PSLE-SCI-REALITY-0080
Wait, What? A batch can have an average above a minimum even when one of its items is below the minimum.
A product card says, “Tested strength: average 52 N. Required minimum: 50 N. Batch passed.” It sounds decisive. Then you inspect the five original readings: 61 N, 59 N, 54 N, 51 N and 35 N.
The average is exactly 52 N. Yet one item is far below 50 N.
The arithmetic is not wrong. The scientific question changed. An average of a group answers a different question from whether every individual item meets a minimum.
Reality Lab Vol No.080 teaches one real-world evidence-transfer job: when a batch, product comparison or quality infographic uses an average to claim that every item passed a limit, check whether the limit applies to the average, to individual items, or to both.
Quick Answer
- Identify the summary. Is the displayed number an individual reading or an average of several readings?
- Identify the rule. Does the stated limit apply to the batch average, each item, or another defined quantity?
- Inspect individual results when the claim is individual. A passing average cannot prove every item passed.
- Keep spread visible. Very different sets of readings can have the same average.
- Check how items were selected and tested. A correct average from a biased sample still cannot support a broad claim.
- Use bounded wording. “The sample average exceeded 50 N” is not automatically “all products exceed 50 N.”
Reality Lab habit: Before using a summary to judge a rule, make sure the summary and the rule live at the same evidence level.
The Owned Learner Job — and the Boundary
This page does not become the general owner of averages, variation, specification limits or quality control. Those scientific and mathematical ideas are broader than this Reality Lab. The owned job here is narrower: evaluating a real-world communication object that turns a passing group average into an every-item claim.
- How to Read an Average PSLE Science Result Without Treating It as Every Trial or Every Specimen
- How to Read the Same Whole-System Total When the Parts Can Be Distributed Differently in PSLE Science
- How to Tell Whether a PSLE Science Count Means Objects, Events or Measurements
- Reality Lab Vol No.067 | “Within Tolerance” — Does Passing a Limit Mean Hitting the Exact Target?
Vol No.067 distinguishes a tolerance band from an exact target. Vol No.080 asks a separate question: whose result is being compared with the limit—the group summary or each individual item?
Original Reality Lab Case: The Five Safety Straps
This is an original composite teaching case. It is not copied from a manufacturer, test standard or assessment book.
Five identical practice straps are tested with the same laboratory pull method. The simplified teaching specification says an individual strap should withstand at least 50 N in this test.
| Strap | Measured strength | Individual 50 N minimum met? |
|---|---|---|
| A | 61 N | Yes |
| B | 59 N | Yes |
| C | 54 N | Yes |
| D | 51 N | Yes |
| E | 35 N | No |
The total is 260 N, so the mean is 260 ÷ 5 = 52 N. The mean exceeds 50 N. But the individual-level statement “all five met the 50 N minimum” is false because Strap E measured 35 N.
Observed, Claimed and Inferred
| Layer | What it says |
|---|---|
| Observed | Five individual readings were 61, 59, 54, 51 and 35 N. |
| Calculated summary | The average was 52 N. |
| Claimed | Every tested item passed a 50 N individual minimum. |
| Inferred | A group average above the limit is being treated as proof that no individual result fell below it. |
The first two statements can be perfectly correct while the third remains unsupported.
The Average Is a New Number, Not a New Specimen
An average is calculated from a set of readings. It is a useful summary, but it is not an additional strap, leaf, battery, cup or plant that was measured. It does not replace the individual results from which it was calculated.
This matters because a summary compresses information. Once five readings become one average, differences among the five become harder to see.
Two Very Different Batches Can Have the Same Average
Consider these two original sets:
| Batch | Five readings | Average |
|---|---|---|
| Batch P | 52, 52, 52, 52, 52 | 52 |
| Batch Q | 61, 59, 54, 51, 35 | 52 |
Both averages equal 52. Yet Batch P has no reading below 50, while Batch Q has one. The same mean can sit above very different patterns of variation.
Therefore the average alone cannot answer every question about the members of the group.
Comparison Check: What Does the Limit Actually Apply To?
This is the central question. A rule might legitimately say any of the following:
- The average of a defined sample must be at least a stated value.
