Primary 6 Science Learning Guide | Confidence, Uncertainty, Anomalies & Strength of Evidence for PSLE

Primary 6 Science is not only about getting an answer; it is about knowing how strongly the evidence supports that answer. A single trial may suggest a relationship. Repeated consistent measurements may strengthen confidence. An anomaly may weaken a pattern or reveal a method problem. A graph may show association without proving cause.

This guide develops confidence, uncertainty, anomalies, evidence strength and cautious scientific judgement for PSLE Science.

Return to the Primary 6 Science Learning Hub.

The confidence rule

CLAIM → EVIDENCE → CONSISTENCY → METHOD QUALITY → ALTERNATIVES → BOUNDARY → CONFIDENCE.

This is an eduKate reasoning routine, not an official SEAB marking formula.

Part I — Confidence is not certainty

Science often supports conclusions with different degrees of confidence.

High confidence does not mean “cannot be wrong”. Low confidence does not mean “useless”. Confidence depends on how well the evidence fits the claim.

Part II — What increases confidence?

• Repeated results show similar patterns.
• Relevant variables are controlled.
• The measured outcome matches the scientific question.
• The instrument is suitable.
• The sample is representative enough for the claim being made.
• Alternative explanations are reduced.
• The result is consistent with several independent observations.

Part III — What decreases confidence?

• One trial only.
• Large variation among repeated results.
• Two variables change together.
• The measurement is only a weak proxy for the intended process.
• The sample is very small.
• The conclusion extends far beyond the tested range.
• Anomalous results remain unexplained.

Part IV — An anomaly is a signal, not rubbish

An anomaly is a result that differs strongly from the rest of the pattern.

Example: distances of 79 cm, 81 cm, 80 cm and 43 cm under the same condition.

The 43 cm result deserves investigation. It should not be deleted automatically.

Part V — Possible reasons for anomalies

• Measurement error.
• Procedure not followed.
• Apparatus malfunction.
• Starting condition changed.
• Natural biological variation.
• Environmental condition shifted.
• The assumed pattern is incomplete.

Repeat and inspect before deciding.

Part VI — Repetition and spread

Results that cluster closely suggest good repeatability.

Example A: 80, 81, 79, 80 cm.

Example B: 55, 83, 60, 91 cm.

A is more consistent than B. That increases confidence in the typical value for A, assuming the design itself is valid.

Part VII — Reliability does not guarantee validity

A method can produce nearly identical results repeatedly and still measure the wrong thing.

Example: repeatedly measuring leaf length when the investigation asks for immediate gas production.

Consistency is valuable only when the method answers the right question.

Part VIII — Evidence can be direct or indirect

Direct evidence measures the target quantity closely.

Indirect evidence measures a related indicator.

Bubble count may be used as a proxy for gas production, but bubble sizes can vary.

Plant height may indicate growth, but it does not measure every aspect of plant health.

Indirect evidence can still be useful; its limits should be recognised.

Part IX — One observation versus a pattern

One point cannot establish a trend.

Three or more values can begin to reveal direction, but confidence also depends on variation and method quality.

Do not use “trend” when only one comparison exists unless the question provides enough structure.

Part X — Correlation versus cause

If two quantities change together, they are associated. Causal confidence becomes stronger when:

• one variable is deliberately changed;
• relevant alternatives are controlled;
• the mechanism is scientifically plausible;
• the pattern repeats.

Part XI — Confidence in predictions

Predictions inside the tested range are usually safer than predictions far beyond it.

Interpolation generally carries more confidence than distant extrapolation.

A plateau or turning point should reduce confidence in a simple linear prediction.

Part XII — Confidence language

Evidence situation	Useful wording
Strong direct comparison	supports, shows under these conditions
Reasonable indirect evidence	suggests, is consistent with
Prediction outside range	may, would be expected if the trend continues
Weak/confounded evidence	cannot conclude confidently, insufficient evidence

Part XIII — Original workshop: repeated car trials

Surface P: 80, 81, 79 cm.

