SCIENCE · EVIDENCE · MODELS · TESTING · UNCERTAINTY · REVISION
Science is the disciplined human system for building reliable explanations of the natural world through observation, measurement, evidence, models, testing and revision. It is not simply a collection of facts. It is a way of asking questions whose answers must survive contact with evidence.
Science begins with curiosity, but curiosity alone is not enough. A scientific claim must be connected to observations or measurements, stated clearly enough to be tested, compared against alternatives, and revised when the evidence no longer supports it. That combination—imagination constrained by reality—is what gives science its unusual power.
Science does not ask us to believe a claim because it sounds convincing. It asks what evidence would support it, what evidence could weaken it, and whether another explanation fits better.
This is the canonical public definition within the Science Learning Hub. It owns the broad question What is science? Specialist pages retain the narrower jobs of Primary Science, PSLE Science, Secondary Science, Physics, Chemistry, Biology, laboratory work, scientific vocabulary, evidence construction and examination preparation.
Contents
- 1. The shortest useful answer
- 2. Why science is difficult to define
- 3. A working definition of science
- 4. Science, facts, technology, engineering and opinion
- 5. What makes a question scientific?
- 6. Evidence: what science is built from
- 7. Observation, measurement and instruments
- 8. Models: how science represents what cannot be seen directly
- 9. Hypotheses, predictions and explanations
- 10. Experiments and controlled comparisons
- 11. Correlation, causation and natural experiments
- 12. Uncertainty, error and scientific confidence
- 13. Replication, peer criticism and self-correction
- 14. The main branches of science
- 15. The relationship between science and mathematics
- 16. Science, technology and engineering
- 17. Why civilisation needs science
- 18. What science does in education
- 19. How science learning works
- 20. Why students struggle with science
- 21. How science education fails
- 22. How to repair a science learning system
- 23. Science across school stages
- 24. Assessment and scientific explanation
- 25. Science in the age of artificial intelligence
- 26. Questions people ask about science
- 27. The final answer
1. The shortest useful answer
Science is a method for making explanations answerable to evidence.
In simple words, science asks:
- What is happening?
- How do we know?
- What pattern can we measure?
- What explanation could account for it?
- What prediction follows from that explanation?
- What observation would count against it?
- How uncertain are we?
- Can someone else repeat the work or inspect the evidence?
- Does a better explanation fit the evidence with fewer problems?
These questions reveal the central movement of science: observe → represent → explain → test → compare → revise. Science advances when claims become more constrained by what the world actually does.
2. Why science is difficult to define
Most students first meet science as content: plants, forces, cells, materials, heat, electricity, ecosystems and the Solar System. That is necessary, but it can make science look like a warehouse of correct facts. The warehouse matters. The deeper structure is the process by which claims entered it, remained there, were modified, or were removed.
Science is difficult to define because no single procedure is used in every discipline. A physicist may run a controlled laboratory experiment. An astronomer cannot move a star into a laboratory and may instead rely on observation, spectroscopy, modelling and prediction. An evolutionary biologist may combine fossils, genetics and comparative anatomy. A geologist may reconstruct an event from layers of rock. An epidemiologist may reason from populations where fully controlled experiments are impossible or unethical.
So science cannot be reduced to “the scientific method” as one rigid staircase. The deeper commonality is stronger: scientists make claims accountable to evidence, expose methods to criticism, quantify uncertainty where possible, and prefer explanations that survive increasingly demanding tests.
3. A working definition of science
Science is the organised, public and revisable practice of building explanatory models of the natural world from evidence, testing the consequences of those models against observation, and updating confidence when evidence changes.
Each part of this definition matters.
Organised
Science is more than personal curiosity. It uses shared methods, measurement systems, records, specialised instruments, statistical procedures, archives and communities of practice so that knowledge can accumulate beyond one person’s memory.
Public
A scientific result gains strength when the evidence and method can be inspected. Publication, data sharing, methodological detail, criticism and replication allow other people to test the route rather than merely trust the person who took it.