- Every individual item must be at least a stated value.
- The average must meet one condition and individual items must stay within another permitted range.
- A certain proportion of items may meet a classification rule.
Those are different decision rules. Do not invent one from another. In real standards, the exact rule belongs to the relevant authority or test specification. A Primary learner’s job is not to guess industrial rules. It is to read the stated rule and match it to the evidence.
Representation Check: The Giant “52”
Imagine a packaging graphic that prints “AVERAGE TEST RESULT: 52 N” in huge type while the five individual readings appear in tiny text or are omitted completely. The large number is real, but the design can encourage a reader to forget the variation behind it.
A stronger representation would show the average and the individual readings when the claim concerns individual performance. A dot plot, table or minimum/maximum summary can reveal what one mean hides.
Method and Variable Check
Even if every listed item passes, the scientific claim still depends on how the test was conducted.
- Were all items tested with the same method?
- Were the starting conditions comparable?
- Was the measuring instrument suitable?
- Were the items selected fairly from the batch?
- Was one item tested repeatedly while others were not?
- Was a damaged or failed item excluded after seeing the result?
An honest average from an unfair or selectively reported test remains weak evidence.
Source and Provenance Check
When a public claim says “average tested strength,” ask what was averaged. Five independent items? Five repeated readings from one item? Five batches? Averages are only interpretable when the observation unit is clear.
This is why provenance matters. The chain should be recoverable: item → measurement → set of measurements → calculated average → claim.
Worked Case 1: Average Temperature
Five containers finish an experiment at 22°C, 22°C, 22°C, 22°C and 32°C. The average is 24°C. Can you say every container finished at 24°C?
No. The average summarises the set; it is not the value of every member.
Worked Case 2: Average Above a Minimum
Four original material samples have strengths of 48, 51, 55 and 66 N. Their average is 55 N. A rule says every sample must be at least 50 N.
The average passes, but the set fails the stated every-sample rule because 48 N is below the minimum.
Worked Case 3: When an Average Rule Really Is the Rule
A fictional classroom protocol says, “The average of five repeated temperature readings must lie between 19°C and 21°C.” The five readings average 20°C.
For that stated rule, comparing the average with the range is appropriate. You should not silently replace it with an invented requirement that every reading must lie between 19°C and 21°C unless the protocol also says so.
Worked Case 4: One Object Measured Five Times
A student measures one block five times and averages the readings. The infographic then says, “Five blocks were tested.” That is wrong for a different reason. Five measurements of one block are not five independent blocks. Before interpreting an average, identify what was repeated.
Worked Case 5: The Average Improves While the Worst Item Gets Worse
Version 1 readings are 50, 50, 50, 50 and 50, average 50. Version 2 readings are 40, 55, 55, 55 and 55, average 52. Version 2 has a higher average but a lower worst-case reading.
Which version is “better” depends on the scientific and engineering question. A higher average is not identical to more consistent individual performance.
Alternative Explanations to Keep Alive
- The group may contain one unusually low or high item.
- Natural specimen variation may be large.
- A measurement error may have affected one reading.
- Items may come from different batches or production conditions.
- The average may be based on repeated measurements of the same item rather than multiple items.
- The stated threshold may apply to a different quantity than the displayed average.
The right response is not to declare the average useless. It is to ask what question that average can actually answer.
What Evidence Would Strengthen “Every Tested Item Passed”?
- The individual readings for every tested item.
- A clearly stated individual-item threshold.
- Evidence that every listed reading meets that threshold.
- A consistent test method and comparable conditions.
- A transparent account of exclusions or failed measurements.
- A suitable sample if the claim is extended beyond the tested items.
What Evidence Would Strengthen “The Batch Average Passed”?
- A rule that genuinely applies to a defined sample average.
- A clear sample size and sampling method.
- The individual readings used to calculate the average.
- A correct calculation.
- Enough information to see whether extreme values are being hidden.
What Would Weaken the Claim?
- Only the average is shown even though the wording says “every item.”
- The number of items averaged is missing.
- Failed items were removed from the calculation without a justified rule.
- Repeated readings from one item are described as many tested items.
- The threshold belongs to a different measurement or condition.
