How to Decide Whether a PSLE Science Investigation Needs More Repeats or More Test Conditions

Wait, What? “Collect More Data” Can Mean Two Completely Different Repairs

A learner carries out an investigation and gets one result at each of three conditions.

The teacher asks, “How could the investigation be improved?”

The learner writes, “Repeat the experiment more times.”

Sometimes that is exactly right.

Sometimes it does almost nothing to fix the real weakness.

Repeats answer, “Would I get a similar result again at this condition?” More test conditions answer, “What does the relationship look like across the range?”

Those are different scientific questions.

If repeated readings at one condition vary wildly, repeating can reveal how stable the measurement is. If the investigation tests only two values, repeating those same two values twenty times still may not show whether the relationship is straight, curved, threshold-like, plateauing or changing direction between them.

“More data” is not one repair. The learner must diagnose what information is missing.

Quick Answer

Choose more repeats when the main problem is variation or uncertainty at the same tested condition. Choose more test conditions when the main problem is that the investigation has too few values to reveal the pattern, threshold, turning point or useful range. Use both when you need a reliable reading at several conditions.

Use this decision route:

STATE THE INVESTIGATION QUESTION → IDENTIFY WHAT WAS CHANGED → IDENTIFY WHAT WAS MEASURED → ASK WHETHER THE WEAKNESS IS VARIATION AT A CONDITION OR TOO LITTLE COVERAGE ACROSS CONDITIONS → REPEAT IF YOU NEED STABILITY → ADD CONDITIONS IF YOU NEED SHAPE OR RANGE → USE BOTH IF BOTH QUESTIONS MATTER → KEEP THE METHOD FAIR → STATE WHAT THE IMPROVEMENT WOULD ACTUALLY ALLOW YOU TO CONCLUDE.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: how a Primary 5 or Primary 6 learner decides whether an investigation should be improved by repeating measurements at the same condition, testing more values of the changed condition, or doing both.

It does not replace the canonical owners for fair tests, repeated investigations, variables, simple measurement, method evaluation or data interpretation. It owns the decision between two commonly confused improvements.

The key question is:

What uncertainty am I trying to reduce?

Why This Matters for the 2026 PSLE Science Frame

For examination from 2026, the revised PSLE Science paper assesses attainment in the 2023 Primary Science syllabus. The official assessment objectives include applying scientific facts, concepts and principles and applying scientific inquiry, including making predictions and hypotheses, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning.

When a question asks how an investigation could be improved, the strongest answer is not a memorised “repeat three times”. The learner should connect the proposed change to the weakness in the evidence and explain what better information the change would produce.

Two Different Dimensions of Better Evidence

Improvement	What changes?	Main information gained
Repeat at the same condition	Number of observations at one condition	How consistent or variable the result is there.
Test more conditions	Number or spread of changed-variable values	How the outcome behaves across a broader or finer range.
Do both	Observations per condition and coverage across conditions	Stability plus a better-defined pattern.

One improves depth at a point. The other improves coverage across the relationship.

Repeats: Ask Whether the Same Condition Gives a Stable Result

Suppose a ball is released from the same height three times and the measured travel distances are 118 cm, 121 cm and 119 cm.

Those repeated trials show that the result is reasonably similar across the three attempts under the stated setup.

Now imagine the readings are 82 cm, 121 cm and 147 cm. The large variation tells you something important: the method, release, surface, measurement or another condition may not be sufficiently consistent.

Repeating does not magically make the investigation “correct”. It exposes whether the same nominal setup keeps producing similar observations.

More Conditions: Ask Whether You Have Enough of the Relationship

Suppose a learner tests one plant in low light and one in high light. Even if each condition is measured very carefully, two conditions give only one comparison.

They do not reveal what happens at intermediate light levels. A response could change steadily, rise quickly then level off, show little change until a threshold, or behave differently across different ranges.

Testing more conditions can reveal the shape of the relationship.

The Point Versus Pattern Distinction

Repeats strengthen what you know about a point. More conditions strengthen what you know about a pattern.

