How to Choose Similar Specimens for a PSLE Science Investigation Without Cherry-Picking the Result

Wait, What? “Use More Plants” Can Still Produce Weak Evidence if You Choose the Plants Badly

A learner wants to test how a condition affects plant growth.

They decide to improve the investigation by using ten plants instead of one.

Good idea?

Maybe.

Now imagine the learner chooses the ten strongest, tallest plants for Treatment A and ten smaller, weaker plants for Treatment B.

The sample is bigger, but the comparison is worse.

More specimens strengthen an investigation only when the specimens are chosen in a way that fits the scientific question and does not quietly favour the result.

Natural things vary. Leaves differ. Seeds differ. Plants differ. Pieces of natural material can differ. Even manufactured objects may have small differences. Good inquiry does not pretend those differences do not exist. It chooses specimens carefully enough that one unusual item—or one biased choice—does not decide the conclusion.

Quick Answer

Before selecting specimens for a PSLE Science investigation:

State what group the conclusion is supposed to describe. Similar leaves? Seeds of one type? Pieces of one material?
Choose relevant starting characteristics that should be comparable. Size, age, type, condition or starting measurement—only when these could affect the outcome.
Set the selection rule before seeing the result. Do not choose “good-looking” specimens after you know which result you want.
Use several suitable specimens when natural variation matters.
Avoid choosing only extremes. The biggest, smallest, healthiest or easiest-to-reach items may not represent the intended group.
Apply the same selection rule to all compared groups.
Keep the conclusion within the specimens and conditions actually represented.

Use this reasoning route:

STATE THE SCIENTIFIC QUESTION → DEFINE THE SPECIMEN GROUP → IDENTIFY RELEVANT STARTING FEATURES → SET A FAIR SELECTION RULE → CHOOSE SPECIMENS BEFORE RESULTS ARE KNOWN → APPLY THE SAME RULE ACROSS CONDITIONS → RUN THE TEST CONSISTENTLY → INSPECT VARIATION → CHECK WHETHER ONE UNUSUAL SPECIMEN DOMINATES → STATE A BOUNDED CONCLUSION.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: how a Primary 5 or Primary 6 learner chooses suitable similar specimens for a PSLE Science investigation without cherry-picking specimens that make one condition look better or worse.

It does not replace the guide on repeated trials versus more specimens. That page decides whether more specimens are needed. This page owns the next question:

If I use more specimens, how do I choose them so they provide fair evidence for the question?

Why This Matters in the 2026 PSLE Science Frame

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

Specimen choice belongs inside method evaluation. If the selected specimens differ systematically before the test, the final comparison may have more than one possible explanation.

First Distinction: A Specimen Is One Case From a Larger Group

A specimen is one object, organism or piece selected for observation or testing.

Examples:

one leaf from a plant;
one seed from a packet;
one seedling from a tray;
one strip cut from a material;
one shell, fruit or natural object;
one piece from a batch of manufactured items.

If your conclusion is only about that one specimen, selection is simpler. If your conclusion is meant to say something about the larger group, specimen choice becomes more important.

Representative Does Not Mean Identical

Learners sometimes think a fair investigation needs specimens that are perfectly identical.

Real organisms are not identical. Two leaves can differ slightly in area, age or condition. Two seeds can have different biological histories. Natural variation is part of the living world.

The goal is usually comparability, not imaginary sameness.

Ask:

Which differences could strongly affect the measured outcome?
Can those starting features be kept reasonably comparable across groups?
Can several specimens reduce dependence on one unusual case?

What Is Cherry-Picking?

Cherry-picking means selecting cases because they support the result you want while ignoring suitable cases that do not.

In a Primary Science investigation, this can happen when a learner:

chooses only the tallest plants for one condition;
chooses only the largest leaves because they give a clearer effect;
removes a specimen after seeing that its result does not fit the pattern without a scientific reason;
selects the healthiest seeds for one group and random seeds for another;
keeps repeating selection until the expected result appears.

The problem is not that the chosen specimens are “bad”. The problem is that the selection rule is connected to the desired outcome.

The Pre-Selection Rule

A strong habit is to decide the selection rule before the outcome is known.

