PSLE Science performance depends on more than remembering correct scientific statements. A learner must also control the scope of a claim. Data from two set-ups may support a comparison without proving a universal rule. A graph may show that two quantities change together without proving that one caused the other. One experiment may support a conclusion under the tested conditions without proving what will happen in every possible situation. This guide teaches the learner to keep the claim inside the evidence.
The skill is simple to state but difficult to execute under pressure: say what the evidence allows, no more and no less. An answer that is too weak may fail to complete the job. An answer that is too strong may add a cause, certainty, general rule or explanation that the question has not established. Good Science performance means matching the strength of the conclusion to the strength of the evidence.
This volume builds on Vol 0004: Science — Evidence Before Explanation and routes deeper foundations to the PSLE Science Learning Guide. Existing specialist guides remain the owners of observation, inference, prediction and conclusion. Vol 0008 focuses on the next performance job: deciding how far a claim may travel.
WHAT WAS TESTED? → WHAT WAS OBSERVED? → WHAT CAN I CLAIM? → WHAT CAN I NOT CLAIM YET? → WHAT MORE EVIDENCE WOULD I NEED?
PSLE Science answers become stronger when the learner knows which parts come from the question and which parts come from scientific knowledge. A table, graph, diagram or description may establish what happened. Scientific knowledge may be needed to explain why. Mixing these jobs produces one of the most common Science failures: a true statement that does not answer the evidence in front of the learner.
This guide develops a foundational rule: evidence before explanation. It links to the PSLE Science Learning Guide, the wider PSLE Learning Guide and the shared launch routine in Vol 0001.
READ THE EVIDENCE → NAME THE SCIENCE JOB → SELECT THE RELEVANT CONCEPT → BUILD THE MECHANISM → RETURN TO THE EVIDENCE.
What PSLE Science performance actually requires
Science performance is not a contest to recall the most keywords. The learner has to use knowledge with understanding and apply scientific reasoning to the situation presented. That means the answer must respect the objects, conditions, observations and relationships in the question.
A memorised sentence can be scientifically correct and still be the wrong answer.
The three layers of a Science response
Layer 1: evidence
What does the question actually show, state or measure? This may be a value, trend, observation, comparison, labelled condition or experimental result.
Layer 2: concept
Which scientific idea is relevant? The best concept is not the chapter name. It is the smallest piece of knowledge that can explain or justify the required result.
Layer 3: mechanism
How does the condition produce the outcome? A mechanism connects the concept to the specific case.
A strong explanation often has the shape: condition → scientific process or relationship → effect → observed outcome.
Observation is not explanation
Suppose two identical containers begin at the same temperature. One is wrapped in insulating material. After the same time, the wrapped container has a higher temperature. The observation is that the wrapped container remains warmer. The explanation must connect the insulation to a reduced rate of thermal-energy transfer to the surroundings, which accounts for the higher final temperature.
Repeating “the wrapped container has a higher temperature” does not explain why. Repeating “insulators keep things warm” without connecting it to the measured case is also incomplete. The answer needs both the correct mechanism and the actual condition.
Relationship is not cause
A graph may show that one variable increases as another changes. That pattern is evidence of a relationship in the data. It does not automatically prove the cause. The learner should not add a causal explanation unless the question and the scientific design justify it.
This distinction becomes increasingly important in unfamiliar investigations.
The E–J–K–B routine
- E — Evidence: What is explicitly given, observed or measured?
- J — Job: Do I need to state, compare, predict, infer, explain, conclude or evaluate?
- K — Knowledge: What scientific concept or mechanism is necessary?
- B — Bind: How do I connect the knowledge back to the specific object, condition and result?
During practice, learners can label these steps. In the examination, the routine should become mental rather than a written template.
Worked example 1: compare before explain
Imagine two plants are placed under different light conditions for the same period and a table records their growth. A compare question asks how the results differ. The answer should compare the measured growth using the same basis. An explain question then asks why. Only at that point should the learner bring in the relevant concept about the role of light in the process being assessed, at the level expected by the curriculum.
The first job is evidence. The second job is mechanism. Do not let one impersonate the other.
Worked example 2: a circuit diagram
Suppose a circuit changes after one component is moved. Before explaining, identify the actual connection shown. Is the path complete? Which components share a branch? What changed and what stayed the same? A memorised statement about “more batteries” or “more bulbs” is not useful unless it matches the arrangement.
The diagram is evidence. The circuit concept interprets the evidence. The explanation must return to the arrangement shown.
Worked example 3: fair-test reasoning
If two set-ups differ in more than one relevant condition, a difference in outcome cannot safely be attributed to only one of them. The learner should first identify what varied, what was controlled and what was measured. Evaluation questions are about the strength of the method, not just the chapter content.
