How to Build a Simple Scientific Model From PSLE Science Evidence and Test What It Predicts

Wait, what? A Science question can give you a table full of results, a labelled diagram and three observations—and you can still have no useful model of what is happening.

That is because evidence and a model are not the same thing. Evidence tells you what was observed or measured. A scientific model is a deliberately simplified way of representing the important objects, relationships and conditions that can explain the evidence and help you reason about what should happen next.

Quick Answer

When a PSLE Science situation feels complicated, do not begin by hunting for a memorised sentence. Build a small model:

READ THE EVIDENCE → NAME THE SCIENTIFIC OBJECTS → IDENTIFY THE IMPORTANT RELATIONSHIP → KEEP THE CONDITIONS VISIBLE → REPRESENT THE MODEL SIMPLY → USE IT TO PREDICT → CHECK THE PREDICTION AGAINST NEW EVIDENCE → KEEP, LIMIT OR REVISE THE MODEL.

The model is useful only while it remains faithful to the evidence. If a new observation does not fit, the answer is not to force the observation into the old story. The answer is to ask which part of the model is incomplete, too broad or wrong.

The Exact PSLE Science Learning Job This Guide Owns

This guide owns one learner job: building a simple scientific model from PSLE Science evidence, using that model to predict, and checking whether the prediction survives further evidence.

It does not own the scientific concepts themselves. Existing pages about light, heat, electricity, forces, plants, water, body systems, materials and other scientific objects remain the concept owners. It also does not claim that every PSLE question asks students to draw or name a formal model. The model here is a reasoning tool: a compact representation that helps a learner move from evidence to an explanation or prediction without losing the important relationships.

Why This Belongs in Current PSLE Science

For examination from 2026, SEAB states that Standard PSLE Science assesses attainment in the 2023 Primary Science syllabus. The assessment objectives include knowledge with understanding and the application of knowledge and scientific inquiry. The inquiry component includes making predictions and formulating hypotheses, interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. These abilities can be expressed using words, diagrams, tables and graphs.

MOE’s 2023 Primary Science syllabus organises learning across the connected themes of Diversity, Cycles, Systems, Interactions and Energy. That connected structure is a strong reason to learn model-building rather than treating every chapter as a separate bag of facts. A useful model lets the same reasoning habit travel across different examples while the actual science changes.

Evidence Is Not Yet an Explanation

Suppose two identical cups begin with the same amount of water. Cup A is placed in a warmer location. Cup B is placed in a cooler location. After the same time, less water remains in Cup A.

The observations are:

• the starting amounts were the same;
• the locations differed in temperature;
• the elapsed time was the same;
• less water remained in Cup A.

Those statements are evidence. They do not automatically explain the result. The learner still has to identify a scientific relationship and mechanism that connects the changed condition to the measured outcome.

A simple model might be written as:

warmer condition → faster evaporation over the same time → more liquid water changes into water vapour → less liquid water remains

The arrows matter. They force the learner to expose the middle of the reasoning instead of jumping from “warmer” directly to “less water”.

A Scientific Model Is a Working Map, Not a Photograph of Reality

A model keeps what matters for the question and leaves other details out. A circuit diagram does not show every microscopic detail of the wire. A ray diagram does not reproduce all the physics of light. A food-web diagram does not show every organism or every interaction in an ecosystem. The simplification is not automatically a weakness. It is what makes a model usable.

The danger appears when a learner forgets that a model has a boundary. A model may be good enough to explain the evidence in one question and still be too simple to answer another. So every useful PSLE Science model should carry two quiet questions:

• What does this model help me explain or predict?
• What has this model left out?

The Seven-Step Model-Building Protocol

1. Read only what is actually given

List the observations, measurements, labels, stated conditions and relevant changes. Do not add a cause yet. This protects the evidence from being rewritten to fit the answer you hope is correct.

2. Name the scientific objects or parts

Ask what the question is really about: water, a plant, a circuit, a material, a moving object, a body system, a light source, a measured quantity, or several interacting parts. Keep object identity stable. Many wrong answers happen because a learner switches silently from the whole system to one part or from one specimen to another.

3. Identify the relationship that can connect the evidence

Do not ask only, “Which chapter is this?” Ask, “What is changing, what is affected, and how are they related?” The relationship is usually more useful than the topic label.

4. Keep the condition attached to the relationship

A scientific statement without its condition can become misleading. “It becomes faster” is incomplete if the question depends on higher temperature, a complete circuit, more exposed surface, a changed force, or another stated condition. The model should preserve the condition that makes the relationship relevant.

