How to Keep Your Prediction From Changing What You Record in PSLE Science

Wait, what? A prediction can be wrong and the science can still be excellent.

A prediction can also be right while the science is poor.

The difference is what happens when evidence arrives. If you expected Set-up A to give the larger reading, but Set-up B actually gives the larger reading, the scientific job is not to rescue your prediction. It is to protect the observation. Record what the method genuinely produced, then investigate why.

That sounds obvious until a real question, experiment or practice task produces an inconvenient result. Then the mind starts negotiating: perhaps I read the scale wrongly; perhaps this number should be rounded the other way; perhaps that trial does not count; perhaps the picture “must” mean what I expected; perhaps I should write the result that fits the chapter.

Scientific reasoning becomes stronger when expectation and evidence are allowed to disagree.

Quick Answer

In PSLE Science learning, keep a prediction and an observation in different jobs. A prediction is what you expect before the result is known. An observation or measurement is the evidence actually obtained from the situation, diagram, table, graph or investigation. Record the evidence according to the stated method, scale, units and observation rule even when it surprises you. Only after the observation has been preserved should you check the method, examine repeats, consider limitations, revise an explanation or decide whether a result may be anomalous.

A useful student rule is simple: prediction may guide what you look for; it must not decide what you record.

The Exact PSLE Science Learning Job Owned by This Guide

This guide owns one narrow but important learner job: how a Primary 5 or Primary 6 student protects the integrity and objectivity of scientific evidence at the moment it is observed and recorded, especially when the evidence does not match an expectation.

It does not replace the scientific concept pages for heat, light, forces, electricity, plants, materials, cycles or any other Primary Science topic. It also does not replace separate guides on comparing a prediction with the final result, handling an anomalous result, deciding whether repeats should be averaged, or evaluating a flawed method. Those are later reasoning jobs. Here, the first duty is earlier: do not let the expected answer quietly become the recorded answer.

Why This Matters in the Current Singapore Primary Science Frame

The current 2026 PSLE Science examination assesses attainment in the 2023 Primary Science syllabus. SEAB states that the assessment includes knowledge with understanding and the application of knowledge and scientific inquiry, including making predictions and formulating hypotheses, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.

The MOE Primary Science syllabus also places scientific attitudes alongside scientific knowledge and inquiry. Among these are integrity in handling and communicating data and information, and objectivity in seeking data and information to validate observations and explanations without bias. Those words are not an instruction to memorise an examination phrase. They describe a way of doing science: evidence must be allowed to correct us.

This guide therefore does not teach a secret marking template. It trains a scientific habit that supports inquiry, data interpretation, explanation and self-correction.

The Invisible Fork: What You Expected vs What Happened

Imagine that Alicia is investigating which of two coverings slows the cooling of warm water more effectively. Before taking the final readings, she predicts that Cup A will remain warmer because she thinks its covering reduces heat transfer more effectively.

After the same stated time, the thermometers read:

Set-up	Alicia predicted	Observed temperature
Cup A	Higher final temperature	39 °C
Cup B	Lower final temperature	42 °C

At this moment there are two records in Alicia’s head:

Expectation: A should be warmer than B.
Observation: B is measured at a higher temperature than A.

Only the second belongs in the results table.

Alicia may later discover that the coverings were swapped, one thermometer was misread, the starting temperatures differed, one cup was measured later, or the result was genuine. But she cannot know which explanation is correct by changing 39 to 42 or by deleting the inconvenient reading. The surprising reading is exactly the evidence that tells her a check is needed.

Prediction Is a Model. Observation Is a Receipt.

A prediction is useful because it exposes your current scientific model. It says, in effect: “Given these conditions and this concept, I expect this outcome.”

An observation is useful for a different reason. It is the receipt returned by the world. It says: “Under this method, at this time, with this instrument or observation rule, this is what was recorded.”

If the receipt always has to match the prediction, the investigation cannot teach you anything new. The prediction would be grading the evidence instead of the evidence testing the prediction.

That is why a mismatch is not automatically a failure. Sometimes the prediction was based on an incomplete concept. Sometimes the method was weak. Sometimes natural variation matters. Sometimes the measurement was mishandled. Sometimes the result is simply not what the learner expected. Science needs the original evidence intact before it can distinguish these possibilities.

Before Looking at the Result, Define What Counts

The easiest time to protect an observation is before the observation is made.

