PSLE Science Reality Lab Vol No.098 | “Tested in Triplicate” — Were There Three Independent Samples, or Three Readings of One Sample?

PSLE-SCI-REALITY-0098

Wait, What? Three Numbers Can Still Come From One Sample

A product report says:

Tested in triplicate.

That sounds reassuring. Three tests must be stronger than one, right?

Sometimes. But first you need to know what “three” actually counts.

Did the laboratory measure the same prepared sample three times? Did it prepare three separate test portions from one product? Did it test three different products from the same batch? Did it repeat the whole investigation on three separate days? Those designs can all produce three results, but they do not test exactly the same kind of uncertainty.

This Reality Lab owns one real-world transfer job: when a scientific report says “tested in triplicate”, work out what was repeated before treating the number three as three independent confirmations.

Quick Answer

Ask what object was repeated: the reading, the test portion, the specimen, the trial, the day or the laboratory.
Three readings of one sample mainly tell you about short-term measurement consistency.
Three independently prepared specimens can reveal variation that one specimen cannot.
Three trials are only genuinely separate if the set-up is reset appropriately and each trial can produce its own result.
Agreement among close repeats is useful evidence, but it does not automatically prove reproducibility across different specimens, operators, instruments or laboratories.
Keep the claim matched to the level that was actually repeated.

The Exact Learner Job This Page Owns

This page does not re-teach the whole topic of repetition. Existing PSLE Science pages already own the distinctions among a reading, a trial and a specimen, the choice between repeated trials and more specimens, and repeatability versus reproducibility.

This Reality Lab applies those skills to a public-facing phrase that often compresses all of that method detail into two words: tested in triplicate.

Original Reality Lab Case: The Fictional FlexTile Test

This teaching case is original. It does not reproduce a commercial product test, examination question or competitor material.

A company compares the water absorption of two fictional floor materials, FlexTile A and FlexTile B. Its infographic says:

Each material tested in triplicate.

The results shown are:

Material	Reading 1	Reading 2	Reading 3	Average
A	12.1 g	12.2 g	12.1 g	12.13 g
B	15.0 g	15.1 g	15.0 g	15.03 g

The numbers look very tidy. But now the method note appears: one piece of A was soaked once and weighed three times; one piece of B was soaked once and weighed three times.

That design gives useful information about whether the weighing step is consistent. It does not tell us how much different pieces of material vary, because only one piece of each material was tested. The three readings are not three independent specimens.

Observed, Claimed and Inferred

Layer	Statement
Observed	The same prepared A specimen was weighed three times and gave closely similar readings.
Observed	The same prepared B specimen was weighed three times and gave closely similar readings.
Supported inference	The weighing procedure was reasonably consistent over those repeated readings.
Stronger claim needing more evidence	Three independent A specimens and three independent B specimens all behaved this way.
Still broader claim	All FlexTile A and B products from other batches and conditions will show the same difference.

The First Question: What Does the “Three” Count?

A good learner never lets a count float without an object. “Three” is incomplete until you know whether it means:

three readings from one object;
three test portions taken from one specimen;
three separate specimens from one batch;
three trials after resetting the apparatus;
three different days;
three different operators;
three different instruments;
or three different laboratories.

Each version asks a different question about how stable the result is.

Why Three Readings of One Sample Are Still Useful

Do not make the opposite mistake and call repeated readings worthless. They can be valuable. If a balance gives 12.1 g, 14.8 g and 10.6 g for the same unchanged object within seconds, something about the measurement process deserves attention. If it gives 12.1 g, 12.2 g and 12.1 g, the short-term reading is more stable.

NIST describes repeatability as agreement among successive measurements of the same measurand under closely matched conditions. That can include the same procedure, observer, instrument, location and a short time interval. So repeated readings can test an important part of measurement quality.

But they answer that narrow question. They do not automatically reveal how much different samples vary.

Why Independent Specimens Add a Different Kind of Evidence

Suppose three different pieces of FlexTile A absorb 10 g, 12 g and 18 g of water. One piece behaves very differently from the others. Reweighing only the first piece three times would never discover that specimen-to-specimen variation.

Independent specimens help the investigation travel from “this object behaved this way” toward “objects of this kind showed this pattern under the tested conditions”. How far that conclusion can travel still depends on how the specimens were selected, whether they came from one batch, how many were tested, and what conditions were used.

Repeatability Is Not the Same as Reproducibility

If one operator, one instrument and one session produce closely matching results, that is encouraging. A result that also survives a change of operator, instrument, day or laboratory answers a broader question.

NIST distinguishes repeatability under closely similar conditions from reproducibility under changed conditions. The exact technical definitions become more advanced than Primary Science, but the practical learner habit is simple:

Agreement is strongest when you know what was allowed to change and what stayed the same.

The Representation Check: A Table Can Make Dependence Disappear

An infographic might show three neat dots for Product A and three neat dots for Product B. Visually, that looks like six separate experiments. But a graph does not automatically tell you whether the points came from six independent objects or six repeated readings of two objects.

The method note matters because the structure of the evidence is part of the result.

The Provenance Check: Follow One Result Backwards

Which specimen produced this result?
Was that specimen independently prepared?
Was the apparatus reset before the next result?
Were the measurements made at the same time or on different days?
Did the same operator and instrument produce all readings?
Was the same source material used for every specimen?