A strong answer makes the consequence visible: because another relevant condition also changed, the comparison does not isolate the effect of the intended variable.
The keyword-dumping trap
Students are often taught important scientific words. The problem begins when the words are treated as marks by themselves. “Heat”, “energy”, “photosynthesis”, “force”, “evaporation” or “oxygen” do not automatically form an explanation.
Use a keyword only when it performs a job in the reasoning chain.
A keyword names an idea. A mechanism connects ideas.
How much detail should an answer contain?
Enough to complete the job, not enough to empty the whole chapter. A useful test is to ask whether every sentence changes the reasoning. If a sentence can be removed without weakening the explanation, it may be unnecessary.
Over-answering creates extra opportunities for contradiction, imprecision and drift away from the question.
The seven Science error families
- Question-reading error: the learner performs the wrong reasoning job.
- Evidence error: the learner ignores, misreads or swaps the data or conditions.
- Concept error: the underlying Science is missing or incorrect.
- Mechanism error: the answer names the concept but does not connect cause and effect.
- Scope error: the claim goes beyond what the evidence supports.
- Communication error: the idea is present but the object, comparison or sequence is unclear.
- Checking error: the answer contradicts the data, diagram or stated condition and the contradiction survives.
Different error families need different repairs. Memorising another model answer will not repair a data-reading error.
A practice method that exposes the source of the answer
- Choose one original or school Science question with a diagram, table, graph or description.
- Underline only the information explicitly given.
- Write the question job in a few words.
- Write the one concept you think is relevant.
- Draft the reasoning chain from condition to mechanism to outcome.
- Check every sentence: evidence, knowledge or bridge?
- Remove knowledge that does not help answer the question.
- Try one changed question using the same concept but a different reasoning job.
This teaches flexibility. The learner stops treating one concept as one fixed model answer.
From basic to advanced Science performance
- Basic: identify what is observed or stated.
- Foundation: distinguish observation, inference, prediction and explanation.
- Core: connect one condition to one mechanism and outcome.
- Transfer: apply the same concept to a changed set-up.
- Advanced: evaluate evidence strength, alternative explanations, method limits and the boundary of a conclusion.
- Exam control: choose the required depth quickly and stop when the job is complete.
When the skill is becoming independent
- The learner can point to the data or diagram feature used in the answer.
- The learner can distinguish “what happened” from “why it happened”.
- The learner does not add a cause merely because two quantities changed together.
- The learner can use the same concept for state, compare, predict, explain and evaluate questions.
- The learner notices when the answer exceeds what the evidence can support.
- The learner checks the final explanation against the actual set-up rather than against a memorised sentence.
How this connects to the wider PSLE series
Return to Vol 0001: Read Before You Solve for the shared launch routine. Use Vol 0002: English — Answer the Actual Task for evidence and meaning in English, and Vol 0003: Mathematics — Represent Before You Calculate for relationship control in Mathematics.
For deeper Science routes, use the PSLE Science Learning Guide, which organises question reading, evidence, investigations, data, measurement, diagrams, reasoning, examination craft and revision.
Official examination reference
For the current assessment objectives and examination format, use the correct examination-year document from the Singapore Examinations and Assessment Board. For 2026, see PSLE Science. The official document and school instructions take priority over generic study advice.
The quick answer: evidence has boundaries
Every piece of evidence comes from particular objects, conditions, measurements and methods. A strong answer respects those boundaries. If two plants were tested under two light conditions for one week, the evidence directly concerns those plants, those conditions and that period. The learner may explain the observed difference using relevant Science when the question requires it, but should not automatically claim that the same result must occur for every plant species, every light intensity and every duration.
This is not about writing timid answers. It is about writing exact answers. Scientific reasoning becomes stronger when the learner can distinguish what is known, what is inferred and what would require additional evidence.
The three-column test: can say, cannot say yet, need to know
When a question feels uncertain, divide the reasoning mentally into three columns.
- Can say: claims directly supported by the observations, measurements or valid comparison.
- Cannot say yet: claims that go beyond the data, such as an untested cause, universal rule or exact future result.
- Need to know: additional variables, controls, repeated trials or observations that would be needed to support the stronger claim.
The three-column test is especially useful for investigations, graphs, comparisons and evaluation questions. It turns vague caution into a concrete reasoning process.
A comparison is not automatically a cause
Suppose set-up A has a higher measured value than set-up B. The learner can state the difference. Whether the learner can claim that one particular factor caused the difference depends on the design. If more than one relevant condition changed, the comparison does not isolate a single cause.
