The Demarcation Problem

What separates science from non-science?

At first this seems easy.

Science uses evidence.

Science performs experiments.

Science makes predictions.

Science is falsifiable.

But each answer quickly runs into exceptions.

Astronomy is scientific even though astronomers cannot manipulate stars.

Historical sciences reconstruct events that cannot be repeated.

Some scientific theories are difficult to test directly.

Some pseudosciences collect data, use technical vocabulary, and make predictions.

The boundary is real enough to matter.

It is difficult to define with one simple rule.

This is the demarcation problem.

Why Demarcation Matters

The question is not merely academic.

It affects:

• education,
• medicine,
• public policy,
• funding,
• legal testimony,
• journalism,
• public trust.

Calling something scientific grants epistemic authority.

Calling it pseudoscientific can remove that authority.

We therefore need standards strong enough to discriminate without pretending science follows one rigid template.

Science Is Not Defined by Subject Matter

Astronomy studies stars.

Astrology also talks about stars.

Medicine studies health.

So do many unvalidated healing systems.

Physics studies energy.

Pseudoscientific claims often borrow the word “energy.”

The subject does not determine whether an activity is scientific.

Method and evidential behavior matter more than vocabulary.

Science Is Not Defined by Laboratories

Many sciences use laboratories.

But some do not.

Astronomy, paleontology, geology, and parts of ecology rely heavily on observation.

Researchers cannot rerun the formation of Earth.

They cannot experimentally restart evolution from the Cambrian period.

Science must therefore include reliable inference from naturally occurring evidence.

Science Is Not Simply “Using Mathematics”

Astrology can include charts.

Financial scams can include equations.

Pseudoscientific reports may contain statistical tables.

Mathematical appearance does not create scientific legitimacy.

The mathematics must connect to well-defined quantities, justified assumptions, and empirical tests.

Symbols can disguise weak reasoning as easily as they can sharpen strong reasoning.

Science Is Not Whatever Scientists Say

Institutional membership matters, but it is not enough.

Scientists can make non-scientific claims.

A Nobel Prize does not transform every opinion of its recipient into science.

Likewise, a correct claim can originate outside formal scientific institutions.

Authority is relevant evidence about expertise.

It is not the definition of scientific status.

The Verificationist Idea

Logical positivists emphasized verification.

A meaningful empirical claim should be connected to possible observations that could verify it.

This seemed to offer a clean boundary.

But strict verification runs into trouble.

Universal laws cannot be conclusively verified from finite observations.

No number of observed white swans logically proves that all swans are white.

Scientific theories often reach beyond directly observed cases.

Popper’s Falsifiability

Karl Popper proposed a famous alternative.

Scientific theories should be falsifiable.

They should make predictions that could, in principle, turn out false.

“All swans are white” is falsifiable because one black swan would conflict with it.

A claim compatible with every possible observation cannot be tested strongly.

Falsifiability captures a core scientific virtue:

risk.

But Falsifiability Is Not Sufficient

Suppose someone predicts:

“Tomorrow a coin will land heads.”

The statement is falsifiable.

But one falsifiable claim is not automatically science.

Science also requires:

• systematic method,
• evidential context,
• reproducibility,
• theoretical structure,
• openness to revision.

Falsifiability is important.

It does not define science by itself.

Falsifiability Is Not Always Simple

Scientific predictions depend on auxiliary assumptions.

If an experiment fails, perhaps:

• the theory is wrong,
• the detector malfunctioned,
• the sample was contaminated,
• initial conditions were misestimated.

No isolated observation necessarily points to one proposition alone.

This is one reason real scientific falsification is more complicated than textbook logic.

Kuhn’s Challenge

Thomas Kuhn emphasized that mature sciences operate within paradigms.

Scientists share:

• exemplary problems,
• methods,
• standards,
• background assumptions.

Most research is not an attempt to falsify the paradigm every morning.

It is normal science:

solving puzzles within a framework.

Anomalies can accumulate without immediate theory abandonment.

Scientific change is historically structured.

Lakatos and Research Programmes

Imre Lakatos tried to combine insights from Popper and Kuhn.

He described research programmes containing a relatively stable theoretical core and a changing belt of auxiliary hypotheses.

A programme can be progressive if modifications generate successful new predictions.

It can become degenerating if changes merely protect it from failure after the fact.

This shifts demarcation from isolated theories to patterns of development over time.

Scientific Practice Is Clustered

Perhaps science is identified not by one criterion but by a cluster of features.

Scientific work tends to involve many of the following:

• empirical evidence,
• testability,
• quantitative precision,
• reproducibility,
• explanatory coherence,
• predictive risk,
• methodological transparency,
• error correction,
• peer criticism,
• cumulative progress.

No one feature may be necessary in every case.

Together, they form a recognizable pattern.

Pseudoscience

Pseudoscience imitates the appearance of science without reliably adopting its error-correcting practices.

Common warning signs include:

• vague predictions,
• selective evidence,
• immunization against criticism,
• reliance on anecdotes,
• moving goalposts,
• misuse of technical language,
• refusal to specify failure conditions,
• conspiracy explanations for every contrary result.

One sign alone is not decisive.

The overall pattern matters.

Unfalsifiable Escape Routes

Consider a claim that survives every outcome.

If the prediction succeeds:

the theory is confirmed.

If it fails:

hidden forces interfered.

If critics object:

their skepticism proves the theory threatens established interests.

Such a system converts every possible event into support.

