Prediction questions test whether a pupil understands a relationship well enough to move it. The surface example changes. A condition is added, removed, increased, reversed or extended. The learner must decide what should happen next and explain why without turning a reasonable prediction into an unsupported certainty.
This guide develops prediction, what-if reasoning, extrapolation and transfer for Primary 6 and PSLE Science.
Return to the Primary 6 Science Learning Hub.
The prediction rule
IDENTIFY THE CHANGED CONDITION → RECALL THE RELATIONSHIP → CHECK THE RANGE → PREDICT THE DIRECTION → JUSTIFY THE LINK → QUALIFY IF THE EVIDENCE IS LIMITED.
This is an eduKate reasoning routine, not an official SEAB answer formula.
Part I — A prediction is not a guess
A scientific prediction uses a known relationship or pattern to estimate what will happen under a new condition.
Guess: “I think Plant B will grow taller.”
Prediction: “If Plant B receives more light while other relevant conditions remain comparable, it may show greater growth over the tested period because more light is available for the plant processes supported by light.”
The second answer names the condition, relationship and expected effect.
Part II — Prediction from a mechanism
Mechanism-based prediction uses Science rather than trend alone.
Example: if a spring is compressed farther within a safe tested range, it can exert a greater elastic effect on a toy, potentially producing greater motion on release.
The explanation should use the Primary-level relationship required by the question and avoid unsupported advanced detail.
Prediction from a trend
Trend-based prediction uses measured data.
If a graph shows output rising steadily from X = 1 to X = 4, a prediction at X = 5 may be reasonable if the question asks and if the relationship appears to continue.
But the prediction is weaker than a directly measured point.
Part III — Extrapolation has a boundary
Extrapolation means predicting outside the measured range. It can be useful, but confidence usually decreases as the new value moves farther from the evidence.
Measured spring loads: 1, 2, 3 units.
Predicting at 4 units may be modest extrapolation.
Predicting at 50 units assumes the same behaviour far beyond the evidence and may be physically unrealistic.
Interpolation is safer
Interpolation predicts within the measured range.
If results are known at 10 cm and 30 cm, estimating at 20 cm is within the tested span. The model may still be nonlinear, but the prediction does not travel beyond the boundaries.
Part IV — What-if reasoning changes one condition
Many questions ask what happens if one condition changes.
Use the counterfactual route:
- What was true originally?
- What is changed now?
- What relationship directly depends on that condition?
- What immediate effect follows?
- What downstream effect is the question asking for?
Original example: photosynthesis
Original: plant has sufficient water, carbon dioxide and light.
Change: light is greatly reduced.
Immediate effect: less light energy is available for photosynthesis.
Prediction: photosynthesis-related output should decrease under otherwise suitable conditions.
Do not jump directly to “the plant dies” unless the question extends over a suitable time period and provides enough evidence.
Original example: friction
Original: car rolls on a smooth surface.
Change: surface becomes rougher while release is kept comparable.
Prediction: greater frictional effect should slow the car more quickly, so it is expected to travel a shorter distance under the tested conditions.
Original example: environment
Original: predator has two prey sources.
Change: one prey decreases.
Prediction: predator population may be affected, but the second prey source can reduce the impact. An absolute decline is not guaranteed.
Part V — Reverse prediction
Sometimes the question gives the outcome and asks what condition could have changed.
Example: bubble count decreased.
Possible relevant conditions might include lower light availability or another limiting photosynthesis requirement, depending on the setup.
The correct answer must fit the evidence. Do not list every possible cause if the question gives information that narrows the options.
Part VI — Transfer means moving the relationship to a new surface
Transfer is stronger than memorisation because the pupil recognises the same relationship in an unfamiliar example.
Friction:
- shoe on floor;
- tyre on road;
- book sliding on table;
- toy car on different surfaces.
The nouns change. The contact-interaction relationship remains.
Transfer across representation
A relationship may appear as:
- paragraph;
- table;
- graph;
- diagram;
- sequence;
- apparatus setup.
A pupil has transferred the idea when they can identify it despite the representation change.
