What Is Science?
Published:
Science is often described as a method for discovering facts.
That is true but incomplete.
Science is also:
- a community,
- a set of practices,
- a system of error correction,
- a way of building models,
- a culture of measurement,
- a method for connecting claims to evidence.
There is no single five-step recipe that captures all science.
Astronomy does not work exactly like molecular biology.
Particle physics does not work exactly like geology.
Yet these fields share a family resemblance.
They make claims answerable to evidence.
Science Is Not a List of Facts
Facts matter.
But science is not simply a warehouse of correct statements.
Scientific knowledge changes.
Measurements improve.
Models are revised.
Old theories survive as approximations or are replaced.
A science textbook records the current state of understanding.
Science itself is the process that produced, tests, and revises that understanding.
The Myth of One Scientific Method
School diagrams often present:
- ask a question,
- form a hypothesis,
- perform an experiment,
- analyze data,
- draw a conclusion.
This is useful pedagogically.
Real research is less linear.
Scientists may begin with anomalous data.
They may construct instruments before having one precise hypothesis.
They may compare simulations.
They may infer inaccessible history from traces.
They may revise the question halfway through.
Scientific practice is iterative.
Observation
Science begins partly with observation.
But observation is not passive looking.
A telescope transforms invisible signals into data.
A particle detector reconstructs events from electronic traces.
A microscope uses optics, electronics, and computation.
An astronomical spectrum requires calibration.
Scientific observation is mediated by theory and instruments.
That does not make it arbitrary.
It makes methodology essential.
Measurement
Observation becomes more powerful when quantities can be measured.
Measurement creates comparisons.
Temperature.
Mass.
Time.
Distance.
Frequency.
Voltage.
Gene expression.
Uncertainty.
A measurement is never just a number.
It includes a procedure, a unit, an instrument, a calibration, and an error model.
Hypotheses
A hypothesis is a proposed claim that can be assessed against evidence.
Good hypotheses are not merely imaginable.
They should connect to observations in a way that risks being wrong.
The form of that risk varies by field.
Some hypotheses are tested through controlled experiments.
Others through natural experiments, surveys, historical records, or observational predictions.
Experiments
Experiments manipulate conditions to test causal relationships.
Control groups.
Randomization.
Blinding.
Repeated trials.
These methods help separate causal effects from confounding factors.
But not every science can manipulate its subject.
Astronomers cannot rerun a supernova.
Geologists cannot recreate continental drift at full scale.
Science therefore includes more than laboratory experimentation.
Natural Experiments
Nature often supplies variation that scientists did not create.
An eclipse.
A volcanic eruption.
A genetic mutation.
A policy change.
A meteorite impact.
Different environments.
Researchers can use these events to test models.
The distinction between experimental and observational science is real, but both can produce strong evidence.
Models
Science builds models.
A model is a simplified representation of some aspect of reality.
It may be:
- verbal,
- diagrammatic,
- mathematical,
- computational,
- physical.
Models omit detail deliberately.
A map that included every grain of sand would be useless as a map.
The question is not whether a model is complete.
It is whether it captures the structure relevant to the problem.
Theories
In everyday language, theory can mean guess.
In science, a theory is usually a structured explanatory framework supported by evidence and capable of connecting many observations.
Examples include:
- evolutionary theory,
- general relativity,
- quantum theory,
- germ theory.
A scientific theory is not automatically uncertain because it is called a theory.
Theory is often the highest level of explanatory organization.
Laws
Scientific laws describe stable relationships.
Newton’s laws.
Thermodynamic laws.
Conservation laws.
But laws and theories are not arranged on a ladder where theories eventually “become laws.”
A law may describe a regularity.
A theory may explain why that regularity occurs.
They play different roles.
Prediction
Prediction is a major source of scientific power.
A theory can forecast outcomes not yet observed.
Neptune was inferred from orbital anomalies.
Antimatter was predicted theoretically.
The cosmic microwave background was predicted before its discovery.
Gravitational waves were predicted long before direct detection.
Successful novel predictions increase confidence because the theory succeeds outside the data that originally motivated it.
Retrodiction
Science also predicts backward.
Cosmology reconstructs the early universe.
Evolutionary biology reconstructs common ancestry.
Geology reconstructs ancient environments.
Forensic science reconstructs past events.
These are retrodictions.
They are not weaker merely because the events already occurred.
The question is whether present evidence strongly constrains the past.
Explanation
Science does more than predict.
A model can predict tides without explaining them deeply.
A theory explains by connecting phenomena to mechanisms, laws, structures, or causes.
Different sciences use different forms of explanation.
A molecular mechanism in biology differs from a symmetry argument in physics.
Scientific explanation is plural.
Reproducibility and Replication
Science depends on results surviving independent scrutiny.
Reproducibility often means obtaining the same result from the same data and methods.
