EDUKATE SINGAPORE · GOLD STANDARD PRACTICAL CASE STUDY 163
Sixty genuine measurements are not sixty independent experiments.
A school science trial measures sixty seedlings but assigns soil treatments to only six trays. This practical case explains why the unit of assignment matters, calculates the genuine descriptive result and shows how a teacher can correct the claim without discarding the work.
All plant data, learners and treatments here are fictional. Research-method sources inform the scientific principles, not any real school experiment.
At a glance
- Recorded: 60 plant heights.
- Independently assigned: six trays, three per method.
- Observed mean difference: 3 cm between tray averages.
- Not proved: sixty independent treatment effects or a causal gain.
- Next action: preserve data, correct the reporting unit and review the study design.
Chapter guide
- 1. Sixty measured plants—but treatment was allocated just six times
- 2. The experimental unit comes before the spreadsheet denominator
- 3. The worked three-centimetre difference is descriptive, not conclusive
- 4. Published research calls this problem pseudoreplication
- 5. Why adding more seedlings may not add independent replication
- 6. An evidence correction should not erase the students’ work
- 7. Teach the distinction with a tray diagram
- 8. Plan genuine independent comparisons next time
- 9. Two reporting counts are better than one green sample-size field
- 10. Practice: what does the experiment really establish?
1. Sixty measured plants—but treatment was allocated just six times
A fictional secondary-school science club tests two soil mixtures. Thirty seedlings are grown under Treatment A and thirty under Treatment B. Students measure every plant carefully and proudly report sixty independent experiments. The measurements are real, but the wording conceals the design.
Each treatment was applied to three trays, with ten seedlings sharing one tray. A tray—not every seedling—received an independent treatment assignment. All sixty measured heights matter, yet the six tray assignments are the independent units for assessing this soil treatment under the fictional design.
The Gold Standard problem is the difference between recording observations and independently repeating an intervention. One spreadsheet row is not automatically one replicate. The teacher should preserve the data while correcting the scientific conclusion.
2. The experimental unit comes before the spreadsheet denominator
The experimental unit is the smallest unit independently allocated the experimental treatment for the chosen research question. Here, that unit is the tray. Seedlings inside a tray share soil, watering conditions and parts of the growing environment. They may therefore be more alike than independently grown plants.
Six trays received assignments: three Treatment A and three Treatment B. Sixty seedling observations were nested inside those six units. Both counts are accurate descriptions, but they answer different questions. Describing sixty measurements as sixty separate treatment assignments overstates how much independent information was collected.
The experimental unit could differ in another study where plants are independently treated. The proper definition follows the allocation and intended inference, not a rule that trays always matter more than seedlings.
3. The worked three-centimetre difference is descriptive, not conclusive
Suppose the three Treatment A tray means are 14, 20 and 26 cm. Their mean is (14 + 20 + 26)/3 = 20 cm. The three Treatment B tray means are 16, 17 and 18 cm, averaging 17 cm. The measured difference between the average tray means is 3 cm.
The arithmetic is correct. But the wide variation among Treatment A trays matters to any proposed inference. No statistical significance or general causal conclusion follows from six tray means alone without a suitable study design and analysis. A 3 cm difference is worth reporting as an observation under defined conditions.
The class can improve the report by displaying each tray’s mean rather than concealing tray differences inside a cloud of sixty plant points. A clear graph supports honest reasoning better than a dramatic but unjustified p-value.
4. Published research calls this problem pseudoreplication
The article Empowering Statistical Methods for Cellular and Molecular Biologists explains why repeated technical observations are not necessarily independent biological replicates. The 2025 experimental-design review describes pseudoreplication as counting dependent measurements as though they were independent units for statistical inference.
An additional review in physiology shows how ignoring nested observations can lead to unreliable certainty estimates. These are general research-method references, not assessments of the invented school trial.
The useful rule is to identify what actually received treatment independently. Multiple accurate observations from one shared unit can improve its description without independently replicating the treatment.