This sentence is simple enough for a Primary learner but powerful enough to organise many inquiry questions.

Worked Example 1 — Cooling Water: More Repeats or More Times?

Original practice setup: A learner measures the temperature of hot water at 0 minutes and 20 minutes.

Question A: “Are the 20-minute measurements stable when the procedure is repeated?”

Useful improvement: repeat the same setup and compare the 20-minute readings.

Question B: “How does the temperature change during the 20 minutes?”

Useful improvement: measure at additional times such as 5, 10 and 15 minutes, while keeping the measurement procedure consistent.

Repeating only 0 and 20 minutes would not reveal the temperature path between them. Adding intermediate times answers a different missing-information problem.

Worked Example 2 — Finding a Threshold

A response is absent at Condition 20 and present at Condition 40.

If the learner wants to know whether those observations are reproducible, repeats at 20 and 40 are useful.

If the learner wants to locate the transition more closely, repeats alone are insufficient. The investigation needs additional conditions between 20 and 40.

For example, Conditions 25, 30 and 35 may narrow the region. The exact choices should suit the scientific context and practical measurement limits.

Worked Example 3 — A Noisy Measurement

A setup tests four surface types once each. One measured time is unexpectedly much larger than the others.

Adding ten new surface types does not first solve the question of whether that strange reading was stable.

A sensible first repair may be to repeat the measurement for the original surface under the same controlled conditions. If the unusual value repeatedly appears, it may be a genuine property of the setup. If it does not, inspect the method for error or inconsistency.

Worked Example 4 — Too Narrow a Range

A learner tests an elastic object at extensions of 1 cm, 2 cm and 3 cm and sees a simple increasing response.

The question is whether the same relationship continues across the safe tested range.

Repeating 1, 2 and 3 cm many times can strengthen confidence in those points. It cannot show what happens at 4 or 5 cm.

If scientifically and safely appropriate, more test conditions extend the evidence. The conclusion must still remain within the tested range and within Primary Science boundaries.

Worked Example 5 — Both Are Needed

An investigation measures how long a wet material takes to dry under four controlled airflow settings. The readings vary noticeably even when the same setting is repeated.

The learner wants both a trustworthy result at each airflow setting and a useful pattern across airflow settings.

Here, both dimensions matter:

repeat within each airflow condition to see the variation;
retain several airflow conditions to describe the pattern across the range.

The design becomes a grid: multiple observations at multiple conditions.

A Data Grid Helps You See the Difference

Condition	Trial 1	Trial 2	Trial 3
Low	12	13	12
Medium	18	17	19
High	25	24	26

Read across a row to inspect repeated observations at one condition.

Read down the condition levels to inspect how the outcome changes across the tested range.

These are two different evidence questions inside one table.

Do More Repeats Automatically Mean “More Accurate”?

No.

Repeating a flawed procedure can produce the same flawed measurement many times.

If a thermometer is read incorrectly in the same way every time, more repeats may make the pattern look consistent without making it correct. If an uncontrolled factor changes with every trial, repeats may reveal variation but not identify the cause.

Repeated observations are useful evidence about consistency. They do not replace sound measurement and fair-test design.

Do More Conditions Automatically Mean a Better Investigation?

No.

Ten poorly controlled conditions can be worse than four well-chosen conditions. More values can also exceed the safe or meaningful range of the apparatus or phenomenon.

The conditions should be chosen because they answer the investigation question, not because a longer table looks scientific.

Spacing Between Conditions Matters

If an effect changes rapidly between Conditions 30 and 35, testing 0, 50 and 100 gives broad range but poor local resolution around the important region.

A good follow-up may use closer test values where the relationship is changing most.

This connects to sparse sampling: you choose measurement spacing according to the question you need the evidence to answer.

Range and Resolution Are Different

Design	Strength	Weakness
0, 50, 100	Wide range	Large gaps; fine changes may be missed.
45, 50, 55	Fine local spacing	Narrow range.
0, 25, 50, 75, 100	Broader coverage with moderate spacing	Still may miss a sharp local transition.

There is no universal best spacing. The scientific question controls the design.