Suppose Beatrice is comparing how quickly two wet cloths dry. If her observation rule is only “looks dry”, expectation can easily leak into judgement. If she already believes Cloth X will dry first, “nearly dry” may become “dry” for X but not for Y.

A better investigation states the evidence job more clearly. Depending on the actual question and method, the learner might measure mass at stated times, record the time when a defined observable condition is reached, or use another suitable measure. The point is not that every investigation needs a number. The point is that the observation criterion should be stable enough that the expected winner does not receive a different rule.

Before reading a result, ask:

What scientific object or set-up am I observing?
What quantity, event or visible feature am I supposed to record?
What unit or category is used?
At what time or stage is the observation made?
What exactly counts as the event or category?
Will I apply the same rule to every comparable set-up?

These questions make expectation less powerful because the evidence has already been given a job.

The Core Reasoning Chain

For this learning job, use the full PSLE Science reasoning chain without skipping the evidence step:

OBSERVE / READ GIVEN INFORMATION → IDENTIFY THE SCIENTIFIC OBJECT OR RELATIONSHIP → DISTINGUISH OBSERVATION FROM INFERENCE → SELECT THE RELEVANT CONCEPT → EXPLAIN THE CAUSAL MECHANISM → CONNECT TO THE QUESTION’S CONDITION → STATE THE OUTCOME → CHECK AGAINST THE EVIDENCE.

Notice where observation sits: at the beginning and again at the end. Evidence starts the reasoning and checks the reasoning. If the learner replaces the first evidence with an expected result, every later step can be beautifully written and still be built on the wrong foundation.

Worked Case 1: The Scale Says the “Wrong” Number

Ciara predicts that a sample will have a greater mass after a stated process. She expects the balance to show about 68 g. The display actually reads 64 g.

A weak reaction is: “It should be 68 g, so I probably read it wrongly.”

A stronger reaction is: “The display reads 64 g. I will record 64 g, then check whether the balance was zeroed correctly, whether I measured the intended sample, whether the timing and procedure matched, and whether a repeat is justified.”

The important sequence is record first, investigate second. A method check may later show that 64 g was invalid. If so, the correction needs a reason. “I expected 68 g” is not that reason.

Worked Case 2: A Qualitative Observation Can Be Biased Too

Denise is comparing two otherwise similar set-ups and expects the liquid in Set-up P to become darker. At the end, both samples look quite similar. She writes “P is much darker” because that is what her concept notes led her to expect.

This is not a vocabulary problem. It is an evidence problem.

If the method uses a qualitative category, the category still needs a consistent basis. Denise should record the observation that the method supports. If the difference is too small to distinguish reliably, she should not manufacture a larger difference from expectation. A suitable improvement might be to use a clearer observation criterion or a measurement method capable of resolving the relevant difference, but that is a method-design decision made after the evidence limitation is recognised.

Worked Case 3: A Diagram Can Trigger Expectation Bias

Emily studies a diagram showing two plants, one drawn taller than the other. She remembers a familiar lesson in which one particular condition increases growth, so she immediately concludes that the taller drawing proves that condition caused greater growth.

But the question never states that the diagram is drawn to scale, and no measured height is supplied.

Emily’s scientific knowledge may be relevant later. Her observation must still respect the representation. A picture that merely looks taller is not automatically a measured difference. She first asks what information the diagram actually encodes through labels, measurements, captions or stated relationships. Only then does she infer or explain.

This is the same integrity problem in a different form: familiar science must not manufacture evidence that the question did not give.

Worked Case 4: When a Prediction Is Correct for the Wrong Reason

Faith predicts correctly that Set-up R will give a larger measured outcome than Set-up S. She feels confident because the result matches. During checking, however, she discovers that her reason depended on a condition that the question never supplied.

The matching result does not rescue the reasoning. A prediction can match by chance, by an incomplete model or by a mistaken assumption. The evidence receipt says that R produced the larger outcome under the stated conditions. It does not say that Faith’s original explanation was valid.

This is why scientific integrity protects the learner from two errors, not one: forcing evidence to match a prediction, and assuming a matching prediction proves the explanation.

Failure Signatures: How This Weak Link Looks in Student Work

Expectation bias often hides inside work that looks neat. The earliest weak link may appear before any formal explanation is written.