5. Why adding more seedlings may not add independent replication
Suppose the science club doubles its planting to twenty seedlings per tray while keeping three trays per treatment. The number of measured plants rises to 120, yet the independent treatment assignments remain six. The new data can improve some within-tray estimates without doubling the number of independently treated trays.
If one tray receives unusually strong light or a watering error, many plants may respond together. Treating those plants as unrelated experiments risks mistaking one shared condition for repeated confirmation of the soil treatment.
A stronger design might allocate treatments independently to more suitable trays under competent planning and consider relevant variation between units. There is no universally sufficient number of trays in this article. A teacher should define the question, resources and intended inference before selecting a replication plan.
6. An evidence correction should not erase the students’ work
The school can retain every measurement, original worksheet, calculation and earlier presentation. The revised result should identify sixty observed plants, six independently treated trays and the descriptive three-centimetre difference without claiming proven superiority.
Students should not be labelled dishonest because they misunderstood statistical independence. They can learn to trace the evidence from allocation through measurement, calculation and conclusion. An honest correction preserves what the experiment showed and identifies what remains uncertain.
GS-R08 requires evidence that matches the claim; GS-R09 requires an appropriate exception response. A report that explains its limits teaches more scientific responsibility than one that simply hides the first mistaken slide.
7. Teach the distinction with a tray diagram
The tutor draws six large boxes for trays and ten small marks inside each box. Labels A and B go on the large boxes, because that is where soil treatments were allocated. The small marks represent seedling measurements. The diagram makes an invisible dependency visible.
Students can then calculate tray means and compare between-tray variation before trying to interpret a treatment effect. They learn to use terms such as observations, units, groups, independent assignment, mean and inference accurately.
A capable researcher may use hierarchical or mixed-effects analysis rather than only simple tray averages. The Gold Standard is not a universal command to discard nested information; it is the requirement that analysis respects how observations are related.
8. Plan genuine independent comparisons next time
Before a new project, the class writes its scientific question, names the unit receiving each treatment, considers suitable random allocation and identifies the measurements needed. It also records which plants belong together, who will measure them and what will happen if some seedlings fail to grow.
Rulers, spreadsheets and neatly printed labels improve work quality, but they do not repair a missing design decision. The correct experimental unit is an early part of readiness, not a detail chosen after seeing which group looks better.
GS-R03 assigns review authority, GS-R04 prepares learners to reason correctly and GS-R05 concerns realistic resources. Clear teaching helps students understand the structure behind the mathematics of their experiment.
9. Two reporting counts are better than one green sample-size field
The club’s original reporting form has a single box for ‘sample size’, filled with 60. The improved form distinguishes independently assigned trays per treatment and seedling measurements within each tray. It also records the actual analysis and limits of interpretation.
The same mistake can arise with many pupils nested in a few classrooms, several images from one specimen or repeated readings from one person. The independent unit depends on the scientific question, not the number of file rows.
An accurate dashboard can celebrate the sixty careful measurements and still say that six treatment allocations constrain the general claim. Those statements are compatible, useful and more trustworthy than one unsupported green statistic.
10. Practice: what does the experiment really establish?
Three Treatment A trays have means of 14, 20 and 26 cm; three Treatment B trays have means of 16, 17 and 18 cm. Each tray contains ten seedlings, giving sixty measured plants. How many independent soil-treatment assignments exist, and what difference can be described?
There are six independently assigned trays, three in each arm, with sixty seedling observations nested within them. A’s mean tray height is 20 cm, B’s is 17 cm, and the descriptive difference is 3 cm. The result does not establish statistical significance or causal improvement without an appropriate analysis.
The next acceptance test should require students to identify the unit of treatment before reporting n, show the true level of observation and state a bounded conclusion. The Gold Standard lesson is that careful measurement alone cannot create independent replication.
Questions parents, students and teachers ask
Are all sixty measurements still useful?