Revisable
Scientific knowledge can be strong without being sacred. A good model earns confidence because it explains evidence and survives testing. New evidence can narrow, extend or replace it. Revision is not a defect in science. It is one of the mechanisms that makes science dependable over time.
Models
Science rarely works with reality in its total complexity. It builds models: simplified representations of systems, relationships or mechanisms. A cell diagram, food web, atomic model, climate model and field equation are different forms of the same intellectual move—represent enough of reality to explain, predict or investigate.
Evidence
Evidence is not merely information that agrees with a claim. Its value depends on how it was produced, how directly it bears on the question, what alternatives it excludes, how reliable the measurement is and whether the data have been interpreted appropriately.
4. Science, facts, technology, engineering and opinion
| Term | Main job | Relationship to science |
|---|---|---|
| Scientific fact | A well-supported observation or measurement statement | An output or input of scientific work, not the whole system |
| Scientific theory | A broad explanatory framework supported by evidence | One of science’s strongest forms of explanation, not a casual guess |
| Hypothesis | A proposed explanation or relationship that can be investigated | A testable component within scientific inquiry |
| Technology | Tools and systems designed to achieve purposes | Often uses science and also creates instruments that extend science |
| Engineering | Design under constraints to make systems work | Uses scientific knowledge but is judged by function, safety and requirements |
| Mathematics | Precise reasoning about quantity, structure and relation | Provides languages, models and inference tools used throughout science |
| Opinion | A judgement or belief | Can motivate questions, but scientific acceptance requires evidence beyond preference |
These distinctions matter because a strong scientific explanation is not “just an opinion,” while a scientific theory is not merely an idea waiting to become a fact. The word theory in science refers to an explanatory framework capable of organising evidence and generating testable consequences.
5. What makes a question scientific?
A scientific question must connect to possible evidence. That does not mean every useful question is scientific. Questions about beauty, morality, meaning and justice may be profound while requiring forms of reasoning that science alone cannot settle.
- Observable connection: the question bears on events, objects, processes or measurements in the natural world.
- Operational clarity: important terms can be defined well enough for investigation.
- Evidence sensitivity: different possible observations would change our confidence.
- Comparison: more than one explanation or outcome can be considered.
- Reproducibility or traceability: the route from evidence to conclusion can be inspected.
The strongest scientific questions are often narrow enough to test yet connected to larger explanatory problems. “Why do plants grow?” is too broad for one experiment. “How does light intensity affect the rate of photosynthesis under these controlled conditions?” creates a tractable scientific job.
6. Evidence: what science is built from
Evidence is information used to support, weaken or discriminate between claims. It may come from direct observation, measurement, experiments, surveys, field studies, samples, images, spectra, fossils, genetic sequences, records or computational comparisons.
Not all evidence carries equal weight. Scientists ask:
- How was it collected?
- Was the sample representative?
- Was the instrument calibrated?
- Could bias enter the procedure?
- Is the measurement precise enough for the claim?
- Do alternative explanations remain?
- Does independent evidence point in the same direction?
One observation can matter enormously when it sharply distinguishes between explanations. Ten thousand weak observations may add less than one well-designed test. Scientific reasoning therefore asks not only how much evidence, but what kind of evidence.
7. Observation, measurement and instruments
Observation is the beginning of many scientific enquiries, but scientific observation is rarely passive looking. It is structured by questions, instruments and categories.
Human senses are limited. Instruments extend them. Microscopes reveal structures too small to see unaided. Telescopes collect faint radiation from distant objects. Spectrometers separate wavelengths. Sensors turn temperature, pressure, acceleration or chemical concentration into data. Sequencing machines reveal biological information encoded in molecules.
Every instrument introduces a measurement system with resolution, calibration, range and possible error. A reading is therefore never “just a number.” It is a number generated by a method under conditions. Understanding those conditions is part of scientific literacy.