For example:

Choose seedlings of the same species whose starting heights are between 8 cm and 10 cm, then assign comparable numbers to each condition.

This rule is based on a relevant starting feature, not on the final result.

Do not turn this into a universal recipe. The appropriate criterion depends on the scientific question.

Worked Example 1 — Tall Plants Versus Short Plants

Question: How does light level affect height increase over seven days?

Weak selection:

Bright-light group: five plants starting at 12–14 cm.
Low-light group: five plants starting at 6–8 cm.

Even if final heights differ, starting height already differs strongly between groups.

Better design:

choose plants of the same intended type;
use starting heights from a comparable range;
record each starting height;
compare height change rather than final height alone when the question is about growth.

The goal is not perfect cloning. It is to prevent starting height from becoming the obvious alternative explanation.

Worked Example 2 — Choosing Only the Largest Leaves

A learner investigates water loss using leaves. They choose the five largest leaves for Condition A because “the effect will be easier to see”. For Condition B, they use five leaves chosen without considering size.

This selection is biased if leaf size can affect water loss.

A fairer approach is to apply a comparable selection rule to both groups, such as choosing leaves within a stated size range or matching leaves by a relevant characteristic before assigning them to conditions.

But do not make the range so narrow that your conclusion silently becomes “only leaves exactly this size behave this way”. Selection criteria also shape claim scope.

Worked Example 3 — Healthy Seeds Only in One Group

A class compares two germination conditions. For Group P, the learner chooses seeds that look large and intact. For Group Q, the learner uses the remaining seeds.

If seed condition affects germination, the groups do not begin comparably.

A better approach applies the same eligibility rule before dividing seeds into groups.

For example, obviously damaged seeds might be excluded from both groups if the investigation is about ordinary intact seeds and the rule is set before results are observed.

Worked Example 4 — Material Strips From One Convenient Corner

A natural sheet of material is not perfectly uniform. The learner cuts every strip for Condition A from the thick edge and every strip for Condition B from the thin edge.

Thickness may now be entangled with the tested condition.

Specimen selection matters even when the object is not living.

Worked Example 5 — The Most Convenient Specimens

Suppose a learner measures leaves from a plant but chooses only those easiest to reach at the outer edge.

Convenience is not automatically bias. But if outer leaves systematically differ from inner leaves in light exposure, age or size, the sample may not represent the intended group.

Ask whether the selection method favours a particular kind of specimen.

Worked Example 6 — Removing an “Ugly” Result

Five specimens give results 8, 9, 8, 9 and 5. The learner deletes the specimen with 5 because “it spoils the pattern”.

That is not a valid reason.

Investigate:

Was the specimen selected using the same rule?
Was the method carried out correctly?
Was there a recording error?
Was the specimen damaged or genuinely different in a scientifically relevant way?
Does repeating or using more specimens show that 5 is rare but real?

An inconvenient result does not become invalid merely because it is inconvenient.

Matching Specimens Can Be Good Science

Choosing specimens with comparable starting features can improve a fair comparison.

Examples:

similar starting plant height;
same species or stated type;
similar initial mass;
same material and similar dimensions;
similar developmental stage when relevant.

This is not cherry-picking when the rule is scientifically relevant, applied consistently and set before the result is known.

But Over-Matching Can Narrow the Question

If you choose only specimens within an extremely narrow range, your evidence may apply mainly to that narrow range.

Example: selecting only leaves exactly 10.0 cm long might reduce variation, but it may also make the sample unlike the normal range of leaves you want to discuss.

Good specimen selection balances comparability with the scope of the scientific question.

Same Type Does Not Mean Same Individual

Two plants can be the same species but different individuals. Two leaves can come from the same plant but differ in age and position. Two seeds can come from the same packet but have different histories.

Use language carefully:

same individual means one organism followed over time;
same type/species means different individuals belonging to the same category;
similar specimens means selected cases share relevant starting features.

More Specimens Versus More Repeats

These solve different evidence problems.

