eduKate Singapore · Primary 6 · Learner’s Guide Vol 0083
Choose the part that helps you now
1. Understand the job · 2. Study the worked examples · 3. Diagnose and repair · 4. Practise independently · 5. Check, control and continue
Full chapter index · Practice · Subject hub · PSLE Learning Guide
Chapter index
Chapters 1–4
1. What a repeat adds to the evidence
2. Identify the measured quantity before comparing readings
Chapters 5–8
5. Worked dataset: similar averages, very different spread
6. Treat an unusual reading as a question, not rubbish
Chapters 9–12
9. Practice progression: identify, compare, conclude
10. Longer practice: a drying investigation
Repeated Science results help you decide how much confidence to place in a conclusion. They do not automatically make an experiment fair, prove that a measuring instrument is correct, or justify a claim about every possible situation.
When you read a results table, first identify what was measured and how the trials were carried out. Then compare all the relevant results, look for variation and decide what the pattern actually supports. Your answer should be as strong as the evidence allows, with the tested conditions kept attached.
This Primary 6 workshop focuses on judging conclusions from repeated results. You will repair overconfident answers, compare sets of readings, distinguish a repeat from a changed investigation, and practise explaining what extra evidence would resolve a doubt. The datasets are invented for learning; they are not reports of real experiments or predictions of PSLE questions.
Suppose a pupil releases the same toy car from the same point on a ramp and measures how far it travels along the floor. The recorded distances are 78 cm, 80 cm and 79 cm. The readings are close, though not identical.
The small variation could come from details of releasing the car, identifying its stopping point or measuring the distance. You cannot identify the exact source from the numbers alone. What you can say is that these three recorded trials produced similar distances under the stated conditions.
Now suppose the readings are 48 cm, 81 cm and 79 cm. One is much lower than the other two. You should notice it and investigate. Do not simply select 81 cm because it looks impressive or ignore 48 cm because it is inconvenient.
A repeat provides another opportunity to observe the result of the method. If similar outcomes occur when the same procedure is repeated, the evidence is less dependent on a single unusual trial. This is the basic reason repeated trials can make results more dependable.
But repetition does not explain itself. You still need to know whether the starting conditions, measured quantity and procedure were comparable. Releasing a car from three different heights produces three conditions, not three repeats of the same condition.
Nor does repetition guarantee truth. If a ruler is misread the same way every time, the results may be close and still wrong. Similar readings show consistency in what was recorded. They do not by themselves establish that the recorded values match the true distances.
A first question to ask
Before writing “repeat the experiment”, ask what the repeat is meant to check. Is the concern that one trial may have been unusual? Is a measuring step inconsistent? Or is the original comparison unfair? The proposed action should address the actual weakness.
For the established Primary 6 repeat-trial learning route, use Repeat Trials: Reliability Before Conclusion. For the wider concept, use the existing repeatability guide. This volume concentrates on the examination-performance job of reading repeated-result evidence and choosing a defensible conclusion.
A number in a table needs a meaning. Forty may be a temperature, a time, a mass or a distance. The unit and heading tell you what the result represents. Before comparing trials, state the measured outcome in ordinary words.
If a table records the time taken for a sugar sample to dissolve, a lower time means faster dissolving under the stated conditions. If it records the amount dissolved after a fixed time, a larger amount means more dissolved during that period. The numerical direction changes with the measured quantity.
This matters when you judge a conclusion such as “Setup A worked better”. Better at what? Dissolving sooner, keeping water warm longer, allowing more light through, or supporting a heavier load? Replace the vague word with the actual outcome.
Consider two invented sets of readings for the time taken to dissolve an equal amount of sugar. In stirred water, the times are 34 s, 36 s and 35 s. In unstirred water, they are 70 s, 74 s and 72 s. The stirred trials have consistently shorter recorded times.
A cautious conclusion is: “Under the stated conditions, stirring reduced the time taken for the sugar to dissolve.” Do not write “stirring made more sugar dissolve” unless the investigation measured the amount dissolved and supplied evidence for that different outcome.
