In PSLE Science, an observation can be true without strongly supporting the explanation a learner wants to give. The evidence may simply be compatible with the explanation. Stronger evidence does more: it helps distinguish the proposed explanation from reasonable alternatives. This difference is subtle, but it is one of the clearest signs of mature scientific reasoning.
This volume teaches one advanced performance skill: separate evidence that supports from evidence that merely fits. “Fits” means the evidence does not contradict the idea. “Supports” means the evidence is what we would expect if the explanation were right and is less easily explained by a competing idea.
The skill extends Vol 0004 on evidence before explanation, Vol 0008 on claim boundaries, Vol 0012 on alternative explanations, Vol 0017 on measurement identity, Vol 0021 on unchanged conditions and Vol 0025 on comparing change rather than endpoints.
CLAIM → WHAT WOULD WE EXPECT IF IT WERE TRUE? → WHAT ELSE COULD PRODUCE THE SAME RESULT? → DOES THE EVIDENCE DISTINGUISH? → STATE THE STRONGEST DEFENSIBLE CONCLUSION.
Consistent evidence and supporting evidence are different
Evidence is consistent with an explanation when both can be true at the same time. But many different explanations can fit the same observation. Supporting evidence is more useful because it helps favour one explanation over alternatives.
For example, a plant being taller under one condition may fit several explanations if water, starting size and light also differed. The result alone does not isolate the cause.
Ask what the explanation predicts
Before looking at the result, state what pattern the proposed explanation would predict. If insulation reduces energy transfer, what temperature pattern should appear? If greater exposed surface area affects evaporation, what change should be expected over the same time?
A prediction makes the explanation testable against the data instead of allowing the learner to invent a reason after seeing the result.
Then ask what a rival explanation predicts
A strong evaluation compares at least one alternative when the design allows it. If both explanations predict the same result under the available data, that result cannot strongly distinguish between them.
The learner does not need to invent endless possibilities. One relevant alternative is often enough to reveal whether the evidence is discriminating or merely compatible.
Changed variables matter because they discriminate causes
When one relevant condition changes and other important conditions are held constant, a difference in the measured result can provide stronger support for a causal explanation.
When several relevant conditions change together, the same result may fit multiple causes. The observation can still be accurate while the causal conclusion remains weak.
Unchanged conditions are evidence too
Controls and constant conditions are not background decoration. They help eliminate alternatives. If both set-ups use the same amount, time, starting temperature and container type, those similarities strengthen the interpretation of the deliberately changed factor.
This is why Vol 0021 treats what stayed the same as part of the evidence structure.
Measurement choice affects support
If the explanation is about rate, measuring only a final amount may not always distinguish rate from starting differences. If the explanation is about temperature change, final temperature alone can mislead when starting temperatures differ.
A measurement supports an explanation better when it corresponds closely to the quantity the mechanism predicts.
Starting points can make endpoints misleading
Two set-ups can end at different values simply because they began at different values. Comparing the change from a common or properly accounted starting point is often more informative.
This is the main performance job in Vol 0025: compare like with like before explaining.
Repeated trials improve reliability, not every kind of validity
Repeating a fair comparison can strengthen confidence that the observed pattern is not a one-off fluctuation. But repeating an unfair comparison does not magically isolate the intended variable.
Reliability and causal support are different questions. The learner should know which one the evidence addresses.
A true fact can be irrelevant evidence
A learner may recall a scientifically correct statement that does not distinguish the explanations in this question. True information is not automatically useful evidence.
The test is: does this fact connect the changed condition to the measured result, or help rule out an alternative?
The absence of an expected pattern can weaken an explanation
If an explanation predicts that increasing a condition should increase a response, but the measured response repeatedly stays unchanged, the evidence may weaken that explanation under the tested conditions.
Do not force a favourite mechanism onto data that fail to show its predicted direction.
One observation rarely proves a universal rule
Even strong support in one fair investigation is normally tied to the tested conditions and syllabus-level relationship. The learner should avoid turning a bounded result into an always-or-never claim.
Use Vol 0008 to keep the conclusion inside the evidence.
Good evidence often changes what alternatives remain
Think of evidence as narrowing a set of possible explanations. Weak evidence leaves many explanations alive. Stronger evidence removes some of them because their predictions do not match the result.
This way of thinking is useful in MCQ as well as open-ended questions: which option survives all the evidence with the fewest unsupported assumptions?
Evidence can support without proving
Scientific reasoning is not all-or-nothing. An observation can increase support for an explanation without making it certain. At primary level, the learner does not need advanced probability language; simple phrases such as “supports the conclusion under these conditions” are enough.
The goal is calibrated reasoning: neither overclaim nor become so cautious that clear evidence is weakened unnecessarily.
Use comparison design as part of the answer
If a question asks why a set-up provides stronger evidence, refer to what differs and what remains the same. The design itself explains why one comparison can isolate the intended relationship better than another.
Do not answer only with “it is a fair test” if the question expects the specific conditions that make it fair.
A six-step support-versus-fit routine
- State the claim or explanation.