Part VII — Prediction needs controlled conditions
If several conditions change together, the predicted cause becomes less certain.
Example: Plant B receives less light and less water than Plant A. If B grows less, we cannot confidently assign the difference to light alone.
What-if reasoning is strongest when one relevant condition changes at a time.
Part VIII — Predictions can be conditional
Use “if”, “assuming” and “under the stated conditions” when appropriate.
Example: “If the same trend continues beyond the measured range, the spring extension would be expected to increase.”
This wording makes the assumption visible.
Part IX — Predicting a plateau
If a graph is already levelling off, the best prediction may be little or no additional increase rather than continuing the earlier steep trend.
Always use the most recent local relationship when extending a graph.
Turning points matter
If a graph rises and then falls, predicting from only the first half can be wrong. Identify where the new condition lies relative to the turning point.
Part X — Time-scale predictions
Immediate and long-term predictions can differ.
Removing a food source may have no visible effect in the first hour but affect a population over weeks.
Reducing light may affect photosynthesis quickly but growth changes appear later.
Use the time scale stated in the question.
Part XI — Original case study: solar fan
A solar fan rotates 20 times in 10 s at low light, 35 times at medium light and 48 times at high light.
If light increases slightly beyond the highest tested value, a cautious prediction is that rotations may increase if the measured relationship continues. But the exact number is uncertain because the system may approach a limit.
Original case study: cooling water
Water cools from 80°C to 65°C in five minutes and to 56°C after ten minutes.
A pupil predicts 41°C after fifteen minutes by subtracting another 15°C.
This may be weak because the cooling rate is already changing. A straight-line assumption is not supported by the two intervals.
Original case study: food web
A hawk eats rabbits and rats. Rabbit population falls sharply while rat population stays stable.
Prediction: hawks may shift feeding toward rats, so the hawk population does not necessarily fall immediately.
Prediction confidence depends on how much each prey contributes and other environmental conditions.
Part XII — “What would happen if nothing else changed?”
This phrase invites a controlled counterfactual. Hold the rest of the model fixed and change one factor.
This is often different from the real world, where many conditions co-vary. Follow the question’s model.
Part XIII — Transfer drills
Change the organism
Plant photosynthesis experiment → aquatic plant → leaf section → greenhouse plant.
Change the object
Sliding block → shoe sole → toy car → tyre.
Change the representation
Table → graph → diagram → paragraph.
Change the command
Describe → explain → predict → evaluate.
If the relationship survives all four changes, the concept is portable.
Part XIV — Prediction traps
- Assuming every trend is linear.
- Extending far beyond the data.
- Ignoring a plateau or turning point.
- Changing two variables but naming one cause.
- Using absolute language when alternatives exist.
- Ignoring the time scale.
- Predicting from a memorised story instead of the current setup.
Part XV — PSLE answer builder
“If [condition changes], then [measured or system outcome] is expected to [increase/decrease/remain similar] because [scientific relationship]. This prediction applies under the stated conditions / if the observed relationship continues.”
Use only as much of the frame as the question needs.
Where to connect
- Models, Assumptions, Limits & Scientific Claims
- Integrated Multi-Concept Reasoning & Unfamiliar Setups
- Primary 4 Prediction, What-If Reasoning & Transfer
Retrieval checklist
- I can distinguish a scientific prediction from a guess.
- I can predict from a mechanism or measured trend.
- I can distinguish interpolation from extrapolation.
- I can identify the boundary of the data.
- I can perform one-change what-if reasoning.
- I can make conditional predictions when evidence is limited.
- I can recognise plateaus and turning points.
- I can account for time scale.
- I can transfer the same relationship to a different surface example.
- I can reject predictions that depend on changed uncontrolled conditions.
The quiet return
Prediction is the test of whether a scientific relationship is alive in the learner’s mind. If the pupil can move the relationship to a new condition without losing its boundaries, the knowledge has become transferable.
Change one condition. Follow the relationship. Predict the direction. Respect the boundary. Test it somewhere new.
Return to the Primary 6 Science Learning Hub.