Yes. They describe seedlings and may help estimate within-tray variation. They should be analysed in a way that respects shared tray conditions.
Does this study have six data points or sixty?
It contains sixty plant-height observations nested in six independently treated trays. The scientifically useful answer identifies both levels.
Would more seedlings inside the same trays fix pseudoreplication?
No. More within-tray measurements may improve precision for those trays without creating new independent treatment assignments.
Does the three-centimetre difference prove Treatment A is better?
No. The difference is descriptive under the given data. Statistical uncertainty and the experimental design need suitable review before any general causal claim.
How can a Secondary 1 learner remember the rule?
Ask what received the treatment independently. Then count measurements within each unit separately. The question is more important than memorising a particular statistical formula.
What the next reviewer should verify
- The scientific question and the unit independently assigned treatment are stated before analysis.
- The data preserve tray identity and the individual seedling observations.
- Any graph distinguishes within-tray measurements from independent comparisons.
- The numerical calculation is checked against the underlying records.
- The final publication states uncertainty and does not invent statistical significance.
- The next design is approved using the actual resources and suitable competence.
A reviewer can ask students to explain which environmental conditions were shared within a tray. If ten seedlings grew under the same watering mistake, those ten observations do not represent ten unrelated treatment assignments. The student who recognises the grouping has understood a meaningful part of experimental design, even before learning formal inference.
The reporting form should retain the old version as part of the learning record and clearly label the corrected interpretation. Nothing in this fictional example demonstrates dishonesty or warrants a punitive conclusion. Its useful outcome is that the class learns to distinguish a descriptive difference from a reliable generalisation.
Use an evidence ledger that separates every level
A practical evidence ledger begins with a tray identifier, a treatment identifier, an allocation date and a reasoned description of the growing environment. Each of the ten seedlings within that tray then has its own measured height and observation time. In the fictional experiment, there would be six tray identifiers and sixty child measurement records, not sixty separate treatment allocations. This structure lets the student trace every number back to the condition under which it was produced. It also allows a reviewer to notice if a seedling has been entered twice, if a tray label was lost, or if a late measurement used a different observation window. An accurate record is not just a neat table; it preserves the connections needed to interpret the observations.
When the data are exported to a spreadsheet or graphing application, the tray field must travel with each row. A chart drawn from only the height column may look impressive while concealing which plants shared a soil container. Labels can make the dependency visible without overloading the class with advanced statistical software. This is the same evidence principle used in the Gold Standard of Information and Record Integrity: the meaning of a value should survive processing, presentation and review. If the unit identifier vanishes between collection and publication, a reliable inference can become much harder, even if the heights themselves remain unchanged.
A deeper look at the six tray-level numbers
The worked example contains three Treatment A means of 14, 20 and 26 centimetres, and three Treatment B means of 16, 17 and 18. The A mean is 20 and the B mean 17. The 3 cm contrast is a real difference between averages in the invented observations. Yet Treatment A’s tray means range from 14 to 26, a 12 cm span, while Treatment B’s tray means range from 16 to 18, a 2 cm span. That difference in variability is a reason to investigate the study and the context, not a licence to cherry-pick the tallest tray. Perhaps a shared light or watering condition mattered; perhaps the treatment affected variability; perhaps random differences were influential. The six figures alone cannot decide between those possibilities.
Students often think that a large number of measurements guarantees strong evidence. That is understandable: a spreadsheet containing sixty rows feels more substantial than one with six. But the confidence of an inference depends on whether the observations supply genuinely independent information about the relevant variation. Ten plants sharing one tray may all shift together when the tray moves nearer the window. Counting them as ten unrelated treatment trials could make that shared movement seem like repeated confirmation. A strong scientific explanation therefore starts with the design, describes relevant uncertainties, and resists transforming an attractive average into a universal claim.