- The sample was chosen because it performed well.
How Far Can the Conclusion Travel?
If five tested items have an average of 52 N, that statement is about those five measurements under the test. To claim that a whole production batch has the same average requires a sampling design that represents the batch. To claim every product sold meets 50 N requires evidence and a decision rule appropriate to individual conformity.
Each step outward needs additional support. The average does not grant unlimited travel.
Model and Measurement Limits
An average is one way to describe the centre of a set. It does not show the full spread, shape or reliability of the results. Real quality decisions can use detailed standards, sampling plans, uncertainty, tolerance rules and process capability methods that go far beyond Primary Science.
For Primary 5/6, the useful transfer is simpler: never let a group summary pretend to be every group member.
PSLE-Style Transfer Case
An original investigation tests the mass supported by four model bridges before breaking. The readings are 12 N, 12 N, 12 N and 4 N. The average is 10 N. A student writes, “Every bridge supported at least 10 N because the average was 10 N.”
Evaluation: The conclusion is false. The average is a calculated summary of all four bridge readings. One bridge supported only 4 N. To determine whether every bridge supported at least 10 N, inspect each individual reading.
Tempting Reasoning That Fails
- “The average is above the limit, so all values must be above it.” A high value can balance a low one.
- “The average represents a typical item, so every item is close to it.” The spread may be wide.
- “One low result should be ignored because the average still passes.” Whether exclusion is justified depends on evidence about the measurement and the stated rule.
- “If one item fails, the average is useless.” The average can still answer a group-level question.
- “A higher average always means a better product.” The relevant outcome, variation, threshold and use all matter.
Explained Practice
Practice A: Results are 8, 8, 8 and 16. The average is 10. Can you say every result was 10? No.
Practice B: Results are 49, 51 and 60. The mean exceeds 50. A rule says each result must be at least 50. Does the set meet the rule? No, because 49 is below it.
Practice C: A rule explicitly says “the average of three readings must exceed 20.” The readings are 18, 21 and 24, average 21. Does the average meet that rule? Yes. Do not invent a different every-reading rule.
Delayed Independent Return: The L-E-V-E-L Check
- L — Level: Is the claim about an individual, a sample or a whole batch?
- E — Evidence: What individual readings exist?
- V — Variation: How different are the readings?
- E — Exact rule: What quantity does the limit apply to?
- L — Limit the conclusion: State only what that evidence level supports.
Try the check later with a different dataset. If you stop treating a mean as a magic replacement for every reading, the transfer is working.
Parent and Tutor Teaching Guide
Write five numbers on cards: 61, 59, 54, 51 and 35. Ask the learner to calculate the average. Then place a line labelled “50 minimum” on the table and ask, “Does the average pass?” The learner should say yes. Next ask, “Does every card pass?” The learner should say no.
Now replace the cards with 52, 52, 52, 52 and 52. The average stays the same, but the individual pattern changes. This makes the information-loss problem visible without advanced statistics.
Finish by changing the rule: “The class requirement is that the average must be at least 50.” Ask whether the first set now meets the stated average rule. This prevents the learner from replacing one overgeneralisation with another. Scientific reasoning depends on matching the evidence to the exact question.
Authoritative Sources
- Singapore Examinations and Assessment Board — 2026 PSLE Science Syllabus
- Ministry of Education Singapore — 2023 Primary Science Teaching and Learning Syllabus
- NIST/SEMATECH e-Handbook — Process Variability
- NIST/SEMATECH e-Handbook — Assessing Process Capability
NIST’s process-quality guidance treats variation and specification limits as distinct from the process average: a process is capable when its output distribution fits the required specifications, not merely because the centre looks acceptable. That industrial framework is far beyond what a Primary learner needs to calculate, but it reinforces the same evidence habit. SEAB’s current PSLE Science objectives ask learners to interpret information, evaluate observations and methods, and communicate reasoning. The first step is knowing whether a number describes one item or a group.
The Quiet Return
Averages are powerful because they turn many readings into one useful summary.
That power is also the danger. Once many readings become one number, it is easy to forget that the individuals still exist.
So when an average passes a limit, ask one more question before you celebrate: which level of the evidence was the limit written for?