If the answers show that all three results share most of the same physical history, the repeats are more dependent than the phrase “three tests” may suggest.

Method Check: Did the Trial Really Restart?

A genuine repeat of a trial usually requires restoring the starting conditions that matter to the question. If a surface stays wet from Trial 1, Trial 2 may inherit a changed starting state. If a spring remains stretched, if a container has warmed, or if a sensor drifts, repeating the procedure without resetting the system may not create an equivalent new trial.

This is why “three measurements” and “three trials” should never be treated as automatic synonyms.

What Evidence Would Strengthen the Claim?

State clearly what was repeated.
Use independent specimens when natural specimen variation matters.
Reset relevant starting conditions between trials.
Preserve individual results instead of showing only the average.
Describe whether samples came from one batch or several.
Where broader reproducibility matters, repeat under appropriately changed conditions such as another day, operator, instrument or laboratory.
Keep the public claim no broader than the repetition structure supports.

What Would Weaken It?

The word “triplicate” appears but the method never explains what was triplicated.
Three readings from one object are presented as three independent samples.
The same prepared solution is divided after the key experimental step and treated as three separate experiments.
The second and third trials inherit changed starting conditions from the first.
Only the average is shown, hiding whether the three values actually agree.
A result from one batch is described as if it represented every batch.

Worked Case 1: Three Weighings

A sponge is soaked once and weighed three times: 18.2 g, 18.2 g and 18.3 g. This is useful evidence about the stability of the weighing step. It is not three independent sponge tests.

Worked Case 2: Three Sponges

Three similar sponges are each dry-weighed, soaked using the same stated method and reweighed independently. Their absorbed amounts are 7.8 g, 8.0 g and 8.5 g. Now the evidence includes variation among separate specimens as well as measurement variation.

Worked Case 3: Three Days

A sensor check is repeated on Monday, Wednesday and Friday using fresh reference material. Agreement across days can reveal a kind of stability that three readings made within one minute cannot.

Worked Case 4: Three Laboratories

Three laboratories test comparable material using a harmonised procedure. If they obtain compatible results, that provides broader evidence than one laboratory repeating its own reading three times. NIST interlaboratory comparisons exist precisely because measurement agreement across laboratories is a different question from within-laboratory repeatability.

Tempting Reasoning That Fails

“Three numbers means three samples.” Not unless the method says so.
“Three samples means three batches.” The three samples may come from one batch.
“Triplicate means the result has been reproduced.” Reproducibility usually involves changed conditions; technical repeats under the same conditions answer a narrower question.
“The average is based on three values, so the evidence is automatically strong.” Evidence strength depends on what those values represent and how independent they are.
“If repeats agree closely, the method must be accurate.” Several readings can agree closely while sharing the same systematic bias.

How Far Can the Conclusion Travel?

If one prepared sample was measured three times, the conclusion can travel confidently to the consistency of those repeated readings under those conditions. It cannot automatically travel to specimen-to-specimen variation, other batches, other laboratories or real-world use.

If several independent specimens and trials agree, the conclusion can travel farther. If different operators, instruments or laboratories also agree, it may travel farther again.

The important habit is not to memorise a hierarchy. It is to ask: what changed between the results, and therefore what kind of uncertainty did the repetition actually test?

PSLE-Style Transfer Case

A fictional advertisement says, “Our filter removed 92% of particles in triplicate testing.” The method note says one filter was used once, and the final water sample was measured three times by the same instrument.

Question: Why is it too strong to say the filter successfully removed 92% in three independent trials?

Reasoned answer: The three values came from repeated measurements of the same final sample, not from three separately run filter trials. They can show consistency of the measurement step, but they do not show that the whole filtering procedure produced the same result three independent times.

Explained Practice

Practice A: A thermometer measures the same cup three times without the cup changing. What is repeated? The reading.

Practice B: Three cups are prepared separately and each tested once. What new information appears? Variation among independently prepared specimens or set-ups.

Practice C: One operator repeats a test today and another operator repeats it next week. What broader question can this help examine? Whether the result remains compatible when conditions such as operator and time change.

Delayed Independent Return: The T-H-R-E-E Check

T — Thing: What exactly was repeated?
H — History: Did the repeats share the same specimen or preparation history?
R — Reset: Was the trial genuinely restarted?
E — Evidence level: Reading, specimen, trial, day or laboratory?
E — Extension: How far may the conclusion travel beyond those repeats?

Parent and Tutor Teaching Guide

Place three identical dots on paper and tell the learner, “These are three results.” Ask for at least four different experimental stories that could have produced them: three readings of one object, three separate objects, three days, or three laboratories. The dots stay the same; the evidential meaning changes.

Then give a short label such as “tested three times”. Ask the learner to write the one method sentence they would need before deciding what the phrase means. This trains method reconstruction rather than keyword memorisation.

Authoritative Sources

The official Singapore assessment frame requires learners to interpret and analyse information, evaluate observations, information and methods, and communicate explanations and reasoning. This page applies those inquiry habits to a real-world method phrase; it does not create a new examiner rule or mandatory answer format.

The Quiet Return

Three results can be useful.

But evidence does not become independent merely because the spreadsheet has three rows.

Whenever you read “tested in triplicate”, ask what was repeated before you decide what was confirmed.

Small Group Tutorials