Be equally careful with temperature. If a cup’s water is 45°C after ten minutes, that is a final temperature, not automatically the amount of temperature decrease. You need the starting temperature to find the decrease. A repeated set of final readings does not repair a missing starting condition.
Name the result before evaluating it: “time until fully dissolved”, “distance travelled after release”, or “water temperature after ten minutes”. This small label helps prevent a conclusion about a nearby but unmeasured feature.
A repair example
Draft: “Material X is strongest because it has the biggest number.” Suppose the numbers are seconds until a strip breaks under a fixed load. The measured result is survival time under that load. A supported answer concerns lasting longer in that test, not a universal ranking of every kind of strength.
If the question supplies a definition of “strongest” for that investigation, follow it. If it does not, keep your conclusion tied to the measured outcome rather than adding an unrestricted claim.
Contents · Previous chapter · Next chapter
A repeated trial attempts the same investigation condition again. A different condition changes something deliberately so that you can compare its effect. You need both kinds of information in many investigations, but they perform different jobs.
Imagine testing how ramp height affects the distance a car travels. You use heights of 10 cm, 20 cm and 30 cm. A single run at each height gives three results across three conditions. It does not give three repeated results for any one height.
If you run the car three times at 10 cm, three times at 20 cm and three times at 30 cm, each height has its own set of repeats. You can inspect consistency within a height before comparing the pattern across heights.
Do not average the results from all three heights into one number and claim that it describes the effect of height. That would combine different conditions and hide the relationship you wanted to investigate.
A table should make the grouping visible. The condition belongs beside its repeated readings. When reading a table, follow one condition across its trial columns, then compare that set with the next condition. Do not compare arbitrary cells without their labels.
The same idea applies when pupils test different materials. Three different materials with one reading each are not three repeats of one material. Three strips of the same named material tested using the same procedure may provide repeats for that material, subject to how comparable the strips are.
Biological investigations may involve naturally varying individuals. Three similar seedlings are not literally the same organism tested again, but they can provide repeated observations for a condition if the design uses comparable groups. State what the task gives you rather than assuming perfect sameness.
Test your understanding
A pupil measures the temperature of three cups after ten minutes. One cup is uncovered, one has a cardboard lid and one has a plastic lid. Are these three repeats? No. They are three different covering conditions.
What would a clearer repeated comparison look like? Several comparable trials for each covering condition, with relevant starting quantities and temperatures controlled. The repeat addresses dependence on one trial; the controlled conditions make the comparison meaningful.
Use the PSLE Science guide library for deeper work on variables and fair comparisons. Here, your audit begins by deciding whether the table contains repeats, changed conditions, or both.
Contents · Previous chapter · Next chapter
Original investigation: “A pupil compared two coverings for identical cups. Each cup contained the same volume of water at the same starting temperature. The cups stood in the same room for ten minutes. The pupil repeated the comparison three times using the stated procedure.”
The temperature decreases for Covering A were 12°C, 11°C and 12°C. The decreases for Covering B were 7°C, 8°C and 7°C. These are teaching data, not a claim about particular real covering materials.
First identify the measured result: temperature decrease over ten minutes. A smaller decrease means the water remained closer to its starting temperature. Under these conditions, B’s decreases were smaller in every recorded trial.
A defensible conclusion is: “Covering B reduced the water’s temperature decrease more than Covering A in these trials.” Another clear version is: “The water with Covering B had a smaller temperature decrease over the ten-minute test.” To explain heat transfer, use scientifically accurate wording about slower heat loss rather than treating temperature itself as a substance that flows.
If asked why the repeats strengthen the conclusion, refer to consistency across the trials. Each repeated comparison showed a smaller decrease with B. The conclusion is not based only on one reading that might have been unusual.