Repeat the Measurement, Not the Mistake

Before repeating, check whether the procedure itself is fit for purpose.

Is the instrument suitable?
Is the scale readable?
Is the starting condition reset?
Is the same method used each time?
Are relevant controlled conditions actually kept comparable?
Is the outcome measured at the same defined point?

Otherwise, repeated trials may simply repeat a methodological weakness.

Do Not Memorise “Repeat Three Times and Take the Average” as a Universal Rule

Repeated measurements can be useful, and an average can sometimes summarise repeated numerical results. But neither “three times” nor “always average” is a universal PSLE Science law.

Some results are categorical rather than numerical. Some repeated results contain an unusual value that should be investigated rather than automatically hidden inside an average. Some methods require a different summary because the observations are not directly comparable.

Use the question, data and method.

When More Conditions Matter More Than More Repeats

You need to tell a trend from a single comparison.
You need to locate a threshold more closely.
You need to see whether a response plateaus.
You need to find a turning point.
You need to test whether equal input steps produce equal output changes.
You need to know whether an observed relationship survives across a wider range.

When More Repeats Matter More Than More Conditions

Repeated readings at the same condition appear inconsistent.
A result looks unusual and you need to see whether it occurs again.
The procedure contains small unavoidable variations from trial to trial.
You need to judge whether one observed value is stable enough to represent that condition.
You suspect the measurement technique itself is variable.

When Both Matter

You want a pattern across several conditions and the measurements at each condition also vary.
You are comparing two competing explanations that predict different curves, not merely different endpoint values.
You need to locate a threshold but the detector response itself is variable.
You need enough coverage to see the relationship and enough repeats to know the points are not one-off observations.

The Method-Improvement Sentence Must Explain the Gain

Weak:

“Repeat the experiment more times for accuracy.”

Stronger reasoning:

“Repeat the measurement at each condition so the learner can check whether the result is consistent rather than depending on one trial.”

Or:

“Test additional values between 20 and 40 so the investigation can locate more closely where the observed response begins.”

The improvement earns its place because it fixes an identified evidence gap.

The Earliest-Weak-Link Diagnostic

Failure signature	Earliest weak link	Repair
“Repeat three times” appears in every method answer.	Improvement is memorised, not diagnosed.	Name what uncertainty the repeat would reduce.
Twenty repeats are proposed to find a threshold between two values.	Point stability is confused with range coverage.	Add intermediate test conditions.
Ten new conditions are proposed after one wildly variable measurement.	Range is expanded before stability is checked.	Repeat the problematic condition first.
“More data is better.”	No link between data type and question.	State whether you need repeat evidence or relationship evidence.
“Take the average” automatically.	Summary method is treated as a ritual.	Inspect the repeated results and the measurement type first.
More conditions are added but other variables also change.	Coverage improvement broke the fair comparison.	Preserve controlled conditions while extending the changed variable.
Repeats give similar values but the conclusion is still too broad.	Reliability at tested points is confused with generalisation beyond them.	Keep the conclusion within the tested range or add justified conditions.

Misconception Repair — Reliability Is Not the Same as Validity

An investigation can give very similar repeated readings yet still fail to answer the intended scientific question if it measures the wrong outcome or confounds important conditions.

Consistency is useful, but the evidence must also be relevant to the claim.

Misconception Repair — A Wider Range Does Not Guarantee a Better Mechanism

A wide data range can reveal more of the pattern. It does not by itself explain why the pattern occurs. The scientific mechanism still comes from the relevant concept, conditions and evidence.

Misconception Repair — More Trials Cannot Repair a Missing Variable

If the investigation never measured the outcome needed to answer the question, repeating the same incomplete method cannot create that missing evidence.

Misconception Repair — More Conditions Cannot Repair Poor Measurement

If every temperature is read from the wrong part of a scale, testing fifteen temperatures does not solve the measurement problem.

