A Primary 6 Science question can hide its difficulty inside a representation. The concept may be familiar, but the pupil has to extract it from a graph, table, sequence, food web, apparatus diagram or set of observations before any explanation can begin.
The most common failure is not “I cannot read a graph.” It is subtler: the pupil reads one value but misses the relationship, sees a trend but turns it into a cause, notices an average but treats it as every trial, or explains a result using facts that were never established by the evidence.
This guide develops the evidence-reading layer of the Primary 6 Science Learning Hub. All practice material is original eduKate teaching content.
Quick answer: what is the evidence job?
Use a five-step route:
READ → DESCRIBE → RELATE → EXPLAIN → LIMIT.
- Read labels, units, conditions, axes, arrows and legends.
- Describe what the representation directly shows.
- Relate the variables or parts.
- Explain the relationship using relevant Science when the question asks for mechanism.
- Limit the claim to what the evidence can support.
This is an eduKate reasoning routine, not an official SEAB formula.
SEAB’s 2026 PSLE Science objectives include interpreting and analysing information, evaluating observations and information, and communicating explanations and reasoning. Official reference: SEAB PSLE Formats Examined in 2026.
Observation, inference and explanation are different jobs
A strong Science answer separates what was directly given from what is concluded from it.
Observation
An observation is what is seen, measured or stated.
Example: “The plant produced 18 bubbles in five minutes at 20 cm from the lamp and 9 bubbles at 40 cm.”
Inference
An inference is a conclusion drawn from observations and scientific knowledge.
Example: “The plant appears to have a higher measured rate of gas-bubble production when placed closer to the lamp under the tested conditions.”
Explanation
An explanation gives the scientific mechanism.
Example: “The closer lamp provides greater light exposure, so more light energy is available for photosynthesis under the other stated conditions, producing a higher measured bubble rate.”
The exact wording should match the evidence and school expectations. The important distinction is that the observation comes from data; the explanation adds the Science.
Do not explain before you describe
Pupils often jump from a graph to a mechanism and accidentally misread the graph. Requiring one descriptive sentence first protects the reasoning.
For a graph showing stopping distance against surface type, first state the observed pattern: “The car travelled a shorter distance on Surface C than on Surface A.” Only then explain using friction if the question asks why.
Reading a table: row, column, comparison
Tables look simple, but they create three common errors:
- reading across the wrong row;
- matching a value to the wrong condition;
- comparing two numbers that do not answer the question.
Before calculating or explaining, point to the column heading and row label for each value.
Example table
| Lamp distance | Bubbles in 5 min |
|---|---|
| 10 cm | 24 |
| 20 cm | 18 |
| 30 cm | 12 |
| 40 cm | 9 |
Direct observation: as lamp distance increases across these tested values, bubble count decreases.
Do not say “photosynthesis stops at 40 cm”. The table shows nine bubbles, not zero. Do not say “every 10 cm halves photosynthesis”. The numerical pattern does not support that.
Reading a graph: axes before shape
A pupil can recognise an upward line and still misread what increased.
Use this order:
- Read the horizontal-axis variable and unit.
- Read the vertical-axis variable and unit.
- Check whether the scale is regular.
- Identify the exact points relevant to the question.
- Describe the relationship before explaining it.
Increase by versus increase to
These phrases describe different quantities.
If temperature rises from 25°C to 40°C:
- it increases to 40°C;
- it increases by 15°C.
Confusing final value with change is a common graph and table error.
Value after the same time versus time to reach a value
Two cooling experiments can be compared in different ways.
- Temperature after 10 minutes: compare values at a fixed time.
- Time taken to cool to 40°C: compare durations to a fixed value.
The graph may contain both kinds of information, but the question decides which one matters.
Average is not every trial
Suppose three spring-extension trials are 4.8 cm, 5.0 cm and 5.2 cm. The average is 5.0 cm.
Correct: “The average extension was 5.0 cm.”
Incorrect: “The spring extended 5.0 cm in every trial.”
Average is a summary. Keep the underlying variation visible when judging evidence quality.
Range and variation: when repeated results disagree
At Primary 6, pupils do not need advanced statistics to notice unstable measurements.
Compare:
- Set A: 41, 42, 41 cm.