8. Models: how science represents what cannot be seen directly
A scientific model is a simplified representation used to explain, predict or investigate a system. Models can be physical, conceptual, diagrammatic, mathematical or computational.
| Model | What it helps us do | What it leaves out |
|---|---|---|
| Particle model of matter | Explain states, diffusion and changes in arrangement | Many details of molecular structure and interaction |
| Food web | Represent feeding relationships | Full behaviour, seasonality and energy losses |
| Bohr-style atom diagram | Introduce electron shells and atomic organisation | The full quantum description |
| Climate model | Simulate coupled physical processes over time | Fine details below model resolution and uncertain future inputs |
| Mathematical disease model | Explore transmission patterns and interventions | Many individual differences and social complexities |
Models are useful because they are incomplete in deliberate ways. The key scientific question is not “Is this model the whole truth?” but “Is this model good enough for this explanatory or predictive job, and where does it break?”
9. Hypotheses, predictions and explanations
A hypothesis proposes a relationship or explanation that can be investigated. A prediction states what we expect to observe if the hypothesis and supporting assumptions are sufficiently accurate. An explanation connects mechanisms or relationships to the evidence.
These are not interchangeable. “The plant grew 4 cm” is an observation. “More light increased growth” is a relationship claim. “The additional light increased photosynthetic production under these conditions” is an explanatory claim. Each step requires more reasoning and usually more evidence.
Strong science makes the chain explicit: claim → expected consequence → measurement → comparison → update.
10. Experiments and controlled comparisons
Experiments are powerful because they deliberately change one aspect of a system while controlling or accounting for others. This can help separate correlation from causation.
- Independent variable: what is deliberately changed.
- Dependent variable: what is measured as an outcome.
- Control variables: relevant factors kept constant or otherwise accounted for.
- Control condition: a comparison baseline where appropriate.
- Replication: repeated measurements or experimental units used to estimate variability.
School experiments often simplify this structure so learners can see it clearly. Real research can be more complex: several variables may interact, controls may be statistical rather than physical, and ethical limits may prevent direct manipulation.
11. Correlation, causation and natural experiments
When two things change together, we say they are associated or correlated. That does not by itself prove that one causes the other. A third factor may affect both, the direction of influence may be reversed, or the pattern may be accidental.
Causal reasoning becomes stronger when evidence converges: controlled experiments, time ordering, plausible mechanism, dose-response patterns, natural experiments, repeated observations and the failure of competing explanations can all contribute.
This is why many sciences cannot be judged by whether they use a laboratory experiment. Astronomy, geology, ecology and parts of medicine often build causal understanding by combining many independent constraints from the world as it is.
12. Uncertainty, error and scientific confidence
Scientific uncertainty does not mean “we know nothing.” It describes the remaining range of plausible values, explanations or outcomes after considering the available evidence.
Sources of uncertainty include measurement limits, sampling variation, uncontrolled factors, incomplete models, future conditions and natural variability. Good science does not hide these. It measures or states them as clearly as the problem allows.
A result can be precise but wrong if the instrument is systematically biased. A result can be accurate on average but individually variable. A large sample can reduce random uncertainty while leaving a badly designed sampling process untouched. Scientific confidence depends on the whole evidence chain.
Uncertainty is not the opposite of knowledge. Often it is part of what responsible knowledge looks like.
13. Replication, peer criticism and self-correction
Science is a social knowledge system. Individual researchers can make mistakes. Instruments can fail. Analyses can be flawed. Expectations can bias interpretation. The system becomes stronger when claims are exposed to other people who have incentives and methods to find weaknesses.
Peer review is one filter, not a guarantee. Replication, independent data, re-analysis, methodological criticism and later research continue the test. Some claims become stronger. Others narrow. Some disappear. The long-run reliability of science depends less on infallible scientists than on institutions that make error discoverable.