Observable student behaviour	Likely weak link	Smallest useful repair
Changes a reading because it “cannot be right”	Expectation has become evidence	Record the displayed or observed result first; check the method separately
Rounds one value upward and another downward to make the expected pattern	Rounding is being used as conclusion repair	Use the stated scale or justified rounding rule consistently
Leaves out the one repeat that breaks the pattern	Selective evidence use	Keep all relevant valid repeats, then investigate the unusual one
Describes an ambiguous qualitative change more strongly in the expected set-up	Observation criterion is unstable	Define and apply the same observation criterion
Adds a condition that was never given because “that is how this topic works”	Scientific knowledge has overwritten question evidence	Separate given conditions from inferred conditions
Sees the expected trend and stops checking units or labels	Confirmation has replaced verification	Perform the same evidence check whether the result is expected or surprising

The Earliest Weak-Link Diagnosis

When a student’s final answer is wrong, do not begin by rewriting the entire science explanation. Find the earliest point at which the record stopped representing the evidence.

Check in this order:

Object: Was the correct set-up, specimen or quantity read?
Observation: Was what was actually shown or measured recorded?
Unit and time: Was the value attached to the correct unit and stage?
Observation versus inference: Did an explanation sneak into the result?
Condition: Was an unstated expected condition added?
Concept and mechanism: Only after the evidence is stable, was the correct science applied?

If the error occurred at step 2, repairing step 6 alone will not solve the learner’s problem. The answer may improve on that one familiar question but fail again when the next result is surprising.

Observation Is Not the Same as Inference

This distinction becomes especially important when expectation is strong.

Observation: “The thermometer reading for Set-up B is 42 °C after the stated time.”

Inference: “Set-up B may have reduced the rate of cooling more effectively under these conditions.”

Explanation: A causal account connecting the relevant property or condition to energy transfer and the measured outcome, if the evidence and question support that account.

If the learner writes the inference as though it were directly observed, the record becomes stronger than the evidence. If the learner writes the expected explanation before reading the observation carefully, the explanation can pull the observation toward itself.

A Scientific Notebook Has Two Different Columns in the Mind

You do not need a special notebook format in an examination. During learning, however, it helps to make the separation visible.

Before evidence	After evidence
My prediction and scientific reason	The observation or measurement actually obtained
What I expect to happen	What happened under the stated method
What result would support my model	What the evidence can and cannot support

After both columns exist, compare them. Do not merge them.

This simple separation is powerful because it makes revision honest. If your prediction fails, you can discover whether the weak link was concept knowledge, a hidden assumption, method quality, measurement, natural variation or interpretation. If the prediction record has already been edited to match the result, that diagnostic information disappears.

When Is It Legitimate to Correct a Recorded Value?

Protecting evidence does not mean treating every first-written number as sacred.

A value can be corrected when there is evidence that the record itself was wrong—for example, a transcription error, a clearly misread scale, a value copied into the wrong row, or a reading that the method identifies as invalid because a stated procedure failed. The reason for correction must come from the record or method, not from disappointment with the pattern.

Suppose a balance displays 56.2 g and the student accidentally writes 65.2 g. Looking back at the display or original record gives a reason to correct the transcription. That is different from changing 56.2 g to 58.2 g because 58.2 would make the graph smoother.

During learning, preserve a trace of genuine corrections when practical: note that a value was corrected and why. A corrected value is not automatically an additional repeat. It replaces or qualifies the earlier record according to what actually happened.

Rounding Is Not a Tool for Making the Pattern Behave

Students sometimes change evidence without noticing because the change is hidden inside rounding.

If two readings are 4.46 and 4.44 on a scale or instrument that supports those values, you cannot round one upward and the other downward simply to create a larger gap. Use the same justified rule for comparable values. More importantly, do not report more precision than the measuring method supports.

The scientific question is not “Which rounded value helps my explanation?” It is “What precision does this measurement legitimately support, and what conclusion survives that limit?”

A Surprising Result Is a Signal to Check, Not a Licence to Delete

An unexpected result creates a new job: check what happened. That job begins after the result is honestly preserved.

You might check whether:

the correct object and set-up were measured;
the scale, zero point and units were read correctly;
the starting conditions were comparable;
the intended changed condition was actually applied;
a controlled condition drifted;
the observation was taken at the correct time;
the result repeats under the same method;
natural specimen variation could matter;
the scientific explanation needs revision.

Notice what is not on the list: “Change the result until it looks right.”

Do Not Cherry-Pick the Convenient Repeats

Suppose four valid repeated measurements are 18, 19, 18 and 23 units. The value 23 deserves attention because it differs from the others. It does not deserve automatic deletion.