- Set B: 18, 51, 33 cm.
Both sets have three results, but Set B varies much more. A pupil should ask whether the procedure was inconsistent, whether the system itself varies, or whether more evidence is needed.
An outlier is a question, not rubbish
If most results are near 80 cm and one is 34 cm, do not delete 34 simply because it looks inconvenient.
Ask:
- Was the car released differently?
- Was an obstacle present?
- Was the ruler read incorrectly?
- Could the system genuinely produce that result?
- What happens if the trial is repeated?
Unexpected data can reveal a mistake, a hidden variable or real variation.
Trend is not automatically cause
If two quantities change together, the graph shows a relationship. Causation requires stronger reasoning about how the data were produced.
If students who sleep more also score higher, the graph alone does not prove that extra sleep caused every score difference. Other factors could be involved.
In a controlled experiment where one factor is deliberately changed and relevant alternatives are held comparable, causal interpretation may be stronger. The method matters.
Direct and indirect evidence
Direct evidence measures or observes the thing closely connected to the claim. Indirect evidence measures something related to it.
Example: measuring gas volume from an aquatic plant is more direct for gas production than measuring plant height. Plant height can be influenced by many factors and occurs over a different timescale.
Indirect evidence is not useless. It simply requires careful interpretation.
Measured outcome versus check on a controlled condition
A thermometer can play different jobs.
In an investigation of cooling rate, temperature may be the measured outcome.
In an investigation of light exposure on photosynthesis, temperature may be monitored to check that both setups remain comparable.
The instrument does not determine the variable’s role. The investigation question does.
Diagram reading: objects, labels, pathways and state
Science diagrams compress information. Before inferring anything, identify:
- what each label refers to;
- whether arrows show movement, force, flow or sequence;
- whether parts are physically connected;
- whether the diagram shows one moment or several stages;
- whether size and distance are drawn to scale.
Connected or just close together?
Two objects drawn near each other may not be connected. Look for tubes, wires, arrows, contact points or explicit wording.
In an electrical diagram, a tiny gap can mean an open circuit. In a plant setup, a tube may carry gas from one container to another. Spatial closeness alone should not replace connection evidence.
Sequence: one object changing or several objects?
A row of diagrams may show the same object at different times or different objects at one time. The interpretation changes completely.
Look for labels such as “after 2 min”, “after 4 min” or repeated specimen names. If the question says “three identical springs”, do not treat them as one spring changing through time.
Cycle diagrams: the starting point may move
A cycle has no unique natural “first” stage. The diagram may begin at any point.
Instead of memorising a fixed top-left sequence, identify the process linking each stage to the next. This matters for life cycles, water cycles and repeating processes.
Food webs: arrows, roles and alternative pathways
Food webs are evidence maps of feeding relationships. A change in one organism can have multiple possible routes.
If insects decline, birds that eat insects may have less food. But if those birds have another food source, the effect may be smaller than a pupil who reads only one arrow predicts.
Trace all relevant arrows before concluding.
Graph workshop 1: Spring extension
A pupil records:
| Load | Extension |
|---|---|
| 1 unit | 2 cm |
| 2 units | 4 cm |
| 3 units | 6 cm |
| 4 units | 8 cm |
Question A: State the pattern.
Answer: In the tested range, greater load is associated with greater spring extension.
Question B: Can we predict exactly 20 cm extension at 10 units?
Answer: Not confidently from these data alone. That would extrapolate far beyond the tested range, and the spring may not continue behaving in the same way.
Question C: Why is “the spring becomes stronger” not supported?
Answer: The table measures extension, not spring strength. The claim introduces a different property.
Graph workshop 2: Plant bubble production
| Light condition | Average bubbles/min |
|---|---|
| Low | 3 |
| Medium | 7 |
| High | 11 |
Question A: Observation
Answer: Average bubble production increased from 3 to 7 to 11 bubbles per minute as the tested light condition increased.
Question B: Explanation
Answer: Greater light exposure provides more light energy for photosynthesis under the stated comparable conditions, resulting in a higher measured bubble-production rate.
Question C: Limitation
Answer: Bubble count is an indirect measure of gas production and bubble sizes may differ; the claim should stay tied to the measured indicator unless gas volume or identity is established.