14. The main branches of science
| Branch | Central focus | Examples |
|---|---|---|
| Physics | Matter, energy, motion, forces, fields, space and time | Mechanics, electricity, thermodynamics, optics, quantum physics |
| Chemistry | Composition, structure, properties and transformation of matter | Atomic structure, bonding, reactions, materials, biochemistry |
| Biology | Living systems and their organisation, function, evolution and ecology | Cells, genetics, physiology, evolution, ecosystems |
| Earth science | Earth systems and their history | Geology, meteorology, oceanography, climate science |
| Astronomy | Objects and processes beyond Earth | Planets, stars, galaxies, cosmology |
| Environmental science | Interactions among physical, chemical, biological and human systems | Pollution, ecosystems, climate, resources |
| Biomedical science | Mechanisms of health and disease | Physiology, microbiology, immunology, molecular medicine |
| Interdisciplinary science | Problems crossing traditional boundaries | Neuroscience, materials science, biophysics, geochemistry |
These disciplines differ in objects, instruments and methods, but they share the deeper demand that explanations answer to evidence.
15. The relationship between science and mathematics
Mathematics gives science languages for quantity, structure, change and uncertainty. It allows a scientific model to become precise enough to calculate consequences, compare observations and estimate confidence.
The two disciplines are close but not identical. Mathematics can prove consequences within defined systems. Science asks whether a mathematical structure adequately represents the natural world. A mathematical derivation may be flawless while the scientific model fails because its assumptions do not fit reality.
Read the companion canonical guide What Is Mathematics? for the full distinction.
16. Science, technology and engineering
Science asks what is true about the world. Engineering asks how to design a system that works under constraints. Technology is the collection of tools, processes and systems people create to achieve purposes. The three constantly exchange knowledge.
A microscope is technology that extends scientific observation. Knowledge of optics makes better microscopes possible. Better microscopes reveal new biological structures, creating new scientific questions. Engineering turns theory into reliable instruments. The loop continues.
Explore this wider relationship through the STEM Hub and How Science Connects Across STEM.
17. Why civilisation needs science
Science allows societies to build shared knowledge that can outlive individual experience. It turns observations into models, models into predictions, and predictions into better decisions when the evidence is strong enough.
17.1 Science extends perception
Instruments let civilisation detect worlds beyond ordinary human scale: cells, molecules, distant galaxies, seismic waves, electromagnetic radiation and genetic sequences.
17.2 Science stores causal knowledge
Understanding mechanisms allows societies to intervene more intelligently. Germ theory changed sanitation and medicine. Electromagnetism transformed communication and power. Materials science changed construction, transport and computing.
17.3 Science improves prediction
Forecasts of weather, disease risk, structural behaviour and environmental change are imperfect, but disciplined models often outperform intuition when systems are complex.
17.4 Science creates error-correction
A society that can measure outcomes and revise models has a better chance of discovering when a policy, treatment or technology is not working as expected.
Science does not decide every value question. It can estimate consequences, mechanisms and risks. People still decide what they value, what trade-offs are acceptable and how benefits and burdens should be distributed.
18. What science does in education
Science education teaches knowledge of the natural world, but its larger purpose is to develop evidence-based reasoning. Students learn to distinguish observation from inference, claim from evidence, model from reality and uncertainty from ignorance.
- Observe carefully.
- Classify by relevant properties.
- Measure with appropriate units and instruments.
- Represent systems with diagrams, tables, graphs and models.
- Identify variables and fair comparisons.
- Build explanations from evidence.
- Recognise uncertainty and limits.
- Revise explanations when evidence changes.
- Transfer ideas across unfamiliar contexts.
This places science inside the larger purpose of education: not merely remembering content, but building a capability that can survive new questions.
19. How science learning works
A strong science learning loop is:
- Observe: notice what the system is doing.
- Name: retrieve the scientific vocabulary and concepts that describe it.
- Represent: use diagrams, tables, graphs, particle models, cycles or equations.
- Relate: identify the mechanism or relationship linking cause and effect.
- Predict: state what should happen if the model is correct.
- Test: compare the prediction with evidence.
- Explain: connect claim, evidence and mechanism.
- Challenge: ask what alternative explanation or exception exists.
- Transfer: use the concept when the context changes.
- Revise: update the model or explanation after feedback.
Memorisation has a legitimate role. Terms, processes and core relationships must be retrievable enough to support reasoning. But recall without explanation is brittle. Explanation without retrieval is slow and unstable. Science education requires both.