Worked case: two changing conditions
Two containers of water are placed in different locations. One is wider and also receives stronger airflow. More water is lost from that container. The data show a larger loss. They do not isolate whether width, airflow or both explain the difference. A strong evaluation points to the confounding change instead of forcing one cause.
A trend is not an explanation
A graph may show that one measured quantity increases as another increases. That relationship is evidence. The explanation requires scientific knowledge and an appropriate design. The learner should not turn a trend into a mechanism merely because the graph looks smooth.
Worked case: temperature and rate
If a graph shows a rate increasing across several tested temperatures, the safe descriptive claim is that the measured rate was higher at the higher tested temperatures within that range. If the question asks for an explanation and the syllabus concept supports one, the learner may explain using the relevant mechanism. But the graph alone does not prove that the rate will continue increasing forever beyond the tested range.
One result is not the same as a general rule
A single observation can be useful evidence, but generalisation requires care. If one material bends more than another in one test, the learner should not immediately conclude that it is always more flexible in every thickness, shape and condition. The claim should remain tied to the tested set-up unless the scientific concept or additional evidence justifies a broader statement.
Repeated trials strengthen reliability, not magic certainty
Repeating a measurement can reduce the influence of random variation and help the learner judge consistency. But repeated trials do not repair a badly controlled comparison. Ten repeats of two set-ups that differ in several relevant ways still cannot isolate which change caused the result.
This distinction matters in evaluation questions. Reliability and fairness are related to evidence quality, but they solve different problems.
Fair-test reasoning: what exactly was isolated?
In a fair comparison intended to test the effect of one variable, other relevant conditions should be kept controlled as far as the investigation requires. The learner should identify the variable changed deliberately, the outcome measured and the relevant conditions that should remain comparable.
Worked case: fertiliser and plant growth
If one plant receives fertiliser and more water while another receives neither, a difference in growth cannot be attributed confidently to fertiliser alone. The evidence may show that the two plants grew differently. It does not isolate the effect of fertiliser. A better method changes fertiliser while keeping water and other relevant conditions consistent.
Mechanism explains evidence; it does not invent missing evidence
Scientific knowledge is used to explain or predict, but it should not be used to pretend that an unmeasured event was observed. If a question shows that one object cooled more slowly, the learner can use knowledge about thermal-energy transfer to explain the difference when appropriate. But the learner should not add an exact temperature change that was never measured.
Science knowledge can connect the evidence. It cannot manufacture evidence that the question never gave.
Prediction has conditions
A prediction extends a pattern or applies a concept to a stated future condition. A strong prediction names the relevant condition and outcome. It does not treat the future as certain when the evidence is limited.
Worked case: extending a graph
If data show increasing values from 10°C to 30°C, the learner may be asked to predict at 35°C. A reasonable prediction may follow the observed pattern if the question invites extrapolation. But claiming what will happen at 100°C may be unjustified if the relationship could change outside the tested range.
Conclusion scope: match the words to the design
Compare these claims: “Object A had the larger increase in this investigation”; “Object A increases more under these tested conditions”; “Object A always increases more.” Each statement travels farther. The correct level depends on the evidence. Examination control means choosing the strongest claim that is still justified.
The danger words: always, never, proves, definitely, only
Absolute words are not forbidden, but they demand strong evidence. When a learner writes always from one table or proves from one comparison, the claim often outruns the method. Ask whether the question actually established that degree of certainty.
Safer wording is not automatically better either. If the evidence directly establishes a result, writing “maybe” can make the answer unnecessarily weak. The aim is calibrated certainty.
Worked case: circuits
Two circuit diagrams differ in battery arrangement and bulb arrangement, and one bulb is brighter. If more than one relevant feature changed, the learner should not attribute brightness to only one feature. First identify the actual circuit differences. Then decide whether the comparison isolates the intended relationship.
Worked case: heat transfer
Two cups begin at the same temperature. One is wrapped in insulating material and the other is not. After the same time in the same environment, the wrapped cup is warmer. If insulation is the relevant intended difference and other conditions are comparable, the evidence supports a conclusion about reduced thermal-energy transfer under those conditions. It does not justify a claim that the wrapped cup will never cool.
Worked case: forces
A toy car travels farther when released from a higher ramp position. The learner may describe the measured relationship across the tested heights. If asked to explain, relevant energy and motion concepts can be used at the expected curriculum level. The learner should not claim the car will travel farther without limit as height increases; the experiment tested a bounded range.