Method choice	What it helps reveal
Repeat the same trial	Whether the procedure gives a reasonably stable result when run again
Use more similar specimens	Whether the pattern survives natural differences among specimens
Use better specimen selection	Whether compared groups begin without a built-in selection advantage

An investigation can use many specimens and still select them badly.

Selection Before Assignment

A useful sequence is:

DEFINE SUITABLE SPECIMENS → SELECT THEM USING ONE RULE → THEN DIVIDE THEM INTO COMPARISON GROUPS.

This makes it harder to favour one condition by selecting different kinds of specimens for it.

At Primary level, you do not need advanced randomisation theory. The durable principle is that the selection process should not know which outcome you want.

When Simple Random Choice Can Help

If suitable specimens have already been identified and there is no scientific reason to choose one over another, a simple unbiased allocation can help avoid favouring one group.

Examples include drawing coded specimen labels or alternating suitable items after they have been ordered by a relevant starting characteristic.

This is a learning principle, not an official PSLE requirement. The actual question may specify the selection method.

When Deliberate Matching Is Better Than Random Choice

If one starting characteristic strongly affects the outcome, deliberate matching can make the comparison clearer.

Example: pair seedlings with similar starting heights, then place one member of each pair into each condition.

Again, the choice is guided by the scientific question, not by a universal rule.

Selection Bias Can Hide Inside “Similar”

A learner may say, “I chose similar specimens,” but similar in what way?

similar size?
same species?
similar age?
similar mass?
same source?
similar condition?

Only similarities that matter to the outcome need special control. Listing irrelevant similarities wastes effort and can hide the real confound.

Specimen Selection and Claim Scope

Your conclusion cannot travel farther than your specimen group without additional justification.

If you test only young leaves from one plant under one set of conditions, the evidence does not automatically establish what all leaves from all plants will do.

Use bounded language:

“For the suitable specimens tested under these conditions…”

The wording is not compulsory. The evidence boundary is.

The Selection Audit

What group is the scientific question about?
What counts as a suitable specimen?
Which starting features could affect the outcome?
Was the selection rule decided before the results?
Was the same rule used for every condition?
Were only extreme or convenient specimens chosen?
Were specimens removed after results were seen?
Do enough specimens represent ordinary variation?
Does one specimen dominate the pattern?
Does the final conclusion match the represented group?

The Earliest-Weak-Link Diagnostic

Failure signature	Earliest weak link	Repair
“Use more plants.”	Quantity increased without a selection rule.	Define which plants are suitable and how both groups will be chosen.
“Pick the healthiest plants for the treatment.”	One group receives a starting advantage.	Apply the same suitability rule before assigning groups.
“Delete the one result that does not fit.”	Outcome-based exclusion.	Investigate the result; exclude only for a clear pre-defined scientific/method reason.
“All specimens must be identical.”	Natural variation treated as experimental failure.	Seek relevant comparability, not impossible identity.
“Choose only the biggest leaves because the effect is clearer.”	Extreme-case selection.	Choose a rule that fits the intended specimen group.
“Same species means same starting state.”	Category equality confused with individual equality.	Check relevant starting measurements or conditions.
“Ten specimens prove the result for every specimen.”	Sample size confused with universal generalisation.	Bound the claim to the represented group and conditions.

Misconception Repair — “More Specimens Automatically Mean Better Evidence”

More specimens can reveal natural variation and reduce dependence on one unusual case. But a large biased sample can still give misleading evidence.

Misconception Repair — “Random Means Careless”

An unbiased selection method can be deliberate and scientifically useful. The point is not to choose without thought. The point is to avoid choosing based on the outcome you hope to get.

Misconception Repair — “Matching Is Cheating”

Matching relevant starting characteristics can strengthen a fair comparison. It becomes problematic only when the matching rule is applied differently to groups or chosen after seeing the results.

Misconception Repair — “An Unusual Specimen Should Always Be Removed”

Unusual specimens may be real members of the group. They deserve investigation. Removal needs a defensible reason, not discomfort with messy evidence.

How This Appears in Multiple-Choice Questions

Identify what kind of specimen the question is about.
Check which starting features are relevant to the outcome.
Reject selection methods that favour one condition.
Reject “use the biggest/healthiest” unless that is the defined population.
Check whether selection occurs before or after results are known.
Choose the method that makes groups comparable while representing the intended specimen type.