20. Why students struggle with science
Low science performance can arise from very different problems that look identical on a marksheet.
- Vocabulary gap: the learner understands the event informally but cannot express it scientifically.
- Model gap: facts are memorised without an internal representation of the system.
- Evidence gap: the learner states a claim but does not use the data or observation that supports it.
- Mechanism gap: the learner names a phenomenon without explaining how or why it occurs.
- Graph gap: tables, axes, variables and trends are treated as a separate mathematics problem.
- Transfer gap: the learner succeeds in rehearsed examples but fails when the context changes.
- Question-reading gap: command words and constraints are missed.
- Answer-construction gap: scientific ideas are present but poorly organised into a complete explanation.
The right diagnosis is therefore more specific than “weak in science.” The first broken link might be knowledge, language, representation, evidence selection, causal reasoning or response construction.
21. How science education fails
21.1 Fact accumulation without mechanism
Students memorise that something happens but cannot explain the sequence of causes that produces it.
21.2 Keywords without relationships
Lists of “must-use words” create answers that sound scientific while the causal chain remains missing.
21.3 Experiments as recipes
The learner follows steps but does not understand the question, variable, comparison or reason for the control.
21.4 Notes without evidence practice
Reading and highlighting build familiarity but not the ability to retrieve, apply and explain under examination conditions.
21.5 Model diagrams treated as literal reality
A learner confuses a simplified representation with the complete physical system and struggles when a more advanced model is introduced.
21.6 Mark-scheme mimicry without transfer
Students memorise familiar response patterns that collapse when the data, organism, apparatus or context changes.
22. How to repair a science learning system
| Observed problem | Likely missing component | Better next move |
|---|---|---|
| Knows notes but cannot answer open-ended questions | Evidence-to-explanation construction | Build claim → evidence → mechanism chains from new data |
| Cannot interpret graphs | Variable and representation control | Name axes, units, trend and scientific meaning separately |
| Uses keywords but loses marks | Relationships are missing | Require causal connectors: because, therefore, leading to, as a result |
| Confuses fair tests | Experimental logic is weak | Identify the question, changed variable, measured outcome and necessary controls |
| Forgets topics quickly | Retrieval is weak | Use short spaced recall linked to diagrams and mechanisms |
| Fails unfamiliar contexts | Concept is bound to one surface form | Vary organisms, apparatus and contexts while preserving the same mechanism |
| Gives overconfident conclusions | Uncertainty and evidence limits are missing | Ask what the data justify and what remains unknown |
Repair is successful when the learner can use the same scientific relationship under a changed surface, explain why the evidence supports the claim, and recognise where the explanation no longer applies.
23. Science across school stages
23.1 Primary foundations
Before formal science becomes dense, children learn to observe, compare, classify, measure, identify patterns and explain simple relationships. These are not preliminary decorations. They are the cognitive roots of scientific thinking.
Use Primary 1 Science Foundations and Primary 2 Science Foundations for the earliest routes.
23.2 Primary 3 to PSLE
Formal Primary Science requires content knowledge, process skills, scientific vocabulary and increasingly precise explanation. The learner must connect observations to concepts and concepts to mechanisms rather than merely list memorised facts.
For examination routes, use the Science Learning Hub and the PSLE Science routes within it.
23.3 Secondary Science
Secondary Science increases model density, mathematical representation, laboratory reasoning and abstraction. Physics, Chemistry and Biology become more distinct while remaining connected by evidence, measurement and systems thinking.
The current Singapore SEC pathways are mapped in SEC Science Pathways Singapore.
23.4 Advanced and research science
At higher levels, students encounter greater mathematical formalism, specialised instrumentation, uncertainty, literature, experimental design and the expectation that claims be situated within existing research. The focus moves from learning established explanations toward producing or evaluating new knowledge.
24. Assessment and scientific explanation
A good science assessment samples more than recall. It asks whether a learner can:
- retrieve relevant concepts and terminology;
- interpret diagrams, tables, graphs and experimental information;
- identify variables and fair comparisons;
- use evidence from the question rather than generic memory alone;
- explain mechanisms and causal chains;
- evaluate claims and procedures;
- recognise uncertainty and limitations;
- transfer concepts to unfamiliar situations.