Worked case: water and evaporation
Two dishes contain the same starting amount of water. One has a larger exposed surface area, and more water is lost over the same time. If other relevant conditions are controlled, the learner can link the larger exposed surface area to a greater rate of evaporation under the tested conditions. The conclusion should still remain tied to the comparison rather than becoming a universal claim about every liquid and environment.
Worked case: plant growth
A plant under one light condition grows differently from another plant under another condition. Before claiming a light effect, check whether the plants, water, soil, duration and other relevant conditions were comparable. Then distinguish the measured growth result from a broader claim about plant health or long-term survival.
Worked case: ecosystems
A field observation records more insects in one microhabitat than another on a particular day. The evidence supports a difference in the observed counts. It may not establish the cause or prove that the same distribution occurs every day or season. More observations, controlled comparisons or environmental measurements may be needed.
Worked case: dissolving
A substance dissolves faster in warmer water in the tested set-ups. The learner should distinguish rate from amount. Faster dissolving does not automatically prove that more substance can dissolve at equilibrium, and the conclusion should not swap one measured idea for another.
Worked case: magnets
A magnet attracts one object but not another. The observation supports a claim about those objects and the conditions tested. The learner should not generalise from appearance alone, such as assuming every shiny metal is magnetic.
Alternative explanations: what else could produce the result?
An advanced learner should sometimes ask whether another factor could explain the observation. This does not mean listing random possibilities. The alternative must be relevant to the design. If two plants differ in growth, unequal water, starting size or light could matter if those conditions were not controlled.
Alternative-explanation thinking helps with evaluation. It shows why a conclusion may be weaker than it first appears.
The scope ladder
- Observed result: what happened in the specific data.
- Comparison claim: how two tested conditions differed.
- Mechanism claim: why the difference occurred, if the design and scientific knowledge support it.
- Generalisation: whether the relationship is expected beyond the exact test.
- Universal claim: a statement meant to hold in all relevant cases.
Each step needs more support. Many PSLE questions operate mainly at the first three levels. The learner should not climb the ladder automatically.
The evidence audit: five questions before you finish
- What exactly was measured, observed or shown?
- Which condition or variable changed?
- Were other relevant conditions controlled well enough for the claim I am making?
- Am I describing a relationship, explaining a mechanism, or claiming a cause?
- Does my final sentence go farther than the evidence?
Scope errors in MCQ
Evidence-scope control also helps in multiple-choice questions. A distractor may contain a scientifically familiar idea but make a claim that is too broad, too certain or about a variable the data did not test. Before selecting an option, ask whether every part of the statement is supported.
One incorrect word can make the whole option unsuitable. Pay special attention to absolutes, causal language and swapped quantities.
Scope errors in structured answers
In structured questions, learners sometimes write extra information to sound scientific. This can introduce an unsupported claim. Build the answer from the required job: evidence, relevant concept, mechanism and bounded conclusion. Stop when the job is complete.
How this differs from observation versus inference
The site already has deeper guides on distinguishing observation, inference, prediction and conclusion. Use those when the learner cannot tell the reasoning categories apart. This volume starts one step later. The learner may already know that a statement is an inference; the question here is whether that inference is stronger than the evidence allows.
For the foundation skill, use How to Tell Observation, Inference, Prediction and Explanation Apart and How to Answer Infer Questions in PSLE Science.
A can-say / cannot-say drill
Give the learner a small table or diagram and ask for three sentences: one statement the evidence clearly supports, one statement the evidence does not yet support, and one piece of additional evidence that would help test the stronger claim. This drill teaches scope directly.
Example
Two identical containers are tested, but one is placed beside a fan. It cools faster. Can say: the container beside the fan cooled faster in this test. Cannot say yet: airflow is the only possible reason if another relevant condition also differed. Need to know: whether the remaining relevant conditions were controlled and whether repeated trials show the same pattern.
A claim-strength rewrite drill
Start with an overstrong sentence and ask the learner to repair it without making it meaningless. Example: “Plants always grow faster in brighter light.” A more defensible version tied to a given investigation might be: “In this investigation, the plant under the higher tested light level showed greater growth over the measured period.”
Then reverse the drill. Give a sentence that is too weak, such as “Maybe set-up A changed more.” If the table clearly shows a larger measured change, strengthen it to a direct comparative statement. Evidence control works in both directions.
A graph-scope drill
Give a graph with data across a limited range. Ask the learner to describe the trend inside the range, predict one nearby value if appropriate, and identify one far-away claim that the graph cannot safely support. This separates interpolation, extrapolation and unsupported generalisation without requiring advanced terminology.
A method-evaluation drill
Show two investigations testing the same idea. One controls the relevant variables; the other changes two conditions at once. Ask which method supports a stronger causal conclusion and why. The learner should connect method quality directly to claim strength.
