Wait, what? Three clues can point toward one answer and one clue can still prove it wrong.
That sounds unfair only if evidence is counted like votes.
Science does not ask how many clues support an idea. It asks how relevant, reliable, direct and discriminating those clues are.
This preserved Hougang Science Tuition P6 URL now owns one precise job: evidence hierarchy and weighting. The old duplicated tuition advertisement, stale schedule, locality conflicts, grade promises and unrelated image stack have been removed.
This page is deliberately distinct from the Hougang P6 guide on integrating multiple evidence sources. Integration asks, “How do all the pieces fit?” Evidence hierarchy asks:
When the pieces do not deserve equal weight, which evidence should control the conclusion?
Evidence is not a vote count
Imagine four observations:
- three rough visual estimates suggest Setup A produced more;
- one calibrated measurement shows Setup B produced a higher measured amount.
The three estimates do not automatically outweigh the measurement.
Why?
- The measurement may be more direct.
- The instrument may be more precise.
- The measured quantity may match the claim more closely.
- The estimates may all repeat the same underlying weak observation.
Counting evidence pieces without judging quality can produce the wrong conclusion.
Four dimensions of evidence weight
A useful Primary 6 framework is:
- Relevance: Does this evidence answer the actual claim?
- Reliability: How trustworthy is the observation or measurement?
- Directness: Does the evidence measure the claim directly or only indirectly?
- Discriminating power: Does this evidence separate competing explanations?
These four dimensions are more useful than simply asking whether evidence “supports” an answer.
Relevant evidence beats interesting evidence
A fact can be scientifically true and still be irrelevant.
Suppose a question asks which setup lost more water. Information about container colour may be interesting, but unless colour affects the tested relationship or the question establishes that it matters, it should not outweigh measured water loss.
Ask:
- Does this fact change the conclusion?
- Does it explain the measured difference?
- Does it identify a changed condition or control?
- Or is it merely part of the story?
Evidence earns weight by relevance to the claim.
Direct evidence often deserves more weight than indirect evidence
Direct evidence measures or observes the quantity or event close to the claim itself.
Indirect evidence requires an extra inference.
Example:
- Direct: measured temperature increased by 8°C.
- Indirect: the container felt warmer by touch.
The tactile observation can be useful, but the measured value usually deserves greater weight when the claim is about temperature change.
The child should learn that “closer to the claim” often means fewer unsupported inference steps.
Reliable evidence is evidence we have good reason to trust
Reliability can improve through:
- appropriate instruments;
- consistent method;
- repeated measurements;
- controlled conditions;
- clear scale reading;
- careful recording;
- agreement across repeated trials.
But repeated weak evidence does not automatically become strong.
Ten repetitions of an unfair comparison remain ten repetitions of an unfair comparison.
This distinction is essential in experimental evaluation.
Discriminating evidence can outweigh many supporting clues
Suppose two explanations both fit most of the evidence.
Explanation A and Explanation B both predict that a bulb should not light in one setup. But only A predicts that the same bulb will light when transferred to a known complete circuit.
If the bulb does light in the known complete circuit, that observation strongly discriminates between the explanations.
It deserves more weight than several vague clues that both explanations already fit.
The strongest evidence is often the evidence that one model predicts and the other cannot explain.
Contradictory evidence can have veto power
A universal claim can be broken by one trustworthy counterexample.
If a student claims “all tested metal objects are attracted to a magnet” and one carefully verified metal object is not attracted, the universal rule must be revised.
The counterexample does not merely subtract one vote. It changes the logical status of the claim.
This is why one high-quality contradiction can outweigh many confirming examples.
But unusual evidence should be checked before it is given veto power
One strange result should not automatically overthrow a strong pattern if the result itself may be unreliable.
Before giving the contradiction decisive weight, ask:
- Was the measurement recorded correctly?
- Was the same method used?
- Did an uncontrolled condition change?
- Does the result repeat?
- Is the object or sample what we think it is?
Evidence hierarchy requires judging quality before judging impact.
Independent evidence is stronger than repeated copies of the same evidence
Three observations may look independent while actually coming from one source.
For example:
- a diagram looks larger;
- the object appears to occupy more space;
- the picture seems to show greater volume.
These may all be the same visual cue repeated in different words. They should not be counted as three separate lines of evidence.
By contrast:
- measured mass;
- measured volume;
- observed behaviour under a test;
may provide different kinds of evidence.
Repeated evidence strengthens confidence when the method is sound
If the method is appropriate and the same result appears repeatedly, confidence increases.
Repeated evidence helps answer:
- Was the first result a one-off?
- Is the measurement stable?
- Does the pattern persist?
- Is the unusual result reproducible?
Repetition supports reliability. It does not automatically increase relevance or directness.
Method quality changes evidence weight
Two results can be numerically precise and still deserve different confidence because their methods differ.
Suppose:
- Experiment A changes one factor and controls the rest well.
- Experiment B changes three important factors at the same time.
Even if both produce a clear difference, A provides stronger evidence about the effect of the intended variable.
The numerical result is only as interpretable as the design allows.
Measurement resolution changes evidence weight
If two setups differ by 0.2 units but the instrument only reads to the nearest whole unit, the apparent difference may not deserve strong confidence.
Ask:
- Is the measured difference larger than the instrument’s useful resolution?
- Would repeated readings distinguish the setups more clearly?
- Could the difference be rounding or reading variation?
Small numerical differences can be weak evidence when measurement resolution is coarse.
Temporal evidence can have different weight
One measurement taken too early may not represent the final system response.
A later measurement may be more relevant if the claim concerns the final state. An early measurement may be more relevant if the claim concerns response speed.
Evidence weight depends on which time point matches the task.