5. Represent the model in the smallest useful form

Use words, arrows, a quick diagram, a small table or another representation that exposes the relationships. The best form is not the most decorative one. It is the one that makes the next reasoning step easier and reduces ambiguity.

6. Make a prediction that the model commits to

A model becomes much more useful when it risks being wrong. Ask: if this model is correct and one relevant condition changes, what should we observe? A prediction turns the model from a description into a testable reasoning tool.

7. Compare the prediction with evidence and decide what to do

If the evidence fits, the model has survived this test. That does not make it universally true. If the evidence partly fits, the model may need a boundary. If the evidence contradicts a necessary prediction, inspect the evidence quality and then revise the model rather than protecting it by inventing excuses.

Worked Case 1: A Circuit That Looks Complete but Does Not Work

Imagine an original practice diagram showing a cell, wires and a bulb. The wire path appears to reach both sides of the bulb, but one connection stops just short of the metal terminal. The bulb does not light.

A weak response is: “The bulb does not light because there is no electricity.” That restates the outcome and introduces vague language.

A model-based learner first identifies the system relationship:

components connected in a complete conducting path → current can flow through the circuit → bulb can light under suitable conditions

Then the learner tests the diagram against that model. The gap means the required conducting path is incomplete. The model predicts that closing the gap with a conducting connection should allow the circuit to become complete. That prediction can then be checked in a changed diagram or practical set-up.

Worked Case 2: Two Graphs, One Relationship

Suppose one graph plots the amount of water remaining over time. Another graph from a different practice question plots the amount of water lost over time. The lines point in opposite visual directions.

A learner who copies graph shape may think the scientific relationships are opposite. A model-based learner asks what each axis represents. If water is leaving the container, “amount remaining” can decrease while “amount lost” increases. Both representations may fit the same underlying process model.

This is why a model should preserve quantity meaning rather than line direction alone. Representations can change while the scientific relationship remains stable.

Worked Case 3: A Plant Result That Supports More Than One Story

A plant in one set-up wilts. That observation is real, but by itself it may not identify one unique cause. Several scientifically plausible conditions could contribute.

The model-building mistake is to choose a familiar cause first and then pretend the wilting proves it. The stronger route is:

• record the observation: the plant wilted;
• identify the possible system relationships that could produce that outcome;
• look for additional given evidence that discriminates among them;
• if the evidence is insufficient, keep more than one possibility alive;
• propose a further observation or comparison that would test the competing models.

A model is not permission to turn possibility into certainty. It is a disciplined way to organise what the evidence supports.

Observation, Inference, Model and Prediction Are Different

Reasoning object	Main question	Example
Observation	What was directly seen or measured?	The thermometer reading increased.
Inference	What interpretation is supported by the evidence plus science knowledge?	The water gained heat under the stated conditions.
Model	What compact relationship explains or organises the observations?	Greater heat transfer into the water can raise its temperature while conditions allow.
Prediction	What should happen in a new stated condition if the model is useful?	Under otherwise comparable conditions, increasing the heating input for the same interval should lead to a different temperature outcome within the model’s limits.

Keeping these jobs separate prevents a common failure: writing a prediction as if it had already been observed, or writing an inference as if the instrument measured it directly.

Failure Signatures: How You Know the Model Is Weak

• You can repeat the facts but cannot say what affects what.
• Your explanation jumps straight from the starting condition to the final outcome.
• You change the scientific relationship when the diagram rotates or the graph axes change.
• You ignore one observation because it does not fit your preferred answer.
• You cannot predict what would happen if one relevant condition changed.
• Your model predicts everything, so no possible result could count against it.
• You keep adding details that are true but irrelevant to the question.
• You treat one successful prediction as proof that the model works in every situation.

Find the Earliest Weak Link

When a model fails, repair the earliest broken stage rather than rewriting the whole answer from memory.

1. Evidence failure: Did you misread a label, value, trend or condition?
2. Object failure: Did you confuse the whole system with one part or switch specimens?
3. Relationship failure: Did you choose the wrong scientific connection?
4. Condition failure: Did you drop a condition that decides whether the relationship applies?
5. Mechanism failure: Is there an unsupported jump in the causal chain?
6. Prediction failure: Did your stated prediction actually follow from the model?
7. Boundary failure: Did you stretch a useful model beyond the evidence or syllabus-level situation where it is appropriate?

This order matters. If the graph was misread at step one, polishing the wording at step six will not repair the science.