The claim-boundary transfer drill
Use one investigation in practice and then change the surface context while keeping the reasoning structure. For example, first compare cooling in two containers, then several days later compare plant growth under two treatments. The learner should still ask what changed, what was measured, what was controlled and how far the conclusion may travel. If the learner can apply the same scope discipline without relying on the original topic, the skill is transferring.
Then make the second task deliberately stronger or weaker. In one version, control all relevant conditions except the intended variable. In another, change two relevant conditions. Ask the learner to explain why the same observed difference supports a stronger causal claim in the first design than in the second. This makes the connection between method quality and claim strength explicit.
Finally, give a result with very clear data and ask the learner to avoid becoming too cautious. Evidence control is not about writing “maybe” everywhere. The strongest defensible answer is the goal: direct when the evidence is direct, bounded when the evidence is bounded, and uncertain only when genuine uncertainty remains.
A seven-day evidence-scope cycle
- Day 1: can say / cannot say / need to know.
- Day 2: relationship versus cause.
- Day 3: fair-test and confounding-variable cases.
- Day 4: graph trends and bounded predictions.
- Day 5: repeated trials, reliability and method limits.
- Day 6: mixed MCQ and structured scope errors.
- Day 7: delayed transfer with unfamiliar contexts.
Evidence control under examination time pressure
The learner does not need to write the three columns or scope ladder during the examination. Training should compress into a mental checkpoint: What was tested? What can I say? Am I adding anything? If the answer contains a cause, universal rule or exact detail that was not established, narrow it.
If the learner is stuck, use the time-and-attention routine in Vol 0005. A short evidence audit can rescue the answer; repeated keyword dumping usually cannot.
What parents and tutors should ask
Instead of asking only whether the answer is correct, ask: What did the experiment actually test? Which sentence comes directly from the evidence? Which sentence comes from Science knowledge? What stronger claim would need more evidence? Did more than one relevant condition change? Could another explanation fit?
These questions teach disciplined reasoning without giving the final wording.
Common evidence-scope mistakes
- Universalising one result: turning one investigation into an always/never rule.
- Cause inflation: turning association or comparison into a proven cause.
- Variable blindness: ignoring a second changing condition.
- Quantity swap: claiming about amount when only rate was measured, or vice versa.
- Mechanism substitution: writing a true scientific fact instead of analysing the given data.
- Reliability confusion: assuming repeated trials repair an unfair method.
- Certainty inflation: using words stronger than the evidence justifies.
- Weakening direct evidence: writing “maybe” when the measured comparison is clear.
Frequently asked questions
Should Science answers always say ‘in this experiment’?
No. Do not add phrases mechanically. Use wording that fits the task and evidence. The important habit is keeping the claim inside the tested conditions, whether or not that exact phrase appears.
Can I use scientific knowledge that is not written in the question?
Yes, when the task requires explanation, prediction or application and the knowledge is relevant. But scientific knowledge should explain or connect the evidence, not invent observations or conditions.
Does more data always mean a stronger conclusion?
More relevant, reliable data can strengthen evidence, but quality matters. Repeating an unfair comparison does not isolate the intended variable.
How do I know whether I can say ‘because’?
A causal explanation should be supported by the question design and relevant scientific concept. If several relevant variables changed, be careful about assigning the outcome to one cause.
What if the question asks me to predict?
Use the stated condition, pattern and relevant scientific concept. Keep the prediction within what the evidence and concept justify, and do not invent precision the question does not support.
What if I know the real-world science is more complicated?
Answer at the level and scope of the syllabus and the question. Do not introduce advanced exceptions that distract from the assessed relationship unless they are necessary to avoid an incorrect statement.
When evidence-scope control is becoming independent
- The learner distinguishes measured result from explanation.
- A graph trend is not automatically called a cause.
- Absolute words trigger a quick evidence check.
- Uncontrolled comparisons are identified without prompting.
- Conclusions stay tied to tested conditions.
- Additional evidence can be proposed for stronger claims.
- The learner can strengthen answers that are too vague as well as narrow answers that are too broad.
Next route
Return to Vol 0004: Science — Evidence Before Explanation for the evidence–concept–mechanism foundation and Vol 0005 for examination control.
For deeper Science foundations, use the Primary 6 Observation, Inference, Prediction and Conclusion guide. For final-answer review, use Science Checking by Question Type.
Official PSLE reference
Use the current official SEAB PSLE information and PSLE Formats Examined in 2026 for examination-year requirements. Official documents and school instructions take priority over generic study advice.
Series: How to Perform in PSLE | Learner’s Guide · Vol 0008 · Foundation Science evidence control