A mark therefore provides evidence about performance, not a complete diagnosis of scientific understanding. The working, explanation and error pattern usually tell us more about what needs repair.
25. Science in the age of artificial intelligence
Artificial intelligence can search literature, propose hypotheses, analyse images, fit models, generate code and summarise large bodies of information. These capabilities can accelerate science, but they do not remove the core scientific responsibilities.
25.1 Fluent answers are not evidence
An AI system can produce a convincing explanation that contains a fabricated source, incorrect mechanism or hidden assumption. Scientific reliability still depends on traceable evidence and valid methods.
25.2 AI can widen the hypothesis space
Machine systems can detect patterns across volumes of data too large for unaided inspection and suggest relationships worth investigating. A pattern becomes scientific knowledge only after appropriate validation.
25.3 Models still need boundaries
An AI model can perform well on familiar data and fail when conditions shift. Scientists must ask what population, instrument, environment and assumptions the model represents.
25.4 Human scientific judgement remains essential
People still define the question, decide which measurements matter, design ethically acceptable tests, inspect causal claims, interpret uncertainty and decide whether an answer is fit for consequential use.
26. Questions people ask about science
What is science in simple words?
Science is a way of learning about the natural world by making observations, gathering evidence, building explanations and testing whether those explanations continue to fit.
Is science just facts?
No. Scientific facts matter, but science also includes the methods, models, measurements, arguments and revision processes by which reliable knowledge is built.
What is the scientific method?
There is no single rigid method used by every scientific field. A common pattern is to observe, ask a question, propose an explanation, derive a prediction, gather evidence, compare results and revise the model.
What is a scientific theory?
A scientific theory is a broad explanatory framework supported by substantial evidence and capable of organising observations and generating testable predictions. It is not merely a guess.
Can science prove something with absolute certainty?
Science usually builds degrees of confidence rather than mathematical proof about the natural world. Evidence can become extremely strong, but scientific explanations remain open to revision if genuinely better evidence appears.
Why do scientific explanations change?
They change when new evidence, better instruments, improved models or stronger analyses reveal that an earlier explanation was incomplete or wrong. This is a feature of scientific self-correction.
What is the difference between science and mathematics?
Mathematics determines what follows from defined assumptions and structures. Science uses observation and measurement to determine whether models adequately explain the natural world.
What is the difference between science and engineering?
Science aims to understand how the world works. Engineering aims to design systems that work under stated requirements and constraints. Each informs the other.
Why are experiments important?
Experiments can isolate variables and create controlled comparisons, making causal explanations easier to test. They are powerful but not the only valid form of scientific evidence.
Why is uncertainty part of science?
Measurements have limits, samples vary and models simplify. Stating uncertainty makes the evidence more honest and helps people judge what conclusions are justified.
27. The final answer
Science is not a finished book of truths. It is a disciplined system for making our explanations increasingly answerable to the world.
It begins with observation and measurement. It builds models that compress complexity into inspectable relationships. It generates hypotheses and predictions. It tests those predictions against evidence. It compares explanations, measures uncertainty, invites criticism and revises when reality refuses to cooperate.
For learners, science develops the ability to move from seeing to explaining, from evidence to claim, from model to prediction, and from confidence to justified confidence. For civilisation, science creates a collective memory of tested explanations that can guide medicine, engineering, agriculture, energy, technology, environmental management and almost every system in which misunderstanding the natural world carries a cost.
Science is the human practice of making explanations accountable to evidence, so that what we think about the world can be corrected by what the world actually does.
Canonical article record
| Article | What Is Science? |
|---|---|
| Question owned | The broad definition, evidence architecture, modelling logic, scientific reasoning, educational meaning and civilisational significance of science |
| Canonical parent | Science Learning Hub |
| Primary wider route | STEM Hub |
| Audience | Students, parents, teachers, educators and readers seeking a complete public definition |
| Published | 4 September 2026 |