Avoid “B completely prevented heat loss”. Its recorded decreases were 7°C or 8°C, so the temperature still fell. Avoid “B is always the best covering”. Only two coverings and a stated set of conditions were tested.
The averages, if requested, would be about 11.7°C for A and 7.3°C for B. But an average is not necessary to see that every B reading is lower than every A reading in this dataset. Do not add calculation merely because the table contains numbers.
Diagnose three draft answers
Draft 1: “B is better because 8 is less than 11.” This selects one reading from each set and uses “better” vaguely. Repair it by naming temperature decrease and using the repeated pattern.
Draft 2: “The results are reliable because there are three trials.” The number of trials alone does not describe their agreement. Repair: “The three trials gave similar decreases for each covering and the same comparison between the coverings.”
Draft 3: “The repeats make the experiment fair.” Repeats and fairness address different concerns. The equal starting volume, temperature, cups and exposure help the comparison be fair. The repeated results provide evidence about consistency.
Contents · Previous chapter · Next chapter
Now examine a different original investigation. A pupil measures the distance travelled by a toy car using two release procedures. Procedure P produces 59 cm, 60 cm and 61 cm. Procedure Q produces 40 cm, 60 cm and 80 cm.
Both sets have an average of 60 cm. Yet they do not tell the same story about consistency. P’s readings are close together. Q’s readings vary widely.
If the question asks which procedure produced more consistent recorded distances, P is supported by the smaller variation. If it asks which had the greater average distance, neither did; both averages are equal. A correct answer depends on the comparison requested.
Do not write that P must be more accurate. Accuracy concerns closeness to the true or accepted value, and that value has not been supplied. Consistency is visible from the repeated readings; accuracy needs additional evidence.
Do not write that Q’s average is meaningless. An average can summarise a set, but it does not show the whole distribution. The problem is treating the average as if it also proves that individual results are close together.
At Primary 6 level, you can explain variation using the actual readings without introducing advanced statistical language. “P ranged from 59 cm to 61 cm, while Q ranged from 40 cm to 80 cm” makes the contrast concrete.
The range is the largest reading minus the smallest. P’s range is 2 cm; Q’s is 40 cm. Use it when the task calls for that calculation or when it helps your explanation, but do not claim that every Science question requires a range.
Repair the average-only answer
Draft: “The two procedures are equally reliable because their average distances are both 60 cm.” This confuses average outcome with repeated consistency. A repair is: “They had the same average distance, but P produced more consistent readings because its three results were much closer together.”
Notice how the repaired answer preserves the true part of the draft. You do not need to reject the average calculation. You need to stop it from supporting a claim it cannot establish.
This is a general performance skill: keep each piece of evidence attached to the question it can answer. A number can be valid and still be insufficient for a broader conclusion.
Contents · Previous chapter · Next chapter
An unusual reading deserves attention. It does not deserve automatic deletion. If you remove a value simply because it weakens your preferred conclusion, you are changing the evidence to fit the answer.
Suppose repeated travel distances are 79 cm, 80 cm and 42 cm. The 42 cm result differs substantially. Possible explanations include a different release, an obstacle, an incorrect measurement or a genuine variation in the trial. The table alone does not identify which explanation is true.
A suitable response is to check the recorded procedure and, where the task allows, repeat the trial under the intended conditions. If the investigation notes state that the car struck a chair leg during that run, you have a documented reason that the trial did not follow the intended unobstructed path.
If no such note exists, do not invent one. Say that the unusual result should be investigated. Keep it in the record while deciding what additional evidence is needed.
A question may ask for an improvement. An answer such as “repeat the test using the same release point and ensuring the path is clear” can address consistency if those are relevant weaknesses. But “delete the lowest value” is not a sound general improvement.
If the problem explicitly instructs you to exclude a trial because a stated error occurred, follow that information. The key difference is evidence for the exclusion, not whether the number is convenient.
Why extra repeats alone may not resolve everything
If a systematic mistake is repeated, more trials can reproduce it. A ruler placed 5 cm away from the start point could produce similar overestimated distances every time. The procedure needs correction. Repetition after correction can then check the revised method.