First preserve the repeated evidence. Then ask whether there is a documented reason to treat one measurement differently: a recording error, method failure, apparatus problem or other identifiable event. If no such reason exists, the variation itself is part of the evidence that must be interpreted.

Nor should a learner blindly “average everything” because an average sounds scientific. Whether results should be combined depends on whether they are comparable measurements of the same scientific quantity under the relevant conditions and whether the task calls for such a summary. Data integrity comes before calculation.

Expectation Can Enter Through Language

Not all evidence distortion changes a number. Sometimes it changes a word.

Consider the difference between these records:

“The colour became slightly darker.”
“The colour became much darker.”
“The colour changed.”
“The colour did not visibly change under the stated observation conditions.”

If the method gives no quantitative colour scale, the learner should not turn a subtle change into a dramatic one because a chapter prediction suggests a strong effect. Scientific vocabulary should sharpen meaning, not amplify expectation.

Expectation Can Enter Through What You Choose Not to Notice

A learner may record the evidence that supports a prediction and ignore the evidence that complicates it.

Imagine a table in which three observations fit an expected relationship but a fourth condition does not. If all four are relevant and valid, the fourth is not optional. It may reveal a turning point, a limiting condition, an uncontrolled factor, variation or a mistaken model. The correct interpretation depends on the evidence, but the first duty is to retain the evidence.

Good scientific reasoning asks, “What does the whole relevant record show?” rather than “Which part lets me say what I planned to say?”

A Counterexample: When the Expected Result Really Is the Correct Result

Suppose Kai Kai predicts that Set-up M will have a higher reading than Set-up N. The recorded evidence shows exactly that. Does integrity matter less?

No. The same checks still apply.

Confirmation can make learners careless. They may stop checking units, overlook that the values belong to different times, accept a graph whose axes were misread, or treat a familiar scientific fact as though it proves the cause. An expected result should be recorded faithfully, then checked with the same discipline as an unexpected one.

Objectivity is not “distrust every result.” It is “do not use agreement with expectation as a substitute for evidence quality.”

Model and Measurement Limits Still Apply

Honest recording cannot make a weak measuring method stronger than it is.

A ruler with coarse divisions cannot support an extremely precise length. A photograph taken from different perspectives may not support a direct size comparison. A qualitative category may not distinguish a tiny difference. A sensor may have a limited range. A diagram may be schematic rather than to scale.

The learner’s job is therefore not only to record faithfully but also to keep the claim within the evidence boundary. If the method can only show “no detectable difference under this measurement”, do not upgrade it to “the two systems are scientifically identical”. If the instrument cannot distinguish two close values, do not invent a difference merely because the theory predicts one.

Practice 1: Record First, Explain Second

Original practice situation: A learner predicts that Set-up A will produce a greater value than Set-up B. At the end of the investigation, the measurements are A = 31 units and B = 34 units. The learner has no evidence of a measurement or recording fault.

Question: What should be entered in the results table, and what should happen next?

Explained answer: Enter A = 31 and B = 34 because those are the observations obtained. Then compare the result with the original prediction and examine the method, relevant concept and conditions to explain the mismatch. Do not reverse the values merely because the prediction expected A to be larger.

Practice 2: A Genuine Recording Error

Original practice situation: A digital instrument displays 7.6. The learner accidentally copies 6.7 into the table, then notices the transcription while the original display record is still available.

Question: Is changing 6.7 to 7.6 dishonest because the first number was already written?

Explained answer: No. The evidence shows that 6.7 was a copying error. Correcting the transcription makes the table more faithful to the observation. During learning, keep a trace of why the correction was made when practical. The justification is the original evidence, not the desired pattern.

Practice 3: The Result Is Too Close to Call

Original practice situation: Two qualitative observations look almost the same. The student expected Q to show a stronger change than P, but the method has no defined scale capable of distinguishing such a small difference.

Question: Should the student record Q as definitely greater?

Explained answer: Not on expectation alone. The student should describe only the difference the method can support. If the scientific question requires a finer comparison, the method may need a more suitable measurement or observation criterion.

Practice 4: The One Awkward Repeat

Original practice situation: Five comparable repeated readings are 25, 24, 25, 31 and 24. The prediction expected values near 25.

Question: Can 31 be deleted because it does not fit the prediction?

Explained answer: No. Preserve it first. Check whether there is evidence of a measurement, procedure or recording problem and examine repeat evidence. The value may be unusual, but “does not match prediction” is not sufficient evidence that it is invalid.

