How to Decide Whether to Repeat One PSLE Science Trial, Repeat the Whole Investigation or Redesign the Method

Wait, What? “Repeat the experiment” is not one repair. It can mean three very different things—and only one may fit the problem.

A PSLE Science investigation goes wrong. One reading looks strange. A student forgets a step. The repeated results vary. Or the whole comparison was designed badly from the start. In all four cases, learners often write the same improvement: “Repeat the experiment.” That sounds scientific, but it hides the most important question: what exactly needs repairing?

Sometimes one trial should be repeated because that particular run was compromised. Sometimes the whole investigation needs more repeated evidence because the method is reasonable but the evidence is too thin or variable. Sometimes repeating the same procedure would only reproduce the same flaw, so the method itself must be redesigned. Choosing the correct repair scope is a scientific reasoning skill.

Quick Answer

Use this three-way decision:

Repeat one trial when there is a justified reason that one run was not carried out or recorded in the intended way.
Repeat the whole investigation when the method still answers the scientific question but you need more evidence about consistency, natural variation or whether the pattern appears again.
Redesign the method when the problem affects every trial: the comparison is unfair, the wrong quantity is measured, the instrument cannot answer the question, the controlled condition drifts, or the method tests something different from the stated question.

The durable rule is simple: repair the level where the failure begins. Do not use repetition to hide a design problem, and do not redesign an entire investigation because one local run had a clear procedural fault.

The PSLE Science Learning Job This Guide Owns

This guide owns one precise learner job: choosing the correct scope of repair after diagnosing an investigation problem. It does not own any particular scientific concept. It does not claim that “repeat”, “improve reliability” or any other phrase is a universal marking formula. It teaches how to decide whether the weakness lives in one trial, in the amount of evidence, or in the method itself.

For the 2026 PSLE, Standard Science assesses the 2023 Primary Science syllabus. The official assessment objectives include scientific inquiry such as interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. Method repair therefore belongs inside scientific thinking, not outside it as an exam trick.

First Separate the Investigation Into Levels

Think of an investigation as three nested levels:

One observation or trial: one run, specimen, measurement sequence or attempt under a stated condition.
The evidence set: the collection of valid trials or specimens used to judge a pattern.
The method design: the question, changed condition, measured outcome, controlled conditions, apparatus, timing and procedure that generate the evidence.

When something looks wrong, do not immediately change all three levels. Find the earliest level that failed.

Case 1: Repeat One Trial

Repeating one trial can be appropriate when there is a specific, evidence-based reason that this run did not represent the intended method. For example, a timer was started late, a quantity was recorded in the wrong row, a specimen was accidentally dropped, an apparatus part was disconnected, or a stated step was missed.

The key is that the problem is local. The question remains sensible. The method remains sensible. Other trials were carried out correctly. Repeating the affected run restores a missing or compromised piece of evidence.

Do not repeat one trial merely because its result is inconvenient. An unusual result is not automatically an invalid trial. Preserve it first, inspect the method and recording, compare it with the other evidence, and exclude or repeat it only when there is a justified procedural reason.

Case 2: Repeat the Whole Investigation

Sometimes every run was carried out as intended, but the evidence is still too limited. The learner may have only one trial for each condition, only one specimen, or a set of repeated values whose variation makes the pattern uncertain.

Here, repeating the whole investigation—or adding appropriate repeated trials or similar specimens—can strengthen the evidence. The purpose is not to force identical results. It is to see whether the relationship is reasonably consistent, whether natural variation is large, whether an apparent pattern survives another run, and whether the conclusion depends too heavily on one result.

A valid method with too little evidence is different from an invalid comparison. Repetition helps the first problem. It does not repair the second.

Case 3: Redesign the Method

If a flaw affects every trial, repeating the procedure simply creates more evidence from the same flawed design. That is when the method needs repair.

Typical method-level failures include:

two important conditions change together when the question is meant to test one relationship;
the chosen measurement does not actually answer the scientific question;
the instrument range or resolution is unsuitable for the expected values;
the measurement itself changes the system being investigated;
a controlled condition drifts during the investigation;
one set-up is measured for a different duration from another without scientific justification;
the procedure cannot be reproduced because quantities, timings or criteria are missing;
the conclusion asks about a relationship that the method never tested.

These are not “repeat more” problems. They are design problems. Fix the design first, then collect new evidence with the repaired method.

A Decision Tree You Can Actually Use

Question	If yes	Likely action
Is there a clear reason one particular run was carried out or recorded incorrectly?	The problem is local to one run.	Repeat that trial if appropriate.
Was the intended method followed, but the evidence is too sparse or variable to judge the pattern confidently?	The design may still be sound.	Repeat the investigation or add suitable repeats/specimens.
Would every new repeat contain the same unfair comparison, wrong measurement or method flaw?	The problem is structural.	Redesign the method before repeating.
Is the unusual result the only reason you want to repeat?	You may be deleting inconvenient evidence.	Preserve and investigate it first.

Worked Example 1: One Trial Was Compromised

A learner times how long identical toy cars take to travel down the same ramp. In Trial 4, the stopwatch is started noticeably after the car begins moving. The other trials use the intended timing rule.

READ GIVEN INFORMATION: only Trial 4 has a known timing error. IDENTIFY THE SCIENTIFIC JOB: compare travel times under the same procedure. SEPARATE OBSERVATION FROM INFERENCE: the recorded time exists, but there is evidence the timing procedure was not followed. DIAGNOSE THE LEVEL: one trial is compromised, not the whole design. REPAIR: repeat Trial 4 using the same start criterion.

This is different from deleting Trial 4 because its value is simply larger than the others. The reason for repeating must come from the method, not from dislike of the number.