Misconception Repair

“A model must be a drawing.”

No. A useful model can be a causal arrow chain, a labelled sketch, a table of relationships, a graph, a set of linked statements or another compact representation. Choose the form that exposes the relationship needed for the question.

“If the model predicts one result correctly, it is proven.”

No. The model has survived that test. Other models may also fit the same result, and the relationship may have limits not tested by the evidence.

“More detail makes a better model.”

Not automatically. Extra detail can hide the important relationship. The target is the smallest model that is scientifically adequate for the learner job.

“The model tells me what the evidence must say.”

The direction is the opposite. Evidence constrains the model. If a well-obtained result does not fit, investigate the mismatch instead of rewriting the result.

A Four-Level Practice Sequence

Level 1: Build from visible evidence

Use a short original table or diagram. Write only the objects, relationship and condition. Then draw a three- or four-arrow model.

Level 2: Predict one changed condition

Change one relevant condition and state what the model predicts. Explain why. This checks whether the model is truly operational rather than memorised.

Level 3: Add evidence that does not fit perfectly

Introduce one result that is unexpected. Ask whether it reflects measurement quality, an omitted condition, a competing explanation or a genuine limit of the model. Do not assume “unexpected” means “wrong data”.

Level 4: Change the representation and surface story

Move from words to a graph, graph to table, diagram to prose, or one surface context to another while preserving the same reasoning job. If the learner can rebuild the relationship without a topic label, the model is becoming transferable.

The Unfamiliar-Transfer Test

After the learner succeeds with a familiar example, wait. On a later day, give a new original question whose objects look different but whose reasoning requires the same kind of evidence-to-model move. Remove the earlier arrows and prompts.

The learner should be able to:

• identify the evidence without being told the topic;
• name the relevant objects and relationship;
• build a compact model;
• make a prediction;
• say what observation would weaken that prediction;
• write a final explanation that still matches the question’s condition.

That is a stronger receipt of understanding than recognising a diagram seen yesterday.

Answer and Checking Receipt

Before accepting a model-based PSLE Science answer, run this short check:

1. Evidence: What exact observation, value, trend or condition am I using?
2. Object: Which part or system does it describe?
3. Concept: Which scientific relationship is relevant?
4. Mechanism: Have I shown the causal middle rather than only naming a concept?
5. Condition: Does the mechanism apply under the stated situation?
6. Outcome: Have I reached the outcome the question asks about?
7. Boundary: Did I claim anything the evidence does not support?

This is not an official marking template. It is a learning check that keeps reasoning disciplined.

Parent and Tutor Teaching Guide

When a learner is stuck, resist the urge to supply the final scientific sentence immediately. Ask for the smallest visible model first.

• “Show me only what the question actually gives.”
• “Which object is changing?”
• “What could connect this condition to that outcome?”
• “Can you draw that relationship with arrows?”
• “What would your model predict if this one condition changed?”
• “Which result would make you reconsider your model?”

If the learner can answer those questions but still writes poorly, the weak link may be scientific communication rather than conceptual reasoning. If the learner cannot build the arrow chain even orally, do not solve the problem by drilling final wording. Repair the missing relationship first.

Use one worked case, then a near-transfer case, then an unfamiliar case after a delay. The support should fade. The goal is not a student who can recognise the tutor’s model. The goal is a student who can build an adequate model when nobody has drawn it yet.

Useful Internal Routes

• How to Use a Scientific Model in PSLE Science Without Mistaking the Model for Reality
• How to Make a PSLE Science Prediction at the Right Precision
• How to Check Whether a PSLE Science Explanation Accounts for All the Important Observations
• How to Recognise the Same PSLE Science Concept When the Surface Example Changes

Authoritative References

• Singapore Examinations and Assessment Board — PSLE Formats Examined in 2026
• Ministry of Education, Singapore — Primary Science Teaching & Learning Syllabus 2023
• Agarwal, Nunes & Blunt (2021) — systematic review of retrieval practice in schools and classrooms, used here only as learning-science evidence for delayed retrieval and independent return, not as a PSLE examination rule.

The Quiet Return

A strong Science learner does not carry hundreds of disconnected answers. The learner carries relationships that can be rebuilt from evidence.

When the question changes, the diagram looks unfamiliar or the result surprises you, return to the same discipline: observe what is actually there, identify the scientific object and relationship, separate observation from inference, build the smallest model that explains the evidence, use it to predict, and let the next piece of evidence decide whether the model survives.

That is not a trick for one paper. It is one of the ways a student learns to think scientifically.