“Choose the simplest explanation.”

That slogan is often presented as Ockham’s razor.

It is too crude.

The principle is not:

the simplest idea is always true.

Nature can be complicated.

A simple theory can be wrong.

A complicated theory can be correct.

The deeper principle is closer to this:

Do not multiply assumptions beyond what the evidence requires.

Simplicity matters because unnecessary structure weakens explanation.

But simplicity must compete with accuracy, scope, and evidence.

William of Ockham

William of Ockham was a medieval philosopher and theologian associated with a methodological preference for ontological economy.

The famous wording “entities should not be multiplied beyond necessity” is not a literal quotation from him in that familiar form.

But it captures the tradition associated with his name.

Do not posit more kinds of things than explanation requires.

Simplicity Is Not One Thing

What makes a theory simple?

Fewer entities?

Fewer parameters?

Shorter equations?

Fewer assumptions?

Less computational complexity?

Greater conceptual unity?

These can disagree.

There is no single universal measure of simplicity.

Ontological Simplicity

Ontological simplicity concerns how many kinds of entities a theory posits.

Suppose one theory explains planetary motion using gravity and ordinary matter.

Another explains the same observations using gravity, ordinary matter, and invisible spirits assigned separately to every planet.

If the spirits add no predictive value, the first theory is preferable.

The extra ontology does no work.

Theoretical Simplicity

A theory can also be simple in structure.

One general principle may replace many independent rules.

Newtonian gravity unified terrestrial falling and celestial motion.

Maxwell unified electricity, magnetism, and light.

Unification reduces independent assumptions.

This kind of simplicity is scientifically powerful.

Parameter Simplicity

Models with many adjustable parameters can fit almost any dataset.

A polynomial with enough terms can pass through every observed point.

But such a model may predict new data poorly.

Fewer free parameters often mean stronger constraints.

A theory that risks being wrong can earn more confidence when it succeeds.

Overfitting

Ockham’s razor connects naturally to overfitting.

An overly flexible model learns signal, noise, and accidental detail.

A simpler model may generalize better.

Modern statistics and machine learning formalize this intuition through regularization, information criteria, cross-validation, and complexity penalties.

But Underfitting Exists Too

Simplicity can go too far.

A straight line cannot model every nonlinear system.

A one-variable climate model cannot represent every regional process.

An excessively simple explanation omits real structure.

This is underfitting.

Good model selection balances simplicity and adequacy.

“As Simple as Possible”

A popular saying associated with Einstein is often paraphrased as:

“Everything should be made as simple as possible, but not simpler.”

Whatever the exact wording history, the idea captures the tension well.

Simplify until further simplification destroys explanatory adequacy.

Simplicity and Probability

Bayesian reasoning can help explain why simpler theories sometimes receive an advantage.

A highly flexible theory spreads probability across many possible outcomes.

A more constrained theory concentrates probability.

If the constrained theory predicts the observed data, the evidence can support it strongly.

The advantage comes from predictive concentration, not mystical preference for elegance.

Bayesian Occam Factors

In Bayesian model comparison, complex models can be penalized automatically when their extra parameter space does not improve prediction enough.

A model that can explain nearly anything gives lower specific probability to any one outcome.

This is sometimes called a Bayesian Occam factor.

Simplicity emerges from probability theory.

Minimum Description Length

Information theory provides another perspective.

The minimum description length principle favors models that compress data efficiently.

A good explanation captures regularity.

Instead of listing every observation separately, it provides a shorter generative description.

This connects simplicity to compression.

Compression Is Not Everything

A short description can still be wrong.

“Magic did it” is short.

It is not a useful scientific explanation because it lacks mechanism, constraint, and predictive content.

Simplicity must operate within empirical adequacy.

A compressed slogan is not automatically a theory.

Dark Matter and Ockham’s Razor

Dark matter is an instructive case.

At first, adding invisible matter seems less simple than modifying gravity.

But simplicity cannot be judged by counting words.

A dark-matter framework helps explain galaxy rotation, gravitational lensing, cluster dynamics, CMB structure, and large-scale structure.

A modified-gravity theory may be simpler at one scale but require additional structure elsewhere.

Global explanatory economy matters.

