do I see?”

The stronger learner asks: “Which condition changes, and which mechanism links it to the outcome?”

And the independent PSLE Science learner asks: “What prediction is justified by this evidence or model, under these conditions, and how far can I extend it before the evidence runs out?”

Small Group Tutorials

How to Answer PSLE Science Prediction and Hypothesis Questions