Worked Example 2: The Evidence Set Is Too Thin

A fair comparison is made between two conditions using one result for each. The values differ, but there are no repeated trials and the quantity is known to vary from run to run.

The method may be reasonable, but one pair of results gives little information about variability. Repeating the complete comparison can show whether the same direction of difference appears again. The repair is evidence-level, not necessarily method-level.

Worked Example 3: Repeating Would Reproduce the Flaw

A learner wants to test whether surface type affects how far a block slides. The smooth-surface set-up uses a light block; the rough-surface set-up uses a much heavier block. Each condition is repeated ten times.

Ten repeats do not isolate surface type because block mass also differs. The design itself is the earliest weak link. The method must first make the comparison scientifically appropriate. Only then do repeats become useful evidence about consistency.

Worked Example 4: The Measurement Answers the Wrong Question

A question asks how quickly a process changes, but the method records only one final value and no timing information. Repeating that same endpoint measurement may give many final values, but it still cannot reconstruct the rate or timing of the process.

The problem is not the number of repeats. The problem is evidence fit. Redesign the measurement schedule so the evidence can answer the question.

Worked Example 5: A Result Looks Strange but the Method Was Followed

One repeated result differs from the others, but the learner finds no recording error, missed step, apparatus problem or obvious condition change.

Do not silently erase it. The result may reflect natural variation, an unseen method difference, limited measurement resolution or a real boundary in the pattern. Keep it visible, inspect the evidence, and decide whether additional repeats are useful. “Unexpected” is not the same as “invalid”.

The Earliest Weak-Link Diagnosis

What scientific question is the investigation trying to answer?
Was the comparison suitable for that question?
Was the correct quantity measured?
Was the intended method followed in every run?
Is the problem local to one trial or shared across all trials?
Is there enough repeated evidence to judge variability?
Would repeating the same method actually remove the weakness?

If the answer to Question 7 is “no”, redesign before repeating.

Common Failure Signatures

Writing “repeat for accuracy” for every method-evaluation question.
Repeating only the trial with the least convenient result.
Redesigning the whole method because of one known recording mistake.
Repeating an unfair comparison many times and calling the result more valid.
Adding repeats when the real problem is measuring the wrong outcome.
Changing the scientific question while supposedly “improving” the method.
Treating every difference between repeats as proof the method failed.

Misconception Repair

“More repeats fix every weakness.” No. Repeats can strengthen evidence from a suitable method, but they cannot turn the wrong comparison into the right one.

“A bad-looking result should be repeated.” Only if there is a scientific reason to think that run was compromised. An inconvenient value is still evidence.

“Redesign means change everything.” Good redesign is surgical. Change the part that breaks the scientific question while preserving what already works.

“If a method is fair, one trial is enough.” Fairness and evidence quantity are different questions. A fair method may still need repeats or more specimens when variability matters.

The PSLE Science Repair Chain

READ THE QUESTION → IDENTIFY THE INTENDED METHOD → LOCATE THE FIRST FAILURE → DECIDE ITS SCOPE → REPEAT ONE TRIAL / REPEAT THE EVIDENCE SET / REDESIGN THE METHOD → COLLECT NEW EVIDENCE → CHECK WHETHER THE ORIGINAL QUESTION IS NOW ANSWERED.

Original Practice: Choose the Repair Scope

Scenario A: One timing value was copied into the wrong row. The apparatus and all other trials were used correctly. What level failed?

Scenario B: A fair comparison uses only one specimen in each condition. The result may be affected by natural variation. What extra evidence would help?

Scenario C: The two set-ups use different containers, different starting quantities and different tested conditions. Would five more repeats solve the problem?

Scenario D: A process is meant to be compared by rate, but only one final measurement is taken. What needs redesigning?

Retrieval and Transfer Sequence

Round 1: classify six method problems as one-trial, evidence-set or design-level failures.
Round 2: justify the chosen repair in one sentence: “This fixes ___ because ___.”
Round 3: change the science context and repeat the diagnosis without relying on topic cues.
Round 4: return after several days and diagnose a fresh investigation without the decision table.

Delayed Independent Return Test

Give the learner a new investigation containing one local mistake, one evidence-volume weakness and one possible design flaw. Ask for the smallest scientifically sufficient repair. The learner has understood the job if they can distinguish what needs repeating from what needs redesigning and explain why.

Answer-Checking Receipt

I know the scientific question the method is meant to answer.
I can say whether the problem affects one run or every run.
I do not delete an anomalous result merely because it is unusual.
I know whether more evidence or a better design is needed.
I can explain why my proposed repair fixes the diagnosed weakness.
I have not changed the original scientific question by accident.
I can state what new evidence should look like after the repair.

Parent and Tutor Teaching Guide

When a learner gives “repeat the experiment” as an automatic answer, ask: “What problem does the repeat solve?” Do not supply the category immediately. Make the learner identify whether the problem is local, evidence-wide or structural.

A useful teaching contrast is to place three short investigations side by side: one with a single procedural error, one with a sound method but very few repeats, and one with an unfair comparison. Ask the learner to propose the smallest repair that genuinely changes the quality of the evidence. This teaches diagnosis before prescription.

Keep the language flexible. “Repeat one trial”, “collect more repeated evidence” and “redesign the method” are thinking categories, not compulsory examination phrases.

Useful Internal Routes

Authoritative References

Quiet Return

Good scientific repair is not “do more”. It is “fix the level that failed”. One compromised trial needs a local repair. Thin but sound evidence may need more repeats. A flawed comparison needs a better method. Once you learn to locate the failure before choosing the fix, investigation questions become much less mysterious—and your improvements become scientific rather than automatic.

Small Group Tutorials