Do not multiply entities beyond what the explanation requires.
Near-universal rule
Explanation
Two accounts explain everything you've observed. One needs an extra piece the other manages without. The razor's verdict: that extra piece has not earned its place — not because it couldn't exist, but because nothing in the evidence asked for it, and nothing shifts if it's removed.
The mechanism is one question: is this part doing work? Take it out, mentally. If nothing the account predicts or explains changes, the part has no claim to stay. The leaner account is not thereby *right* — it is the one to hold until evidence forces an addition.
One condition governs: the razor only runs between accounts that explain the same observations equally well. Where one account explains something the other cannot, the fuller account wins on its own — the razor was never part of that call.
Watch
No video has yet earned its place here.
When it applies
When two accounts cover all the same observations equally well and one requires an entity the other doesn't, the razor applies: the extra entity has not proved its necessity.
When an account introduces an assumption that does no work beyond what existing assumptions already cover, the razor applies: the addition is idle.
When two models predict equally well but one uses more free parameters to do it, the razor applies: the leaner model is the right default.
The pattern across all three: equal explanatory work, unequal load. The razor cuts at that gap.
Where it stops
Since this is a near universal concept, when underlying conditions are met, it can always be used.
The misuse
The typical error is reaching for the razor before checking whether both accounts explain the evidence equally well. That check is the precondition. Skip it and you are not running a tiebreaker — you are cutting a piece that may have been doing the work you never looked for.
A second misread strips "beyond necessity" from the formula and hears only "fewer entities." The razor prohibits what is idle; it does not push toward the smallest possible entity count. A necessary entity is not surplus because it is an entity.
The deeper promotion: the razor gets lifted from a default into a truth-test. It tells you which account to *hold* in the absence of reasons to add more — not which account is *true*. Treating the leaner story as correct because it is leaner turns a methodological tiebreaker into an ontological verdict.
A worked example
William of Ockham applied parsimony against the prevailing realist position on universals. The realist account required positing universals — real entities in the world, separate from any individual — to explain what individual things have in common. Ockham's nominalist account required no such thing: individual things exist; the universal is a concept the mind forms, not an entity the world contains independently. Both accounts explained what needed explaining. One required an entity the other managed without. That settled the comparison on parsimony.
Holding the leaner position came at personal and institutional cost. The individuals remained. The universals were not missed.
Push
When two accounts explain the evidence equally well, hold the one that needs no extra entity — and put the burden of proof on every addition, never on the omission.
Veto
Do not carry an entity, cause, or step that makes no difference to what the account predicts or explains. If removing it changes nothing observable, the razor rules it out — not a weaker option, ruled out.
And equally: "it is simpler" is not grounds to discard an entity that *is* doing explanatory work. Simplicity does not license cutting what the evidence requires.
Connects to
Connections appear as the deck grows.
Go deeper
Shows the principle at work in the disputes that forced it into practical form
Book
lets you see what applying it actually required.
William of Ockham
Book
*Summa Totius Logicae*** (c. 1323)
Mastery question
**Question:** You are weighing two explanations of the same situation. The leaner one leaves part of what you've seen unaccounted for; the fuller one accounts for all of it but requires an extra part. Which does the razor say to prefer — and why is that a trick question?
**The answer:** The razor says nothing here — its precondition isn't met. It adjudicates only between accounts that explain the evidence equally well. These two don't: one leaves observations unexplained. The fuller account wins — but on explanatory adequacy, not on parsimony. The razor never licenses cutting a part that is doing work; it only forbids carrying a part that isn't. Until both accounts are tied on the evidence, the razor doesn't fire.
**The answer that misses it:** "The leaner one — Occam's razor prefers fewer assumptions." (Or, hedged: "the simpler one, unless the extra part is really needed.")
**Why it matters:** Someone who has memorised "prefer the simpler explanation" reaches for it the moment one account looks lighter, without checking whether both actually cover the same observations. Someone who has internalised it knows the rule only fires when both accounts are tied on the evidence — and that condition is the part that does the real work. The question is built so the lean account is inadequate; only a reader who checks adequacy before counting parts notices the razor doesn't apply at all.