A doctor sees a patient with a cough, a fever, and fatigue. There are two ways to explain it: one illness causing all three symptoms, or three unrelated problems that happen to be striking at once. Most of the time, doctors are trained to look for the first explanation before reaching for the second. That instinct has a name, a 700-year history, and a surprising amount of influence over how astronomers pick between competing models of the universe, how machine learning engineers stop an algorithm from overfitting, and how scientists across every field decide which of two rival theories deserves more attention. It’s called Occam’s razor, and it may be the most widely used, least understood idea in science.

What Is Occam’s Razor?

Occam’s razor is the principle that, when two or more explanations account for the same evidence equally well, the simpler explanation — the one that makes the fewest new assumptions — should generally be preferred. It is sometimes stated as “entities should not be multiplied beyond necessity,” a Latin formulation historians have long associated with the medieval friar the idea is named after, even though he never wrote that exact phrase himself, according to Encyclopædia Britannica.

The word “razor” is doing real work in the name: the principle is a tool for cutting away unnecessary assumptions, not for shrinking an explanation until it stops fitting the facts. Occam’s razor does not say “the simplest answer is always correct.” It says that simplicity is a legitimate reason to prefer one theory over another, once both are equally consistent with the evidence.

Where Does Occam’s Razor Come From?

The principle is named for William of Ockham, an English Franciscan friar and philosopher who lived from roughly 1287 to 1347. Britannica credits Ockham with the maxim pluralitas non est ponenda sine necessitate — “plurality should not be posited without necessity” — which he used constantly in his theological and philosophical arguments, stripping away what he considered unnecessary categories and assumptions in the thinking of his contemporaries.

Ockham didn’t invent the underlying idea. The Stanford Encyclopedia of Philosophy traces the preference for simple explanations back to Aristotle, who argued that a demonstration built on fewer assumptions is superior to one built on more, all else being equal. What made the principle “Ockham’s” was the frequency and force with which he applied it — enough that later scholars started reaching for his name whenever they needed to justify cutting something out.

How Does Occam’s Razor Work in Science?

In practice, Occam’s razor rarely settles a scientific question by itself. Its real job is narrowing the field between competing theories that already fit the observed evidence.

The clearest historical example is the centuries-long argument over whether the Earth or the Sun sits at the center of the solar system. Both the old Earth-centered model and the new Sun-centered model could, with enough patches and add-ons, predict where the planets would appear in the sky. But the Earth-centered model needed a tangle of extra devices — epicycles stacked on epicycles — to keep matching what astronomers observed. Copernicus’s Sun-centered model explained the same planetary motion with a far simpler structure. That gap in simplicity, more than any single new observation, is a large part of why the heliocentric model eventually won out — a story Specialty Digest examined in more depth in The Copernican Principle.

The same logic runs through the everyday mechanics of the scientific method: when two hypotheses both explain an experiment’s results, researchers don’t get to declare a winner. They design a new test that would separate the two — and until then, Occam’s razor offers a reasonable, if provisional, way to decide which hypothesis to investigate first.

Is Occam’s Razor Always Right?

No — and treating it as an ironclad law of nature is one of the most common misreadings of the principle. Nature is not obligated to be simple. Some of the best-confirmed theories in physics, including general relativity and quantum mechanics, are mathematically intricate, not minimal. Simplicity is a tiebreaker between theories that already explain the evidence equally well; it is not evidence in itself, and it cannot rescue a theory that the data actually contradicts.

Medicine offers the sharpest illustration of where the razor’s limits show up. Physicians are taught a rule of thumb often summarized as “when you hear hoofbeats, think horses, not zebras” — look for one diagnosis that explains all the symptoms before assuming several rare diseases are striking at once. But an aphorism attributed to physician John Hickam pushes back: “a man can have as many diseases as he damn well pleases.” A 2015 case report published in Thorax, the peer-reviewed journal of the British Thoracic Society, illustrates exactly this tension: two rare respiratory conditions occurring in the same patient at once, a case the authors explicitly frame as Occam’s razor versus Hickam’s dictum. As the case shows, insisting on a single unifying diagnosis can sometimes cause a doctor to miss a second, coexisting illness — particularly in older patients or those with several chronic conditions (source: PMC/Thorax, 2015). Simplicity is a starting assumption clinicians are trained to test, not a conclusion they’re allowed to stop at.

How Do Scientists — and Machines — Use Occam’s Razor Today?

The principle has outlived medieval philosophy by finding new, very concrete jobs.

In statistics and model selection, researchers comparing two mathematical models that fit a dataset about equally well will typically favor the one with fewer parameters, on the theory that a model with more moving parts is more likely to be fitting noise rather than a real underlying pattern.