Question-Reading Protocol for Method Improvement

What scientific question is the investigation trying to answer?
What condition is deliberately changed?
What outcome is observed or measured?
How many conditions are tested?
How many observations are made at each condition?
Is the weakness variation at one condition, insufficient range, coarse spacing, or another method problem?
Would repeats answer the missing question?
Would additional conditions answer it?
Would both be needed?
Can the improvement be made without breaking the fair comparison?

Practice Sequence

Take ten short investigation descriptions.
For each, label the changed condition and measured outcome.
Classify the weakness: variation, insufficient range, coarse spacing, missing control, unsuitable measurement or other.
Choose “repeat”, “add conditions”, “both” or “neither”.
Write one sentence explaining what information the change would improve.
Change the investigation goal and see whether the best improvement changes.
Use a graph with a hidden threshold or turning point and decide what extra conditions are needed.
Return several days later with unfamiliar contexts.

Unfamiliar Transfer Challenge

A mystery material is tested at 10°C and 50°C. Its measured response is 4 units at 10°C and 18 units at 50°C. Each reading was obtained once.

You are asked two different follow-up questions.

Follow-up A: “Are these endpoint values stable?”

Repeat measurements at 10°C and 50°C under the same controlled method.

Follow-up B: “Does the response increase steadily between 10°C and 50°C?”

Test additional temperatures between the endpoints. Repeats may also be needed if measurements are variable, but repeats at only the endpoints cannot reveal the intermediate shape.

Delayed Independent Return

Four to seven days later, solve a fresh investigation without notes:

What is the changed condition?
What is the measured outcome?
How many conditions are tested?
How many observations are made at each?
Do I need stability at a point or shape across a range?
Would repeating fix the actual weakness?
Would more conditions fix it?
Would both be justified?
What can the improved evidence still not prove?

The Answer-Checking Receipt

Did I state the investigation question?
Did I distinguish repeated trials from additional changed-variable values?
Did I identify whether the weakness is variation or coverage?
Did I avoid “repeat three times” as an automatic phrase?
Did I explain what the proposed improvement would let me judge?
Did I preserve fair-test conditions?
Did I keep the conclusion within the tested range?
Did I avoid assuming that more data automatically means better evidence?

Evidence and Model Limits

School investigations simplify the much larger problem of experimental design. In professional science, decisions about replication, sample size, sampling density, measurement uncertainty and range can require formal statistical and domain-specific methods. Primary learners do not need that machinery here.

The durable Primary Science idea is simpler: match the evidence you collect to the uncertainty you need to reduce.

Repeats can reveal consistency. Additional conditions can reveal more of a relationship. Neither replaces a valid question, appropriate measurement, relevant controls or sound scientific reasoning.

Useful Internal Routes

Parent and Tutor Teaching Guide

When a child gives the automatic answer “repeat three times”, do not mark it wrong immediately. Ask:

“What problem would repeating fix here?”

If the learner cannot answer, the method phrase is memorised rather than understood.

Use pairs of investigations that look almost identical but require different repairs. For example:

same three conditions, but one reading is highly variable → repeat;
stable endpoints, but the question asks where a threshold occurs → add intermediate conditions;
variable readings and unclear threshold → both.

Ask the learner to justify the decision using the evidence gap, not the word “accuracy”.

After the learner can choose correctly in familiar experiments, change the topic. Use heat, forces, evaporation, light, plant responses or another appropriate Primary Science context. The inquiry decision should survive the surface change.

Authoritative and Research References

Singapore Examinations and Assessment Board — PSLE Formats Examined in 2026.
Singapore Examinations and Assessment Board — PSLE Science syllabus, for examination from 2026.
Singapore Ministry of Education — Science Teaching and Learning Syllabus, Primary, 2023.
Lazonder and Harmsen — Meta-Analysis of Inquiry-Based Learning.
Research on children’s control-of-variables strategy, scientific reasoning and measurement is used here as broader learning evidence rather than PSLE-specific marking policy.

The Quiet Ending

More observations are not automatically better observations.

The better question is: what do you still need to know?

If you need to know whether a point is stable, repeat it.

If you need to know what happens between points, add conditions.

And if you need both, design the evidence to do both jobs.

Small Group Tutorials