An unusual reading and a repeated bias call for different responses. One raises doubt about a particular trial. The other raises doubt about the measurement method across trials. Your improvement should identify the issue and explain how the proposed action helps.
For deeper measurement instruction, use the existing measurement-error guide. Keep this workshop’s narrower task in view: decide what the readings allow you to conclude before suggesting a repair.
Contents · Previous chapter · Next chapter
Consider an invented test of two plant-growing conditions. Plants in Group A receive more light and more water than plants in Group B. The pupil repeats this arrangement several times and observes that A grows more.
The repeated pattern may be consistent, but it does not isolate the effect of light. Water changed too. If the question asks whether greater light caused greater growth, the design leaves another changed factor that could contribute.
A better investigation for the effect of light would keep relevant factors such as water amount and plant type comparable while changing the intended light condition. Repeating that improved comparison would then provide evidence about consistency under a fairer design.
Do not answer every improvement question with “repeat three times and take the average”. First inspect the comparison. If two relevant variables changed, repetition keeps the same confusion unless you alter the design.
Likewise, an experiment comparing two containers with different starting temperatures cannot attribute every final-temperature difference to their coverings. The starting condition matters. Repeated unequal starts do not become equal because they are repeated.
Use a two-question audit: “Were the compared conditions suitable for the intended question?” and “Were the results reasonably consistent when repeated?” The first concerns the design of the comparison. The second concerns the pattern in the repeated outcomes.
These questions can lead to different judgements. A fair comparison can have variable readings that need investigation. An unfair comparison can have very consistent readings. A strong answer distinguishes the weaknesses instead of using “reliable” as a label for every kind of quality.
Repair a design-blind conclusion
Draft: “More light made the plants taller because A was taller in all three trials.” Repair: “A grew taller in all three trials, but the test does not isolate the effect of light because A also received more water.”
The repair recognises the repeated observation without overstating its cause. It does not say the data are useless. They show an outcome under combined conditions, while leaving the separate contribution of light unresolved.
For more detailed investigation planning, return to the Science Hub. Use the relevant existing owners rather than trying to learn every variable-control method from this one repeated-results workshop.
Contents · Previous chapter · Next chapter
A conclusion needs a boundary. “Under these conditions” is not an empty phrase when it identifies what was actually tested. It prevents a small investigation from being turned into a universal statement.
Suppose a toy car travels farther from a higher release point in three repeated comparisons. The test uses one car, one ramp surface and three specified heights. The evidence supports the observed relationship in that setup. It does not prove that increasing height indefinitely will always increase distance in every environment.
At larger heights, the car might leave the track or strike an obstacle. Those possibilities are not part of the test. You do not need to list them in an ordinary answer, but you should avoid an unrestricted “always” claim.
Similarly, a plant investigation over one week does not automatically establish what will happen over a year. A test of one material sample does not establish every possible product made from that material. Keep the duration, sample and outcome attached when they matter to the question.
This does not mean every conclusion must be so cautious that it says nothing. If all relevant repeated results show the same clear comparison under controlled conditions, state that comparison directly. Then avoid adding a larger claim.
A strong sentence can be both clear and bounded: “In the recorded ten-minute trials, Covering B gave a smaller temperature decrease than Covering A.” It tells the reader exactly what the data support.
Prediction is a different job
A question may ask you to predict a result for a similar untested condition. You can use the established pattern, but recognise that a prediction is an expected result rather than a recorded observation.
If release heights of 10, 20 and 30 cm show increasing distances, you may predict a greater distance at a nearby height if the same setup and pattern reasonably apply. Do not label the predicted distance as something the pupil measured.
Explain the basis of the prediction and keep it proportionate. A qualitative prediction may be appropriate when exact numerical behaviour is not established. Drawing a precise straight-line extension without evidence can add certainty that the task does not support.
