Primary 6 Science depends on knowing what kind of statement is being made. A pupil may look at a graph, table or experiment and write an inference when the question asks for an observation, write a prediction as though it were measured fact, or write a conclusion that exceeds the evidence.
This guide develops observation, inference, prediction and conclusion as four distinct scientific jobs for PSLE Science.
Return to the Primary 6 Science Learning Hub.
The four-job rule
OBSERVE WHAT IS GIVEN → INFER WHAT IT MAY MEAN → PREDICT WHAT MAY HAPPEN NEXT → CONCLUDE ONLY WHAT THE EVIDENCE SUPPORTS.
This is an eduKate reasoning routine, not an official SEAB formula.
Part I — Observation
An observation reports what is directly seen, measured, counted or recorded.
Examples:
- “The temperature decreased from 80°C to 55°C.”
- “The car travelled 42 cm on Surface Q.”
- “Plant A produced 36 bubbles in five minutes.”
- “The insect population decreased from 120 to 90.”
An observation does not require an explanation.
Part II — Inference
An inference uses observations plus scientific knowledge to interpret what may be happening.
Observation: the car travelled a shorter distance on Q than P under comparable release conditions.
Inference: Q may produce a greater frictional effect on the car.
Inference should remain accountable to the evidence.
Part III — Prediction
A prediction estimates what may happen under a new condition or later time using a pattern or scientific relationship.
Example: if an even rougher surface is tested under otherwise comparable conditions, the car may travel a shorter distance, provided the relationship continues.
Prediction is future- or unmeasured-condition reasoning, not observation.
Part IV — Conclusion
A conclusion answers the investigation question using the evidence collected.
Question: How does surface type affect travel distance?
Conclusion: under the tested conditions, the car travelled farther on Surface P than Q, supporting the conclusion that the surfaces produced different frictional effects.
A conclusion should match the design and tested range.
Part V — One setup, four jobs
Setup: an aquatic plant is placed at several lamp distances and bubble count is measured for five minutes.
Observation: bubble count decreased as lamp distance increased.
Inference: the greater distance may have reduced the light reaching the plant.
Prediction: if the lamp is moved still farther away, bubble count may decrease further if the trend continues.
Conclusion: within the tested range, greater lamp distance was associated with lower measured bubble production.
Part VI — Observation is not “what I know”
“Photosynthesis requires light” is scientific knowledge, not an observation from a particular experiment unless the setup directly establishes it through evidence.
Observation should come from the data or stated event.
Part VII — Inference is not imagination
A valid inference must be plausible and supported.
If a cold bottle develops water droplets outside, inferring condensation from water vapour in surrounding air fits the evidence and Science.
Inferring that the bottle material “creates water” does not.
Part VIII — Prediction must preserve the relationship
If a graph plateaus, predicting continued rapid increase ignores the latest evidence.
If a food web has alternative prey, predicting certain predator decline may be too strong.
Prediction quality depends on recognising boundaries.
Part IX — Conclusion is not universal law
One investigation may support a local conclusion.
“For these surfaces, the car travelled farther on P than Q.”
It does not automatically justify “All smooth surfaces always produce less friction than all rough surfaces.”
Part X — Observation and measurement
Measurements are quantitative observations.
Qualitative observation: “The leaf wilted.”
Quantitative observation: “The leaf length decreased from 8 cm to 6 cm.”
Both can be scientifically useful.
Part XI — Inference strength
Some inferences are strongly supported; others are tentative.
Useful language:
- “suggests”;
- “supports the idea that”;
- “is consistent with”;
- “may indicate”.
Use stronger language only when the evidence warrants it.
Part XII — Prediction from mechanism versus prediction from trend
Mechanism-based prediction uses a causal model.
Trend-based prediction uses observed pattern.
The strongest prediction may use both.
Part XIII — Conclusions from controlled experiments
Controlled experiments can support causal conclusions more strongly when one relevant variable changes and alternatives are controlled.
If light and water both change, the conclusion cannot isolate light alone.
Part XIV — Conclusions from observational studies
Field surveys often support association rather than strong causation.
If more insects are observed in shaded areas, shade may be associated with higher counts, but temperature, moisture and plant cover may also differ.
Part XV — Original workshop: cooling cups
Cup P cools from 80°C to 52°C. Cup Q cools from 80°C to 61°C in ten minutes.
Observation: Q has a higher final temperature and smaller temperature decrease.
Inference: Q’s covering may reduce heat transfer more effectively.
Prediction: if the same conditions continue, Q may remain warmer for longer.
Conclusion: under this test, Q reduced cooling more than P.
Original workshop: food web
Bird A eats insects X and Y. X decreases while Y remains stable.
Observation: X decreased; Y remained stable.
Inference: Bird A retains an alternative food source.
Prediction: Bird A may not decrease immediately.
Conclusion: the food web does not support a certain immediate decline in Bird A based on X alone.
Original workshop: spring
Extension values rise with load for loads 1–4.
Observation: extension increased over the tested loads.
Inference: larger loads produced greater extension in this spring over the tested range.
Prediction: load 5 may produce greater extension if the trend continues.
Conclusion: within the tested range, increasing load increased extension.
Part XVI — Common job-switching errors
- Explaining when asked to observe.
- Predicting when asked to conclude.
- Stating background knowledge as observation.
- Writing an inference as certainty.
- Writing a conclusion beyond the tested range.
- Restating data when asked for mechanism.
Part XVII — Command-word cues
Observe/describe: use data.
Infer: interpret data using Science.
Predict: extend to a new condition or future state.
Conclude: answer the investigation question using evidence.
Part XVIII — Evidence chain
A strong scientific response often follows:
observation → inference → scientific relationship → conclusion.
If the question asks for only one stage, do not write the whole chain unnecessarily.
Part XIX — The OIPC test
- O — Observation: what is directly given?
- I — Inference: what does the evidence suggest?
- P — Prediction: what may happen under a new condition?
- C — Conclusion: what relationship does the evidence support?
This is an eduKate teaching mnemonic.
Where to connect
- Data, Graphs, Diagrams & Evidence
- Prediction, What-If Reasoning, Extrapolation & Transfer
- Primary 5 Observation, Inference, Prediction & Conclusion
Retrieval checklist
- I can write an observation without explaining.
- I can infer without inventing unsupported stories.
- I can predict from a mechanism or trend.
- I can write a conclusion that answers the investigation question.
- I distinguish qualitative and quantitative observations.
- I match confidence language to evidence strength.
- I recognise that prediction is not measured fact.
- I keep conclusions inside the design and range.
- I can separate observation from background knowledge.
- I can identify which of the four jobs a question requires.
The quiet return
Scientific reasoning becomes cleaner when each statement has a job. Observation anchors the evidence. Inference interprets it. Prediction moves the relationship. Conclusion closes the investigation.
Observe first. Infer carefully. Predict conditionally. Conclude within the evidence.
Return to the Primary 6 Science Learning Hub.