Neptune and Extra Entities

Uranus’s orbital anomalies led astronomers to posit Neptune.

Was adding an unseen planet a violation of Ockham’s razor?

No.

The new entity was required by the evidence and generated a precise prediction.

Ockham’s razor does not forbid new entities.

It forbids unnecessary ones.

Neutrinos

Pauli proposed the neutrino to explain missing energy and momentum in beta decay.

The particle was initially hypothetical and difficult to detect.

But it preserved conservation principles and generated testable consequences.

The neutrino was later observed.

Again, economy does not mean refusing invisible entities.

It means demanding explanatory work.

Epicycles

Ptolemaic astronomy is often caricatured as an absurd pile of epicycles.

The history is more nuanced.

Epicycles were mathematically sophisticated tools for representing planetary motion.

But as models accumulate adjustable corrections without deeper unification, they become less attractive.

The problem is not circular geometry itself.

It is explanatory and predictive economy.

Simplicity and Mechanism

A mechanism can make a theory look more complex while actually increasing explanatory coherence.

Suppose one statistical law is replaced by a molecular mechanism involving many components.

The micro-description is more detailed.

But it may unify many previously separate phenomena.

Simplicity depends on level.

Ad Hoc Assumptions

Ockham’s razor is especially useful against ad hoc rescue.

A theory fails.

We add a special exception.

Then another.

Then another.

If the additions produce no independent predictions, the theory becomes insulated from evidence.

Complexity grows without explanatory gain.

That is exactly what the razor warns against.

Auxiliary Hypotheses Are Not Automatically Bad

Real theories always use auxiliary assumptions.

Instrument models.

Boundary conditions.

Approximation schemes.

Background theories.

Adding an auxiliary hypothesis can be scientifically justified if it has independent support or generates new tests.

The razor should not be used mechanically to reject necessary complexity.

Elegance and Truth

Scientists often value elegant theories.

Symmetry.

Compact equations.

Unification.

Elegance has guided major discoveries.

But aesthetic preference can mislead.

A beautiful theory still requires evidence.

Nature may not share human taste.

Ockham’s razor is methodological discipline, not an oracle of truth.

Simplicity Depends on Language

A theory may look simple in one representational language and complicated in another.

A mathematical transformation can compress an ugly equation.

A coding language can make one pattern concise and another verbose.

This creates a deep philosophical issue:

simplicity is partly representation-dependent.

There may be no perfectly neutral language for measuring complexity.

Natural Variables

The right variables can transform complexity.

Planetary motion looks complicated in Earth-centered coordinates.

It becomes simpler in a heliocentric gravitational framework.

Thermodynamics becomes clearer using temperature and entropy.

Finding the right representation can make nature appear simpler because the representation aligns with real structure.

Simplicity as Discovery Heuristic

Scientists use simplicity partly as a search strategy.

The space of possible theories is enormous.

Prefer candidates with:

• fewer arbitrary assumptions,
• broader scope,
• stronger constraints,
• better unification.

This helps navigate theory space.

It does not prove the chosen theory true.

The Razor Can Cut Too Deep

Suppose evidence requires a complex mechanism.

Refusing it because “the simpler theory is better” becomes dogmatism.

Evolution is complex.

The immune system is complex.

Climate is complex.

Quantum field theory is complex.

Reality does not owe us minimalism.

The correct principle is economy constrained by evidence.

A Better Formulation

A useful modern version of Ockham’s razor is:

Among theories that explain and predict the evidence comparably well, prefer the one that achieves this with fewer unsupported assumptions and less unnecessary complexity.

This avoids the simplistic idea that shorter always means truer.

Model Selection

Modern science formalizes this tradeoff through tools such as:

• Akaike information criterion,
• Bayesian information criterion,
• cross-validation,
• regularization,
• Bayesian model comparison.

These methods differ mathematically and philosophically.

But they share one concern:

fit must be balanced against complexity.

The Next Problem

Ockham’s razor helps compare scientific explanations.

But before comparing them, we need to know which claims count as scientific at all.

Can any idea be made compatible with evidence?

Can science be separated from pseudoscience?

Is there one criterion that draws the boundary?

This is the demarcation problem.

What separates science from non-science?