Practice 5: Science Knowledge vs Given Evidence

Original practice situation: A student remembers a correct scientific relationship from revision. In the question, however, one condition required for that relationship is not stated and cannot be inferred from the evidence.

Question: May the student treat the missing condition as given because it would make the familiar relationship work?

Explained answer: No. Scientific knowledge helps interpret evidence; it does not authorise the learner to invent a missing condition. The answer should remain within the information and scientifically justified inferences that the question supports.

Misconception Repair: “If the Result Is Strange, the Experiment Failed”

A strange result can come from many places. It may expose a weak method. It may reveal an unusual trial. It may show natural variation. It may mean the learner’s scientific model is incomplete. It may be a simple recording error. Or it may be a valid result under the stated conditions.

Calling the whole investigation a failure before checking destroys information.

A stronger model is:

Unexpected result → preserve it → verify what was measured → check the method and comparison → inspect repeat evidence → revise the conclusion or explanation only as far as the evidence justifies.

Misconception Repair: “Objectivity Means Having No Prediction”

No. Prediction is part of scientific inquiry. The problem is not having an expectation. The problem is allowing the expectation to decide the observation.

In fact, a clear prediction can improve learning because it gives the evidence something specific to test. The scientific strength comes from keeping the prediction visible when the result arrives rather than rewriting the prediction or the result to make them agree.

Misconception Repair: “Objective Means Numbers Only”

Scientific evidence can be qualitative or quantitative. A numerical measurement can be badly handled; a descriptive observation can be carefully defined and consistently recorded.

Objectivity here means seeking and handling evidence so that the learner’s preference does not quietly determine the result. Depending on the scientific job, that may involve a suitable numerical measurement, a stable visible criterion, repeated observations, neutral labels, consistent timing or another method that makes the evidence less dependent on expectation.

A Useful Home-Practice Technique: Neutral Labels

For some home or tuition investigations, a parent or tutor can label two prepared samples simply A and B rather than telling the learner which one is expected to produce the stronger effect. The learner records the observations first, then the preparation is revealed.

This is an optional learning technique, not a PSLE examination rule. Its purpose is to help the learner feel what objective observation is like: the evidence is recorded before a preferred explanation can steer the description.

Use the technique only when it preserves the scientific question and is safe and practical. Do not hide information the learner actually needs in order to interpret the measurement correctly.

How This Transfers Across the Five Primary Science Themes

The 2023 Primary Science syllabus organises Core Ideas through Diversity, Cycles, Systems, Energy and Interactions, and the themes connect rather than functioning as isolated islands. The integrity job travels across all of them.

In Diversity, expectation can make a learner classify an unfamiliar specimen from resemblance instead of the supplied characteristics. In Cycles, expectation can make the learner invent a missing stage instead of distinguishing what is observed from what is inferred. In Systems, a familiar component can tempt the learner to assign a function not supported by the particular diagram. In Energy, an expected direction of transfer can make a learner ignore a measurement that contradicts the initial model. In Interactions, a familiar cause can be credited even when several conditions changed together and the evidence cannot isolate it.

The science concepts differ. The learner job stays the same: read what the evidence actually says before asking what the concept can explain.

Unfamiliar Transfer Test

Do not test this guide by memorising the cup, cloth or balance examples. Test whether the reasoning transfers.

Try this original situation:

A learner predicts that Device X will produce a larger change than Device Y because of a scientific relationship learned in class. The final readings show a slightly larger change for Y. A classmate says, “That cannot be right. Just swap the labels because X is supposed to be better.”

Without knowing the underlying topic, explain what the learner should do next.

A transferable answer should say something like: preserve the readings with the correct labels; check that the values, units, timing and set-ups were recorded correctly; inspect whether the method and relevant conditions made the comparison valid; use repeats or additional checks only when justified; then revise the scientific explanation or conclusion according to the evidence. The learner should not swap labels merely to satisfy the expected relationship.

If you can give that answer without needing to know whether X and Y are cups, materials, plants, circuits or another context, you are learning the underlying scientific reasoning job rather than the surface example.

Delayed Independent Return Test

Return to this idea after a delay with your notes closed.

On a blank page, answer four questions:

What is the difference between a prediction and an observation?
What should you do if a result does not match your prediction?
When is a correction to a recorded value justified?
Why can a matching prediction still be based on wrong reasoning?