It is epistemically closed.

Science instead needs possible observations that would count against a claim.

Ad Hoc Immunization

Auxiliary hypotheses are normal in science.

But a pattern of endless ad hoc rescue is suspicious.

A good modification should do more than explain away failure.

It should ideally:

• have independent support,
• improve predictive precision,
• create new tests.

A theory that only retreats after every failed prediction loses scientific strength.

Replication

Science expects claims to survive independent checking.

A result depending on one charismatic practitioner, one secret instrument, or one unrepeatable demonstration is weak.

Replication is not always literal repetition.

In astronomy, independent telescopes may observe the same phenomenon.

In historical science, different lines of evidence may converge.

The principle is public checkability.

Transparency

Methods should be inspectable.

How were data collected?

Which cases were excluded?

What statistical tests were used?

Which assumptions were made?

Secret methods weaken scientific credibility because other researchers cannot evaluate the evidential chain.

Transparency exposes claims to correction.

Cumulative Knowledge

Science tends to build on itself.

Later theories explain why earlier theories worked in limited domains.

New measurements refine constants.

Data accumulate.

Methods improve.

Pseudoscientific systems often show much weaker cumulative progress.

The same vague claims can persist for decades without increasing precision.

Prediction and Retrodiction

A good scientific framework should constrain observations.

It may predict future events.

Or it may retrodict traces of past events.

Evolutionary theory predicts genetic and fossil patterns.

Cosmology predicts relics of the early universe.

The important feature is not time direction.

It is the existence of risky empirical consequences.

Mechanism

Mechanistic understanding often strengthens scientific status.

A treatment may first show statistical efficacy.

Later research identifies receptors, pathways, and cellular processes.

A claim becomes integrated into wider scientific knowledge.

Pseudoscientific explanations often stop at labels:

“energy imbalance,” “vibration,” “toxins”

without operational definitions or measurable mechanisms.

Integration with Other Knowledge

Scientific theories do not exist in isolation.

They should fit reasonably with well-supported knowledge from other fields unless they provide strong evidence for revision.

A new biological claim that violates chemistry requires extraordinary support.

A new physical theory that contradicts relativity must explain why relativity works so well in tested domains.

Coherence does not forbid revolution.

It raises the evidential burden.

Error Correction

Science is not defined by scientists never being wrong.

It is defined partly by institutions and methods that can reveal error.

Retractions.

Replication.

Reanalysis.

Competing laboratories.

Improved instruments.

Open data.

Science progresses because claims remain vulnerable.

An error-correcting system can outperform individually brilliant but uncheckable intuition.

Boundary Cases

Some fields sit near the boundary.

Early-stage research may have limited evidence.

Highly theoretical physics may make predictions beyond current technology.

Historical hypotheses may depend on incomplete traces.

This does not make them pseudoscience automatically.

Scientific status can depend on whether the work remains connected to possible evidence and disciplined methodology.

String Theory as a Boundary Debate

String theory is sometimes invoked in demarcation debates because direct tests at relevant scales are difficult.

It has deep mathematical structure and connections to established physics.

Critics worry about limited unique empirical predictions.

Supporters point to indirect theoretical successes and the possibility of future tests.

The debate shows why demarcation is not always binary.

Research programmes can be scientifically serious while empirically underconstrained.

SETI as Another Boundary Case

The search for extraterrestrial intelligence studies a phenomenon not yet confirmed to exist.

Yet SETI can be scientific because it defines signals, detection protocols, false-positive checks, and observational methods.

Science does not require the target phenomenon to be known beforehand.

It requires disciplined ways of finding out.

Science and Metaphysics

Some questions may be meaningful but not empirically testable.

Why is there something rather than nothing?

Are mathematical objects real independently of minds?

Does all reality require a sufficient reason?

These are philosophical questions.

Calling them non-scientific does not mean meaningless.

Demarcation is not a ranking of human worth.

It identifies methods and evidential scope.

Science and Technology

A technique can work without a complete scientific theory.

Early metallurgy worked before atomic theory.

Engineering often uses empirical rules.

Practical success alone does not prove the accompanying explanation correct.

A treatment may have an effect while its proposed mechanism is wrong.

Efficacy and theoretical truth should be tested separately.

Pseudoscience Can Contain True Statements

A pseudoscientific system may include some correct observations.

Astrology uses real planetary positions.

Alternative-health systems may contain genuinely active substances.

The issue is not whether every sentence is false.

The issue is whether the system reliably distinguishes truth from error.

Methods matter because accidental truth is easy.

The Demarcation Problem Has No Magic Test

No universally accepted one-line criterion separates all science from all non-science.

That is not failure.

Complex categories often have multiple dimensions.

A robust judgment asks:

• Is the claim testable?
• Does it risk failure?
• Are methods transparent?
• Are data reproducible?
• Does evidence accumulate?
• Are alternatives considered?
• Are failures acknowledged?
• Does the programme generate new knowledge?

Demarcation is methodological diagnosis.

A Working Principle

A useful practical formulation is:

Science is distinguished less by one defining rule than by sustained openness to empirical constraint, public criticism, reproducible method, and correction.

This captures the direction of scientific practice without pretending every field looks identical.

The Next Problem

There is another threat to scientific reasoning.

Some movements do not merely propose weak scientific claims.

They attack the legitimacy of scientific methods themselves while selectively borrowing scientific language when convenient.

That raises a different question:

What is anti-science, and how is it different from legitimate criticism of science?