Replication means obtaining a compatible result from new data or repeated experiments.
Terminology varies by field.
Both matter because individual studies can be wrong.
Science gains strength through repeated independent confirmation.
Peer Review
Peer review is one mechanism of quality control.
Experts evaluate:
- methods,
- reasoning,
- evidence,
- novelty,
- presentation.
Peer review does not guarantee truth.
Bad papers can pass.
Good papers can be rejected.
Its value is procedural.
It creates structured criticism before and after publication.
Publication Is Not the End
Scientific claims remain open to challenge after publication.
Other researchers may:
- replicate,
- fail to replicate,
- find errors,
- propose alternatives,
- collect better data,
- reinterpret results.
Science is self-correcting only when institutions and researchers actually perform correction.
There is no automatic mechanism ensuring error disappears quickly.
Consensus
Scientific consensus is not truth by vote.
It is the convergent judgment of a community after evaluating evidence.
Consensus can be wrong.
But when a mature field contains large amounts of independent supporting evidence, expert consensus is usually a rational source of confidence for non-specialists.
Rejecting consensus requires stronger evidence than simply distrusting authority.
Expertise
Science is specialized.
No individual can personally verify every experiment in modern physics, biology, medicine, and chemistry.
We therefore rely partly on distributed expertise.
This creates a social dimension of knowledge.
Trust is unavoidable.
The challenge is deciding which institutions, methods, and experts deserve it.
Error Bars
Scientific numbers are rarely exact.
Measurements have uncertainty.
A result should often be expressed with:
- confidence intervals,
- standard errors,
- systematic uncertainties,
- model assumptions.
Uncertainty is not failure.
It is information.
A number without its uncertainty can be less scientific than an approximate range honestly reported.
Statistical Significance
Statistics helps evaluate whether patterns could arise from chance under a model.
But statistical significance is not the same as scientific importance.
A tiny effect can be statistically significant in a huge dataset.
A meaningful effect can fail a significance threshold in a small study.
Good science combines statistics with effect size, design quality, prior knowledge, and replication.
Correlation and Causation
Two quantities can vary together without one causing the other.
Ice-cream sales and sunburn both rise in summer.
Neither causes the other.
Temperature is a common cause.
Science needs experimental design, causal inference, mechanism, temporal structure, and other evidence to move from correlation toward causation.
This distinction is foundational.
Falsifiability
Karl Popper emphasized that scientific claims should be falsifiable in principle.
A theory that can explain every possible outcome risks explaining nothing.
Falsifiability captures something important:
scientific claims must expose themselves to evidence.
But real science is more complicated.
Individual hypotheses interact with auxiliary assumptions, instruments, background theory, and statistical interpretation.
No single criterion captures all scientific practice.
Science and Mathematics
Mathematics proves theorems from axioms.
Empirical science tests models against observation.
The relationship is intimate but different.
A mathematically consistent theory can still be physically wrong.
A beautiful equation does not become a law of nature until it survives empirical testing.
Mathematics gives science extraordinary expressive power.
Evidence decides physical applicability.
Science and Philosophy
Science cannot avoid philosophical assumptions entirely.
What counts as evidence?
What is a cause?
What is a law?
What does probability mean?
What makes an explanation good?
Philosophy of science studies these questions.
This does not place philosophy above science.
It examines the conceptual structure of scientific reasoning.
Science and Technology
Science and technology interact but are not identical.
Science seeks understanding.
Technology seeks effective intervention and construction.
A machine can work before its theory is complete.
A theory can be scientifically valuable before it has an application.
Historically, progress often flows in both directions.
Better tools enable better science.
Scientific understanding enables better tools.
Science Is Fallible
Science can be wrong.
Researchers can be biased.
Data can be poor.
Institutions can fail.
Fraud can occur.
The strength of science is not that scientists are uniquely objective people.
It is that scientific practices can expose errors:
- transparency,
- replication,
- competition among explanations,
- quantification,
- criticism,
- predictive testing.
Fallibility is built into the process.
Science Is Not Mere Opinion
Because science is fallible, some conclude that all claims are equally uncertain.
That does not follow.
A claim supported by thousands of independent measurements is not epistemically equivalent to an unsupported guess.
Fallibilism means knowledge can be revised.
It does not erase degrees of evidence.
A Working Definition
A useful working definition is:
Science is a systematic social practice for building, testing, and revising models of the natural world using observation, measurement, reasoning, and publicly assessable evidence.
This definition is broad enough to include diverse sciences while preserving what makes them scientific.
The Next Problem
Science depends on observation.
But observation has limits.
Our senses detect only narrow ranges.
Instruments transform signals.
Data can be noisy.
Theory influences what we choose to measure.
Some parts of the universe may be permanently inaccessible.
So before trusting observation naively, we need to ask:
What can observation actually reveal, and where does it fail?