Contents · Previous chapter · Next chapter
Start with short datasets so that you can concentrate on the reasoning job. Attempt each task before reading its discussion.
Task A: “The times for a sample to dissolve were 45 s, 44 s and 46 s. What quantity was measured, and were the readings close?” The quantity was the time taken to dissolve, in seconds. The readings were close, spanning only 2 s. This describes the data; it does not identify the exact reason for the small differences.
Task B: “Setup X gave travel distances of 30 cm, 31 cm and 29 cm. Setup Y gave 42 cm, 41 cm and 43 cm. Which produced the greater distance in the recorded trials?” Y did in every trial. A clear answer refers to distance, not speed, because speed was not directly measured by these readings.
Task C: “Procedure A gave 20, 21 and 19 drops. Procedure B gave 10, 20 and 30 drops. Both averages are 20. Which set was more consistent?” A, because its readings were much closer together. Equal averages do not establish equal consistency.
Task D: “A pupil used warm water for one sugar sample and cold water for another, but also stirred only the warm water. The pupil repeated the test three times. Can the results isolate the effect of water temperature?” No. Stirring differed too, so the comparison does not separate the effect of temperature.
Task E: “Distances were 65 cm, 66 cm and 28 cm. The notes state that the car hit a bag in the third trial. What should be improved?” Clear the path and repeat the affected condition using the intended procedure. The documented obstruction explains why that run did not match the unobstructed test.
Read your answer against the measured outcome
If you wrote “Y was faster” in B, ask what information speed requires. Distance alone does not establish speed without the relevant time information. If you wrote “A was accurate” in C, ask where the true or accepted value was supplied.
These repairs train you to keep the conclusion tied to the measured quantity. The data table can look familiar while the question asks for a different judgement. Read the heading and unit before choosing a science term.
Contents · Previous chapter · Next chapter
Original teaching investigation: “A pupil placed equal-sized wet cloth pieces on two racks. The pieces had the same material and the same starting amount of water. Rack M was placed in moving air and Rack N in still air. The pupil measured the time taken for each piece to dry and repeated the comparison.”
The recorded times for M were 18 minutes, 20 minutes and 19 minutes. For N, they were 35 minutes, 36 minutes and 34 minutes. Assume the task states that other relevant conditions were kept the same.
Question A: What was the measured outcome? Question B: What repeated pattern supports a conclusion? Question C: Give a conclusion within the tested conditions. Question D: Explain the result using an appropriate mechanism.
For A, the measured outcome was time taken for the wet cloth to dry. For B, the moving-air condition had a shorter drying time in every recorded comparison, and the readings within each condition were close.
For C, a suitable conclusion is: “Under the stated conditions, the cloth in moving air dried sooner than the cloth in still air.” Avoid “moving air removed every drop instantly” or “still air stops evaporation”. Both are contradicted by the recorded drying times.
For D, explain that moving air carries away water vapour near the cloth, supporting faster evaporation under these conditions. Keep the explanation connected to the water in the cloth changing to water vapour. Do not say the wind created water or converted the cloth itself into vapour.
Question E: A learner says, “M had less water because its number was smaller.” What is wrong with this answer? The number records drying time, not the remaining amount of water. The outcome was misidentified.
Question F: Another learner says, “The experiment was fair because the pupil repeated it.” What should be repaired? The equal material, size, starting water amount and other controlled conditions support fairness. The repeats support the judgement that the observed timing difference was consistent.
Change one design detail
Suppose M’s cloth started with less water than N’s. Can you still attribute the shorter drying time entirely to moving air? No. The starting water amount also differs. More repeats of that unequal comparison would not remove the problem.
Write an improvement that addresses the changed detail: use the same starting amount of water for comparable cloth pieces. Then repeat the improved comparison if the task asks how to check consistency.
Contents · Previous chapter · Next chapter
On another day, attempt a new dataset without reading the earlier discussion. Use the same audit questions: measured outcome, repeat grouping, pattern, design and conclusion boundary.