How This Appears in Structured Inquiry Answers

A useful reasoning scaffold is:

Select several ______ that meet the same relevant starting criteria, such as ______, before assigning them to the different conditions. This reduces the chance that differences in ______, rather than the tested condition, explain the measured outcome.

This is not an official marking phrase. Use the actual scientific weakness in the question.

Practice Sequence

Sort specimen-selection methods into fair, biased or unclear.
For each biased method, identify which starting characteristic favours a group.
Rewrite the selection rule before seeing any outcomes.
Compare “more repeats” with “more specimens” and “better specimen selection”.
Use living specimens with natural variation.
Use pieces of natural material so the skill is not tied only to Biology.
Add one unusual result and decide whether it should be kept, checked or excluded for a real method reason.
Move to an unfamiliar investigation after a delay.

Unfamiliar Transfer Challenge

A learner wants to compare how two conditions affect the bending of strips cut from a natural material. They cut five thick strips for Condition A and five thin strips for Condition B because those pieces were easiest to obtain.

What is the problem?

Thickness may affect bending, so specimen selection has created another systematic difference between the groups.

What is a better method?

Select strips within a comparable relevant thickness range using the same rule, then apply the two test conditions. Record starting thickness if it remains scientifically relevant.

What still cannot be claimed?

The result does not automatically apply to every possible thickness, piece or source of that material.

Delayed Independent Return

Three to five days later, take a fresh investigation and answer without notes:

What specimen group is the question about?
What makes a specimen suitable?
Which starting features could affect the outcome?
Was the selection rule set before results were seen?
Was the same rule used across conditions?
Were extremes or convenient cases overrepresented?
Would more specimens help with natural variation?
Would repeated trials solve a different problem?
Does one unusual specimen need investigation?
How far can the conclusion generalise?

The Answer-Checking Receipt

Did I define the specimen group?
Did I identify relevant starting characteristics?
Did I avoid impossible “identical specimen” thinking?
Did I choose specimens before knowing the desired outcome?
Did I apply one selection rule across groups?
Did I avoid selecting only extremes or convenient cases?
Did I distinguish specimen selection from repeated trials?
Did I keep unusual results visible long enough to investigate them?
Did I avoid removing results simply because they hurt the pattern?
Did I bound the conclusion to the represented specimens and conditions?

Evidence and Model Limits

Sampling and experimental design become much more sophisticated at higher levels. Researchers may use formal random sampling, stratification, blocking, statistical power and population inference.

Primary Science does not require that machinery here. The durable learner principle is simpler:

Choose specimens by a fair rule that fits the scientific question before you know which result you want.

Also remember that even a well-chosen small sample has limits. Similar specimens reduce some variation; they do not prove that every member of a larger group behaves identically.

Useful Internal Routes

Parent and Tutor Teaching Guide

When a child says “use more plants”, ask:

“Which plants?”

Then ask:

“What group are we trying to learn about?”
“Which starting differences could affect the result?”
“How will you choose specimens before you see the outcome?”
“Will both groups use the same selection rule?”
“What would count as cherry-picking?”

Use a simple physical demonstration. Put ten paper strips of visibly different widths on a table. Ask the child to choose five for Group A and five for Group B. If they deliberately put all wide strips in one group, discuss how the groups already differ before the experiment begins.

Then let the learner design a fairer selection method. Repeat with plant pictures or seed cards.

Do not teach that every investigation must use random selection. Teach the deeper rule: specimen choice should follow the scientific question, be decided before results, and not quietly favour one comparison group.

Return after a delay with an unfamiliar material or organism. Mastery is shown when the learner asks about specimen selection before being prompted.

Authoritative and Research References

The research sources support broader sampling and inquiry reasoning. They do not create an official PSLE sampling formula or a compulsory number of specimens.

The Quiet Ending

A larger sample can make an investigation stronger.

But only if the sample was not chosen to manufacture the answer.

Choose the specimens before you know the winner. Let the evidence decide what happens next.

Small Group Tutorials