Table workshop 3: Friction
| Surface | Trial 1 | Trial 2 | Trial 3 |
|---|---|---|---|
| P | 78 cm | 80 cm | 79 cm |
| Q | 52 cm | 51 cm | 53 cm |
| R | 31 cm | 30 cm | 32 cm |
Question A: Which surface produced the shortest travel distance?
Answer: Surface R.
Question B: What can be inferred about friction if other relevant conditions were controlled?
Answer: Surface R produced the greatest frictional effect on the moving car among the tested surfaces, causing it to slow and stop over the shortest distance.
Question C: Why are the repeated trials useful?
Answer: They show that the pattern is consistent and not dependent on one isolated measurement.
Data workshop 4: Environment
A pond survey records the following organism counts:
| Month | Water plants | Small fish | Large fish |
|---|---|---|---|
| June | 120 | 80 | 20 |
| July | 90 | 71 | 20 |
| August | 55 | 48 | 17 |
Question A: State one observed pattern.
Answer: Water-plant count and small-fish count both decreased from June to August.
Question B: Can the table alone prove that fewer plants caused fewer small fish?
Answer: No. The counts change together, but the table alone does not isolate cause. Other environmental changes may have occurred.
Question C: What additional evidence would help?
Answer: Information about food relationships, water conditions, temperature, other organisms, predation, reproduction or a controlled investigation could help test possible explanations.
Qualitative and quantitative evidence
Quantitative evidence uses numbers: length, time, temperature, count, mass or volume.
Qualitative evidence describes qualities: colour change, smell, texture, movement pattern or presence/absence.
Neither is automatically superior. The useful evidence is the evidence that answers the scientific question.
A colour change may be decisive in a chemical indicator test. A numerical temperature may be better for comparing cooling. Strong investigations often combine both.
Same relationship, different representations
A table and a graph can show the same data. A written statement can describe the same relationship. Pupils should be able to translate among them.
Table: as lamp distance increases, bubble count decreases.
Graph: downward trend of bubble count with increasing distance.
Sentence: “The plant produced fewer bubbles when the lamp was farther away under the tested conditions.”
The representation changes; the scientific relationship should not.
Two datasets: do they show the same relationship?
Compare two experiments using different absolute values. One may measure 20, 15, 10; another 40, 30, 20. The numbers differ, but both may show the same directional relationship.
Do not ask only whether values match. Ask whether the pattern between variables matches.
Plateau: when more change stops producing more measured effect
A graph may rise and then level off. Pupils sometimes extend the early trend indefinitely.
A plateau tells us that within the observed range, additional change in the horizontal-axis variable is no longer associated with much change in the measured outcome.
Possible scientific explanations depend on the topic and evidence. For photosynthesis, another required condition may become limiting. But do not name a specific limiting factor unless the question supports it.
Threshold: when an effect appears after a point
Some systems show little change until a condition crosses a threshold. A spring may begin moving an object only after enough force is applied; a seed may germinate only when necessary conditions are met.
Read the graph carefully. “No observed effect below this tested value” is stronger evidence than “nothing can ever happen below it.”
Return to starting value does not mean nothing happened
A graph may rise and later return to its starting value. The final value matches the start, but the system changed in between.
Example: temperature rises from 25°C to 60°C, then cools back to 25°C. It is false to say “temperature did not change.”
Always consider the path, not only the final point.
Condition changes partway through a time graph
If a lamp is switched off at minute 10, the graph before and after minute 10 may represent different conditions. Do not describe the whole graph as if one condition applied throughout.
Mark the transition point and analyse each phase separately before explaining the change.
Signal and noise
Not every small fluctuation deserves a scientific story.
If repeated measurements are 10.1, 10.0, 10.2 and 10.1, tiny differences may reflect measurement variation rather than meaningful system change.
Pupils should look for patterns large and consistent enough to support the claim being made.
Missing data are not zero
A blank cell, unmeasured point or absent observation does not mean the value is zero.
If a graph stops at 60 seconds, we do not know the value at 120 seconds unless a justified prediction is requested.
Extrapolation: beyond the observed range
Predicting between measured points is usually safer than predicting far beyond them.