- State what result that explanation predicts.
- Name one relevant alternative explanation if one exists.
- Ask whether the alternative predicts the same result.
- Identify which controlled condition, measurement or comparison helps distinguish the explanations.
- Write the strongest conclusion the evidence can support without going beyond it.
Twenty-four worked evidence-discrimination cases
Insulated cup
One cup is wrapped with insulating material and another identical cup is not; both start at the same temperature and are left for the same time.
The likely reasoning failure is true fact without discrimination. Saying “insulation is a material” is true but does not support why one cup changes temperature less. The useful evidence is the controlled comparison and measured temperature change.
The design supports a mechanism involving reduced energy transfer because the wrapping is the relevant changed condition while starting temperature, time and cup type are kept comparable. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Different starting temperatures
Two cups end at different temperatures, but one began much hotter.
The likely reasoning failure is endpoint treated as causal evidence. The final difference merely fits many explanations because the starting states differ.
Compare temperature change or use matched starting temperatures before attributing the result to the tested material. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Evaporation surface area
Equal volumes of water are placed in dishes with different exposed surface areas for the same duration.
The likely reasoning failure is relevant prediction. A larger exposed area predicts greater loss over the same time if other relevant conditions are controlled.
Measuring remaining volume or mass after the same time can support the relationship better than simply observing that both dishes contain less water. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Evaporation with fan difference
One dish also has moving air while the other does not.
The likely reasoning failure is confounded evidence. Greater water loss fits both surface-area and airflow explanations.
The result alone cannot strongly support the surface-area explanation because a rival changed condition predicts the same direction. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Plant growth
Plants under different light conditions differ in height after a week.
The likely reasoning failure is multiple possible causes. The height difference fits a light explanation, but it also fits unequal water, starting height or plant condition.
Support strengthens when those factors are controlled and growth is compared from appropriate starting points. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Seed germination
More seeds germinate in one set-up by the observation day.
The likely reasoning failure is count treated as proof of cause. The count supports a condition effect only if seed type, number, water and other relevant conditions are comparable.
If several conditions differ, the observed count is real but not discriminating. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Toy car ramp
A car released from a higher position travels farther.
The likely reasoning failure is mechanism disconnected from measurement. A correct statement about energy supports the explanation only when it predicts the observed direction of distance change in the given set-up.
The measured distance and controlled car/surface conditions provide the evidence link. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Different toy cars
Higher ramp uses a lighter toy car while lower ramp uses a heavier one.
The likely reasoning failure is alternative explanation survives. The farther travel fits both release-height and car-difference explanations.
Using the same car would eliminate the car identity as an alternative. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Dissolving temperature
Equal samples are placed in different water temperatures and time to dissolve is recorded.
The likely reasoning failure is rate versus amount confusion. A shorter dissolving time supports a difference in dissolving speed under the tested conditions.
It does not by itself prove a greater maximum amount can dissolve. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Stirring and temperature both change
Hot water is stirred while cool water is not.
The likely reasoning failure is two causes predict same direction. Faster dissolving fits both higher temperature and stirring.
The observation is compatible with both and therefore weak at distinguishing the intended cause. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Shadow length
A light source moves while object and screen remain fixed; shadow length changes.
The likely reasoning failure is measured relationship. The geometry change supports an explanation about source position and light paths.
A statement about brightness would merely introduce a related but unmeasured idea. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Circuit brightness
One circuit arrangement produces a brighter bulb.
The likely reasoning failure is observation versus unmeasured exact current. Brightness supports a qualitative circuit explanation at the syllabus level, but an exact current value should not be invented without measurement.
Evidence must match the strength and type of the claim. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Magnet test
A magnet attracts one object but not another.
The likely reasoning failure is appearance-based alternative. The observation supports a claim about the tested objects, not a universal rule about every shiny metal.
Testing additional materials can challenge an overbroad appearance-based explanation. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Plant in darkness
A plant condition changes and growth differs.
The likely reasoning failure is chapter keyword dumping. A true fact about photosynthesis supports the answer only if it connects the changed condition to the measured plant response.
The explanation needs a mechanism chain rather than a list of related facts. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Repeated cooling trials
The same insulation comparison is repeated three times with similar results.
The likely reasoning failure is reliability improvement. Repeated trials support reliability of the observed pattern.
They do not repair a design in which another relevant condition differs every time. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
One anomalous trial
Two trials show one pattern and one differs sharply.
The likely reasoning failure is ignoring conflicting evidence. The anomaly does not automatically disprove the explanation, but it weakens confidence and invites checking of method or measurement.
A strong answer acknowledges the full evidence rather than selecting only the convenient trial. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Food-web population count
One organism count falls after another population changes.
The likely reasoning failure is correlation treated as complete cause. The pattern may fit a food relationship, but other environmental changes could also influence counts unless the question structure rules them out.
Use the food-web relationship plus the specific observed change, while keeping the conclusion bounded. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Water-level experiment
Water level rises when an object is placed in a container.