Evidence close to the causal mechanism can be especially useful
Suppose a whole-system outcome changes. One piece of evidence shows the final outcome; another directly shows the intermediate process that the explanation claims changed.
The intermediate-process evidence may strengthen the causal account because it connects the changed condition to the final outcome.
This is why mechanism evidence can be more informative than an endpoint alone.
Negative evidence can be strong only when detection was possible
“We did not observe X” can weaken a model only if the method should have detected X if it were present.
Ask:
- Was the instrument sensitive enough?
- Was the observation long enough?
- Was the relevant part actually examined?
- Would the predicted effect be large enough to see?
Absence of evidence is stronger when the search was capable of finding the expected evidence.
An evidence hierarchy is claim-specific
There is no universal ranking where one source type is always best.
A graph may be decisive for a trend claim. A diagram may be decisive for a connection claim. A controlled comparison may be decisive for a causal claim. A direct observation may be decisive for whether an event occurred.
The learner should therefore ask:
Strong evidence for which claim?
This keeps hierarchy tied to the question rather than to a memorised ranking.
When evidence conflicts, diagnose before averaging
Students may try to “split the difference” when sources disagree.
Instead ask:
- Do the sources measure the same quantity?
- Do they refer to the same time?
- Were the methods equally reliable?
- Is one source indirect?
- Is one result anomalous?
- Does one source directly discriminate between the competing explanations?
Conflict should trigger diagnosis, not compromise by arithmetic.
The decisive-clue test
When several clues are present, ask:
- Which clues merely support both explanations?
- Which clue is most directly tied to the claim?
- Which clue has the strongest method behind it?
- Which clue would be hardest for the wrong explanation to account for?
The last clue may deserve the greatest weight.
The weakest-link evidence audit
Sometimes a strong-looking conclusion depends on one weak evidence link.
Example:
- Measurement is precise.
- Graph is clear.
- But the comparison was unfair.
The unfair comparison limits the causal claim despite the polished data.
Ask:
What is the weakest evidence-generating step that my conclusion depends on?
This prevents students from being impressed by precision after validity has already failed.
Evidence hierarchy in MCQ
In Booklet A, several details may seem relevant. Instead of letting every clue pull equally, identify the controlling condition.
- Which stated fact eliminates an option completely?
- Which measurement is more direct than the visual impression?
- Which condition makes one familiar rule inapplicable?
- Which single contradiction defeats the distractor?
Strong MCQ solving often depends on finding the decisive evidence early.
Evidence hierarchy in structured answers
In Booklet B, the final answer should foreground the evidence that actually supports the mechanism.
Do not list every observation simply because it is available.
- State the most relevant comparison.
- Use the scientific mechanism.
- Add supporting evidence only where it strengthens the explanation.
- Avoid clutter that does not advance the claim.
Evidence hierarchy can improve answer efficiency as well as accuracy.
A simple evidence-weight table
| Evidence | Relevant? | Reliable? | Direct? | Discriminating? |
|---|---|---|---|---|
| Clue A | ? | ? | ? | ? |
| Clue B | ? | ? | ? | ? |
| Clue C | ? | ? | ? | ? |
This is a learning scaffold for difficult questions, not an exam requirement.
Five Primary 6 evidence-weighting failure modes
1. Majority-vote thinker
The student counts supporting clues. Repair by judging relevance, quality and discrimination.
2. Precision-impressed thinker
Detailed numbers are trusted even when the method is unfair. Repair by auditing validity before precision.
3. Repetition-equals-strength thinker
Repeated weak evidence is treated as automatically strong. Repair by asking whether the underlying method is sound.
4. One-anomaly-overthrow thinker
One strange result is given decisive weight before its reliability is checked. Repair by investigating anomaly quality first.
5. Evidence-list writer
Every clue is included equally in the answer. Repair by foregrounding the evidence with the strongest relation to the claim.
A Phase 4 Primary 6 evidence-hierarchy lesson
- Claim: state exactly what must be decided.
- Collect: identify all relevant evidence sources.
- Relevance: remove story details that do not bear on the claim.
- Reliability: inspect method and measurement quality.
- Directness: identify evidence closest to the claim.
- Discriminate: find the clue that separates competing explanations.
- Contradict: inspect any trustworthy evidence against the claim.
- Weight: decide which evidence should dominate.
- Integrate: build the conclusion using the strongest evidence first.
- Bound: keep confidence proportional to the evidence quality.
Why small groups help with evidence weighting
Three students may reach three answers because each gives a different clue the greatest weight.
- Why did you prioritise that evidence?
- Is it direct or indirect?
- How reliable is the method behind it?
- Does it actually discriminate between the options?
- What contradictory evidence must your explanation survive?
The disagreement becomes a lesson in evidence quality rather than confidence.
What parents can practise at home
- Ask which clue is most important and why.
- Ask whether three clues are actually independent.
- Ask whether a measurement is more direct than a visual impression.
- Ask what evidence would defeat the current explanation.
- Ask whether repeated evidence comes from a sound method.
- Ask which fact separates two plausible answers.
How this page fits the Hougang Science network
This eduKateSingapore page owns evidence hierarchy and weighting. It complements multi-source evidence integration, competing explanations, evidence and method evaluation, and scientific checking and uncertainty.
For the complete P3-to-PSLE map, use Hougang Primary Science Learning Library.
Official 2026 examination reference
For Standard Science examined in 2026, SEAB lists the revised PSLE Science subject as syllabus 0009. The official assessment includes interpretation and analysis of information, evaluation of observations and methods, and communication of explanations and reasoning. See PSLE Formats Examined in 2026.
A strong Primary 6 scientist does not ask, “How many clues support my answer?” They ask which evidence is most relevant, reliable, direct and discriminating—and whether one trustworthy contradiction should force the entire explanation to change.