A classroom sequence that teaches the idea from first principles
A tutor can begin without formal statistical terminology. Place six labelled cards on a table, three marked A and three marked B. Put ten small markers on each card. Ask students where the soil mixture was chosen and applied. They will point to the cards, not the individual markers. Next, ask how many heights were measured. The answer is sixty. The two answers appear to conflict only if the learner assumes that ‘measurement’ and ‘independent experiment’ are the same thing. The teacher can then introduce the language of experimental unit and nested observation. This follows the direct worked-example, guided-practice and error-correction approach described in eduKate’s Secondary 1 Mathematics tutorials reference without claiming that the fictional science trial occurred there.
Next, ask learners to reconstruct the two treatment means from the tray means and explain the limit of their conclusion in one sentence. Students who only memorise a formula may calculate 20 minus 17 correctly yet overlook the fact that the treatments were assigned at tray level. A stronger answer names both the calculation and the boundary of the evidence. The teacher can model a fair statement, invite students to critique an overconfident alternative, and repeat the exercise with new numbers. The goal is not to make Secondary 1 students perform complex biological statistics. It is to build the habit of asking whether a mathematical answer corresponds to the real-world structure of the problem.
What a genuine follow-up experiment would need
A better study should define its research question and determine where treatments can be allocated independently before any planting begins. If the relevant unit remains a tray, suitable replication may require additional independently prepared and assigned trays, while keeping other conditions as comparable as feasible. A reasonable plan also identifies what counts as an unsuccessful seedling, how missing measurements will be handled and what source data the analysis requires. No fixed number of trays guarantees a trustworthy conclusion in every experiment. The right design depends on expected variability, acceptable precision, research purpose and resources, and may warrant input from a teacher or statistician with suitable competence.
Changing the design after results are known creates a separate governance concern. If the group sees that Treatment A looks better and then changes which observations are counted as independent, the resulting analysis may reflect the desired conclusion rather than an accepted research question. The Gold Standard’s controlled-change requirement encourages keeping the original design, explaining subsequent decisions and separating exploratory analysis from a genuinely planned test. An exploratory pattern can be educationally important. It simply deserves the correct label, just as a preliminary weather forecast and a final observed measurement deserve different labels.
A practical review exercise and model decision record
Before presenting the revised science poster, the class should exchange its evidence with another group. The reviewing pupils receive the tray allocation plan, the individual height records and the proposed headline. Their task is to identify the unit of independent assignment, recompute the two descriptive means, and underline any sentence implying more certainty than the evidence warrants. They should also identify one plausible shared-tray factor whose existence cannot be ruled out merely from the recorded heights. This tests understanding of reasoning, not a student’s ability to click the statistical software’s significance-test menu.
A short model decision record would read: ‘Our class observed sixty seedlings in six trays. Three trays per soil-treatment group were independently assigned under the experiment design. The treatment-group means, calculated from the reported tray values, were 20 cm and 17 cm. The observed difference was 3 cm in this study. We have not established a general causal advantage or justified a significance claim from these figures alone. Further suitable independent replication and analysis would be needed.’ It identifies purpose, source counts, method, result, uncertainty and next step without implying that every learning experiment needs expensive equipment.
The reviewer should also distinguish a procedural limitation from a personal failing. A teacher may accurately teach arithmetic and still need to refine the experimental design. A pupil may make a thoughtful observation and then misuse the word ‘independent’. Correcting those mistakes is part of building scientific competence. A system focused solely on producing the most impressive experiment result may discourage questions about uncertainty; a Gold Standard educational system rewards the ability to notice and explain the limits of evidence before other people rely on it.
References and connected Gold Standard case studies
- Experimental Replication and Design (2025)
- Empowering Statistical Methods for Biologists
- Pseudoreplication in Physiology
- Gold Standard Core Requirements
- Gold Standard Evidence and Verification
- Case 122 · Research Assessor Blinding
- Case 118 · Microplastics Sampling Controls
- Gold Standard Casebook
GS-R01–GS-R12 is an eduKate educational management reference, not an external laboratory accreditation. Appropriate experimental units and analysis depend on the actual hypothesis and competent scientific review.