The likely reasoning failure is measurement renamed. The observed rise supports displacement-related reasoning, but the height change is not automatically a directly measured displaced volume.
A calibration or geometry bridge is needed for a quantitative volume claim. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Graph trend
A graph shows response increasing as condition increases across several tested points.
The likely reasoning failure is trend versus universal rule. The trend supports a relationship across the measured range.
It does not prove the same pattern continues indefinitely beyond the tested range. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
No-change result
A condition changes but the measured response remains similar across repeated fair trials.
The likely reasoning failure is forcing the expected mechanism. The unchanged response weakens an explanation that predicted a clear directional change under those conditions.
The learner should not invent hidden effects just to preserve the preferred answer. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Alternative set-up
Two proposed investigations test the same claim; one changes one relevant factor, the other changes two.
The likely reasoning failure is design-strength comparison. The one-factor comparison provides more discriminating evidence because fewer rival causes remain.
The learner can explain why evidence quality differs even before seeing results. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
MCQ distractor
Two answer options are scientifically possible, but only one directly explains the given result.
The likely reasoning failure is possibility mistaken for support. A possible statement may merely fit the topic. The better option uses the evidence and mechanism that discriminate the case.
Choose the option that accounts for the specific conditions, not the one that is merely true in general. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Prediction test
A learner proposes a mechanism before a new trial.
The likely reasoning failure is post-hoc explanation. Write the predicted direction before seeing the new result.
A result that matches the prediction supports the mechanism more strongly than a reason invented only after the result is known. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
Controlled condition
Two set-ups use equal time, amount and apparatus but differ in one target condition.
The likely reasoning failure is controls ignored. The unchanged factors are part of the evidence because they reduce alternative explanations.
A good evaluation names the specific controlled conditions rather than using “fair test” as an unexplained label. The question is not merely whether the observation can coexist with the explanation. The question is whether the observation and design make that explanation more defensible than relevant alternatives.
Now ask what extra evidence would discriminate better. A stronger comparison may control another variable, match starting conditions, measure a closer quantity, repeat the trial, or test a prediction that competing explanations treat differently.
For delayed transfer, change the Science topic but preserve the evidence structure. The learner should still distinguish “fits” from “supports” without relying on memorised chapter wording.
The prediction-first drill
During practice, hide the results and show only the question, set-up and proposed explanation. Ask the learner to predict what should happen if the explanation is correct. Then reveal the data. This prevents the learner from shaping every explanation after the fact.
Next ask what result would weaken the explanation. A learner who can state both supporting and weakening outcomes is reasoning from a testable model rather than from keyword association.
A seven-day evidence-discrimination cycle
- Day 1: claim versus observation.
- Day 2: prediction before result.
- Day 3: one relevant alternative explanation.
- Day 4: controlled conditions and fair comparisons.
- Day 5: measurement choice and starting-point control.
- Day 6: reliability, repeated trials and anomalous results.
- Day 7: delayed mixed transfer using MCQ, graphs and open-ended explanations.
Parents and tutors: ask what else could fit
When a learner gives a correct-looking explanation, ask: “Could another cause produce the same observation?” This question should not become endless scepticism. One relevant alternative is enough to test whether the evidence actually discriminates.
Then ask which feature of the design or data makes the preferred explanation stronger. This turns a model answer into a reasoning habit.
Frequently asked questions
Does supporting evidence prove the explanation?
Not always. It can make the explanation more defensible under the tested conditions without proving a universal rule.
What does ‘merely fits’ mean?
It means the evidence does not contradict the explanation, but the same evidence could also be explained by other relevant possibilities.
Do I always need an alternative explanation in the exam?
No. Use alternatives when the question asks for evaluation, when more than one relevant factor changed, or when distinguishing explanations is central to the reasoning.
Why do controlled variables matter?
They reduce rival explanations. If important conditions are held constant, differences in the result can be linked more strongly to the intended changed condition.
Do repeated trials make evidence stronger?
They can improve reliability. They do not automatically make a confounded comparison fair.
Can a true scientific fact be weak evidence?
Yes. A fact can be correct but irrelevant to the specific measured result or unable to distinguish competing explanations.
Official 2026 PSLE Science frame
The 2026 PSLE Science syllabus assesses knowledge with understanding and application of knowledge and scientific inquiry, including making predictions, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. Distinguishing supporting evidence from merely compatible evidence strengthens those inquiry skills. See the 2026 PSLE Science syllabus and the 2026 PSLE formats page.
Next route
Return to Vol 0004 for evidence-before-explanation foundations, Vol 0008 for claim boundaries, Vol 0012 for alternative explanations, Vol 0017 for measurement identity, Vol 0021 for unchanged conditions and Vol 0025 for change comparison. Continue to Vol 0030: Turn Every Correction Into a Future Trigger. The wider Science route is the PSLE Science Learning Guide and PSLE Learning Guide.
The performance rule
Do not ask only whether the evidence fits your explanation. Ask whether it helps distinguish your explanation from other relevant possibilities.
Series: How to Perform in PSLE | Learner’s Guide · Vol 0029 · Science evidence discrimination