Suppose spring extension is measured for loads 1–4. Predicting at 2.5 lies within the observed range. Predicting at 20 assumes the relationship continues much farther than tested.
When extrapolating, use cautious language and recognise reduced confidence.
Scientific evidence has boundaries
A result may support one claim and not another.
Evidence: a plant produced more bubbles under higher light.
Supported: higher light was associated with higher bubble-production rate under tested conditions.
Not directly supported: the plant grew taller, produced more seeds, absorbed more minerals or will always perform better under unlimited light.
The strongest pupils learn to say “the evidence does not tell us that” without feeling that they are failing to answer.
Evidence error clinic
- Error: turning correlation into causation. Repair: inspect how the data were produced.
- Error: using a final value as a change amount. Repair: subtract start from finish when change is required.
- Error: treating an average as every trial. Repair: distinguish summary from individual values.
- Error: assuming missing data are zero. Repair: mark them as unknown.
- Error: ignoring units. Repair: read axis and table headings before values.
- Error: explaining before describing. Repair: state one observation first.
- Error: forcing a smooth pattern through an outlier. Repair: investigate unexpected data.
- Error: extrapolating as certainty. Repair: reduce confidence beyond tested conditions.
- Error: confusing a diagram’s appearance with physical scale. Repair: use labels and stated measurements.
- Error: reading one food-web arrow and ignoring alternatives. Repair: trace all relevant pathways.
A representation-switch drill
Take one relationship and express it four ways.
Relationship: the car travels a shorter distance as surface roughness increases.
- Words: state the relationship.
- Table: invent three plausible roughness categories with distances.
- Graph: sketch the expected trend.
- Diagram: draw the car, surface and measured distance.
If the learner changes the scientific meaning while changing format, the representation skill needs repair.
How to answer “compare” from data
A comparison should use the same basis.
Weak: “Plant A is taller; Plant B has fewer leaves.”
Stronger: “Plant A is taller than Plant B.” Then, if asked for another difference: “Plant A also has more leaves than Plant B.”
Use like-with-like comparisons so the reader can see the contrast.
How to answer “describe the trend”
Describe what changes as the other variable changes. Avoid mechanism unless asked.
Example: “As time increased from 0 to 10 minutes, temperature decreased rapidly, then decreased more slowly from 10 to 20 minutes.”
That is more informative than “temperature went down.”
How to answer “use the data to support”
Quote the relevant values or pattern and connect them to the claim.
Claim: Surface Q produces more friction than Surface P.
Support: “The car travelled only 52 cm on Q compared with 79 cm on P under the same release conditions, showing that it was slowed more on Q.”
The data are not decoration. They are the evidence in the argument.
How parents can practise evidence discipline
Use everyday claims:
- “This plant grows better here.” What measurements would show “better”?
- “This fan is stronger.” What outcome would be measured?
- “This cup keeps water warm longer.” Compare temperature after the same time or time to reach the same temperature?
- “This shoe has more grip.” What controlled test could provide evidence?
Ask the child to separate observation from explanation before debating the answer.
How teachers can diagnose representation failures
If the pupil understands a concept orally but fails on a graph question, isolate the representation job.
- Can the pupil read axes and units correctly?
- Can the pupil match values to conditions?
- Can the pupil state a trend without explaining it?
- Can the pupil identify an outlier?
- Can the pupil translate the graph into a sentence?
- Can the pupil distinguish direct evidence from inference?
Do not reteach the whole chapter if the failure is a graph-reading operation.
Bridge to PSLE explanation
Evidence reading is only useful if it changes the answer. The pupil must select the command, choose the relevant Science, build the causal chain, and write only what the question needs.
That is the final guide in this batch.
Continue the Primary 6 Science Learning Guide series
- Guide 1: Concepts, Systems, Interactions & Energy
- Guide 2: Investigations, Variables, Fair Tests & Method
- Guide 4: PSLE Questions, Explanations, Error Analysis & Revision
Return to the Primary 6 Science Learning Hub.
The quiet return
Data do not speak. A pupil has to read the representation, preserve the labels, distinguish what is seen from what is inferred, and keep the claim within the evidence.
Read exactly. Describe first. Explain second. Carry the evidence into the sentence. Stop where the evidence stops.