Delayed task: “Identical strips of two materials were tested with the same load. The times until breaking were A: 24 s, 25 s, 23 s; B: 25 s, 9 s, 26 s. No procedural notes were supplied.”
Can you conclude that B is consistently longer-lasting than A? No. B’s readings vary substantially and do not establish a consistent advantage. One B reading is much lower. Investigate the procedure and gather appropriate further evidence rather than ignoring that reading.
Can you conclude that A’s recorded times are more consistent? Yes. They are closer together. Can you conclude that A must always be stronger than B? No. That broader claim exceeds the stated outcome and conditions.
The distinction matters under examination pressure. A familiar average calculation may tempt you to skip the data-reading job. Pause long enough to identify what the question asks you to judge. Then use the relevant readings, including awkward ones.
When a question asks for evidence, cite the pattern or values that support the answer. When it asks for an explanation, connect the outcome to the scientific mechanism. When it asks for an improvement, name the weakness and explain how the change addresses it.
The SEAB 2026 PSLE formats page is the current-year official reference checked for this guide. The revised Standard Science framework assesses knowledge with understanding and application of knowledge and scientific inquiry, including evaluation and communication of reasoning.
This workshop’s audit is a teaching tool, not a mandatory official answer template. Use your school’s instructions and the relevant Standard or Foundation examination document. Do not assume their paper formats are identical.
A useful final check is: “Have I claimed more than these results show?” Remove an unsupported “always”, an invented cause of variation, or a conclusion about an unmeasured quantity before polishing the sentence.
Contents · Previous chapter · Next chapter
Adults can isolate this difficulty by giving a learner two datasets with the same average but different variation. Ask what is the same and what is different. Then ask what conclusion each feature can support.
If the child says “repeat for reliability”, ask, “What would similar repeats tell us here?” This encourages an explanation rather than a memorised phrase. If the child cannot identify the measured outcome, begin there before discussing consistency.
Do not require advanced statistics to explain simple repeated patterns. A learner can compare actual readings and describe whether they are close. The important skill is connecting the evidence to the judgement requested.
Does three trials mean the results are reliable?
Not automatically. Inspect whether the trials used comparable conditions, whether the readings agree reasonably and whether the method has a relevant weakness. A count of trials is information about the design, not proof of every kind of quality.
Should I always take an average?
No. Use an average when requested or when it helps summarise comparable repeated readings. Also inspect the individual readings. An average can hide large variation, and combining different conditions can conceal the relationship being investigated.
Can I ignore one strange result?
Not merely because it looks inconvenient. Look for a documented procedural problem or other evidence that explains why the trial did not represent the intended test. Without that evidence, investigate the unusual result and keep the uncertainty visible.
Are close readings accurate readings?
They are consistent recorded readings. To judge accuracy, you need evidence about the measurement’s closeness to the true or accepted value. A repeated mistake can produce close numbers. Use the right word for the evidence you have.
How long should my conclusion be?
Long enough to state the measured relationship and any condition needed to prevent overclaiming. A concise sentence grounded in the whole pattern is better than a long paragraph that mixes observation, explanation and speculation.
What should I practise after this?
Use one new repeated dataset and write four separate statements: what was measured, what the readings show, what conclusion follows and what remains unresolved. Then check whether each statement does its own job.
Return to the PSLE Learning Guide, the PSLE Science guide library or the Science Hub for deeper topic work. Your next performance habit is to let the full set of evidence control the strength of the conclusion.
Contents · Previous chapter · Choose your next PSLE route
Continue through these four performance workshops
Vol 0081: English: Keep Each Comprehension Answer Inside the Requested Time Window
Vol 0082: Mathematics: Audit a Transfer by Recording Both the Giver and the Receiver
Vol 0083: Science: Judge What Repeated Results Let You Conclude
Vol 0084: English, Mathematics and Science: Test Whether Your Answer Can Be Understood Without You