Surface Q: 45, 47, 46 cm.

The groups are internally consistent and clearly separated. Confidence is relatively strong that the surfaces produced different travel distances under the tested conditions.

Original workshop: confounded plant trial

Plant A receives bright light and 100 mL water. Plant B receives dim light and 50 mL water. A grows more.

Confidence that light alone caused the difference is weak because water changed too.

Original workshop: anomalous cooling result

Three cooling trials for Cup Q show temperature decreases of 16°C, 17°C and 41°C.

The 41°C result is anomalous relative to the other two. Check starting temperature, timing, thermometer reading and setup before averaging.

Original workshop: field survey

More insects are found in shaded than open areas at one site.

The result suggests an association but does not prove shade alone caused the difference because moisture, plant cover and time of day may also vary.

Part XIV — Missing data creates uncertainty

If a variable was not measured, its value is unknown.

If a food-web arrow is absent, do not invent the relationship.

If a graph stops at 30 units, behaviour at 50 units is not directly observed.

Part XV — “No difference” can mean several things

No measured difference may mean:

• there is genuinely little difference;
• the instrument resolution is too coarse;
• the effect needs more time;
• sample size is too small;
• variation hides the effect.

Use the information in the question to decide which explanations remain plausible.

Part XVI — Contradictory evidence

If one new result conflicts with earlier data, do not ignore it.

Ask whether:

• the method changed;
• the condition lies outside the previous range;
• the earlier model was too simple;
• the result may be erroneous;
• a new mechanism or boundary has appeared.

Part XVII — Confidence should match the claim size

Small claim: “In these trials, Q produced shorter travel distances than P.”

Large claim: “All rough surfaces always produce greater friction than all smooth surfaces.”

The second claim needs far more evidence.

Part XVIII — Evidence hierarchy in a PSLE context

A useful classroom ordering is:

1. single observation;
2. repeated consistent observation;
3. controlled comparison;
4. controlled repeated comparison with suitable measurement;
5. multiple consistent lines of evidence.

This is a teaching structure, not an official scientific hierarchy.

Part XIX — The TRUST test

1. T — Trials: repeated enough to see consistency?
2. R — Relevant controls: alternatives controlled?
3. U — Uncertainty: anomalies, proxy limits or missing data?
4. S — Scope: does the claim stay within the evidence range?
5. T — Target measure: does the measurement answer the question?

This is an eduKate teaching mnemonic.

Part XX — MCQ confidence traps

• Choosing an absolute statement from one trial.
• Ignoring an anomaly.
• Assuming repeated results fix a confounded design.
• Treating an association as proof of cause.
• Extending a pattern beyond the graph.
• Assuming “no measured difference” means exact equality.

Part XXI — Open-ended evaluation structure

“Confidence in the conclusion is limited because [specific weakness]. This could affect [measured outcome or interpretation]. To strengthen the evidence, [targeted improvement].”

Use only if the question asks for evaluation or improvement.

Where to connect

• Models, Assumptions, Limits & Scientific Claims
• Confounding Variables, Controls & Experimental Validity
• Primary 4 Confidence, Uncertainty & Evidence

Retrieval checklist

• I distinguish confidence from certainty.
• I know what increases and decreases confidence.
• I investigate anomalies instead of deleting them automatically.
• I distinguish reliability from validity.
• I can identify direct and indirect evidence.
• I can distinguish correlation from cause.
• I match confidence language to evidence strength.
• I recognise uncertainty created by missing data.
• I keep claim size proportional to evidence.
• I can suggest a targeted improvement that strengthens evidence.

The quiet return

Scientific confidence is earned, not assumed. Strong evidence comes from suitable measurements, fair comparisons, repeated consistency and conclusions that stay inside the boundary of what was actually tested.

Check the method. Inspect the variation. Respect the anomaly. Limit the claim. Let confidence rise only when the evidence earns it.

Return to the Primary 6 Science Learning Hub.