In machine learning, this shows up as a core design principle rather than a philosophical aside. According to Google’s Machine Learning Crash Course, simpler models tend to generalize better to new data than complex ones, even when the complex model scores slightly higher on the data it was trained on. Techniques called regularization exist specifically to penalize unnecessary model complexity during training, pushing algorithms toward simpler solutions — a direct, mathematically formalized descendant of a 700-year-old philosophical heuristic. It’s a useful caution to keep in mind alongside more ambitious claims about what artificial intelligence can do: even cutting-edge systems are, in part, built on a medieval friar’s instinct that simpler is usually safer.

In everyday scientific practice, the razor functions less as a rule and more as a discipline: a reminder to ask, before adding a new variable, a new cause, or a new hidden factor to an explanation, whether the evidence actually requires it — one reason the replication crisis in some fields is, in part, a story about researchers finding complex explanations for what turned out to be statistical noise.

Evidence Strength

Well-supported: That Occam’s razor functions as a practical heuristic across science, statistics, machine learning, and clinical medicine is well documented and largely uncontroversial among practitioners in those fields.

Contested: Why simplicity should track truth — whether it’s a fact about how the universe happens to be built, a reflection of how human minds are wired to reason, or simply a practical convenience for choosing between theories — remains a genuinely open question in the philosophy of science, with no settled consensus. The Stanford Encyclopedia of Philosophy devotes an entire, still-unresolved entry to competing justifications for the principle.

Key Takeaways

• Occam’s razor says that when two explanations fit the evidence equally well, the one that makes fewer new assumptions should generally be preferred — it does not say the simplest explanation is automatically true.
• The principle is named for William of Ockham (c. 1287–1347), but the underlying preference for simplicity traces back at least to Aristotle.
• Occam’s razor was a major reason the far simpler Sun-centered model of the solar system eventually displaced the Earth-centered model, which needed increasingly complex add-ons to keep matching observations.
• In medicine, Occam’s razor is deliberately balanced against Hickam’s dictum, the reminder that a patient can have more than one disease at the same time.
• In machine learning, Occam’s razor is formalized as regularization — a technique that penalizes unnecessarily complex models so they generalize better to new data.
• Simplicity functions as a tiebreaker between theories that already fit the evidence, not as evidence itself — it cannot override data that contradicts a simpler theory.
• Why simplicity should track truth at all — versus just being a convenient way to choose between theories — remains a genuinely unresolved question in the philosophy of science.

Frequently Asked Questions

Is Occam’s razor a proven scientific law?
No. It’s a heuristic, not a law of nature or a mathematical proof. It offers a reasonable basis for choosing between theories that already fit the evidence equally well — it cannot override evidence that contradicts a simpler theory.

What is a simple, everyday example of Occam’s razor?
If you hear hoofbeats outside your window, assuming a horse is a more Occam’s-razor-consistent guess than assuming a zebra escaped from a zoo — not because zebras are impossible, but because “horse” requires fewer additional unlikely assumptions to be true.

Who actually invented Occam’s razor?
The principle is named after William of Ockham (c. 1287–1347), but the underlying preference for simple explanations predates him by centuries, tracing back at least to Aristotle.

Does Occam’s razor apply outside of science?
Yes — it’s widely used in philosophy, law, debugging software, and everyday decision-making as a general caution against over-explaining something that has a simpler cause.

Is a more complex theory ever the correct one?
Yes. Occam’s razor only applies when two theories explain the evidence equally well. If a more complex theory explains something a simpler one cannot, the complexity is justified — general relativity is a well-known example of a mathematically intricate theory that is also extremely well confirmed.

How is Occam’s razor used in artificial intelligence?
Machine learning engineers use a version of it called regularization, which penalizes models for unnecessary complexity during training so they are less likely to memorize noise in the training data and more likely to perform well on new, unseen data.

The Bottom Line

Occam’s razor survives, nearly 700 years after Ockham’s death, because it solves a problem every field of inquiry eventually runs into: what to do when more than one story fits the facts. It doesn’t promise that the simplest answer is the true one. It promises something more modest and, in practice, more useful — that unnecessary complexity should have to earn its place in an explanation, not be assumed for free.

Sources & References

• The Editors of Encyclopaedia Britannica. “Occam’s razor.” Encyclopaedia Britannica.
• Baker, Alan. “Simplicity.” Stanford Encyclopedia of Philosophy, Stanford University.
• Henderson, S.R., Shah, A., Copley, S.J., et al. “Occam’s razor or Hickam’s dictum? Allergic bronchopulmonary aspergillosis and eosinophilic granulomatosis with polyangiitis.” Thorax, BMJ Journals (peer-reviewed), 2015. DOI: 10.1136/thoraxjnl-2015-207280. Hosted via the U.S. National Library of Medicine (NCBI/PMC), National Institutes of Health.
• Google for Developers. “Overfitting: Model complexity.” Machine Learning Crash Course.