Then solve one new question in which the expected and observed results disagree. Do not use the same objects as the worked examples above. If you can preserve the evidence, diagnose the method or reasoning, and rebuild the explanation independently, the learning has survived a change of surface.

Answer and Checking Receipts

Before declaring this learner job secure, look for receipts in the student’s actual work.

The prediction is recorded separately from the result.
The result preserves the correct object, value or category, unit and time.
A surprising result is not silently changed or omitted.
A correction has an evidence-based reason.
Rounding and observation criteria are applied consistently.
Given information is not replaced by a familiar expected condition.
The learner can explain what the evidence can and cannot support.
The learner can repeat the same discipline on an unfamiliar context after a delay.

These receipts matter more than whether the student can repeat the sentence “be objective”.

Parent and Tutor Teaching Guide

Parents and tutors can accidentally train expectation bias when they react to a surprising result too quickly.

If a learner says, “This answer must be wrong because we learned that X should happen,” resist immediately supplying the expected answer. Ask first: “What did the question or measurement actually give you?”

Then move through a short diagnostic conversation:

Evidence: Show me exactly what was observed, measured or stated.
Provenance: Which set-up, time, unit or part does it belong to?
Expectation: What did you predict before seeing that result?
Mismatch: Where exactly do prediction and evidence differ?
Method check: Is there a real reason to doubt the measurement or procedure?
Science: If the evidence remains valid, what part of your explanation needs revision?

Avoid praising only predictions that turn out correct. Praise the stronger scientific behaviour: accurate reading, honest recording, careful comparison, willingness to revise and the ability to say “the evidence does not yet decide”.

For practice, deliberately include some original examples in which the intuitive or predicted result is not the recorded result. The aim is not to trick the learner. It is to make evidence discipline visible. Afterwards, change the context and see whether the student still protects the record without being reminded.

Also teach the opposite case: give a result that matches expectation but contains a unit, timing or comparison flaw. This prevents the learner from assuming that familiar-looking results deserve less checking.

Common Traps to Avoid

“Theory says so.” A correct concept does not authorise inventing a measurement.
“One odd result can be ignored.” An unusual result must first be preserved and investigated.
“Average makes data scientific.” Averaging cannot repair incomparable or invalid data.
“More decimal places are more objective.” Precision must come from the measurement, not formatting.
“Objective means emotionless.” The learner can be surprised or disappointed; the scientific requirement is that preference does not rewrite evidence.
“If the prediction matches, the explanation is proved.” A correct outcome can still come from wrong reasoning.
“If the prediction fails, the concept is false.” First inspect conditions, measurement, method and the exact scope of the model.

Useful Internal Routes

This guide stops at the point where evidence has been protected. Continue with the next owner that matches the problem you actually have.

Authoritative References and Evidence Boundary

Singapore Examinations and Assessment Board — PSLE Formats Examined in 2026. This is the current SEAB route to the revised 2026 Science format and syllabus information.

Singapore Examinations and Assessment Board — 2026 PSLE Science Syllabus. The syllabus states that the examination assesses the 2023 Primary Science syllabus and identifies Knowledge with Understanding plus Application of Knowledge and Scientific Inquiry, including prediction/hypothesis, interpretation/analysis, evaluation of observations/information/methods and communication of explanations/reasoning.

Ministry of Education, Singapore — Primary Science Teaching and Learning Syllabus 2023. The syllabus provides the connected thematic and inquiry frame used in this guide and explicitly includes scientific values and attitudes such as integrity and objectivity.

Royal Society of Chemistry Education — Tricky Tracks: observation and inference in science. This classroom resource supports the broader science-education distinction between what is observed and what is inferred. It is not a Singapore examination authority and is used here only as supporting science-education context.

No source above establishes a compulsory PSLE wording formula for “objective” answers, and this guide does not claim one. The purpose is to strengthen a learner’s scientific evidence handling within the current curriculum frame.

The Quiet Return

The most important result in an investigation is not the one you hoped to see.

It is the one the evidence can honestly support.

Your prediction matters because it gives your scientific model something to risk. Your observation matters because it gives reality a voice in the argument. Keep them separate long enough for one to teach the other.

When the result surprises you, do not rescue the prediction first. Protect the evidence. Check the method. Rebuild the explanation. Then return later and see whether your new model survives a different question.

That is not merely a way to answer PSLE Science. It is one of the habits that makes science scientific.

Small Group Tutorials