Analytic vs Synthetic Distinction

How the distinction works

Take two sentences. “All bachelors are unmarried.” If you understand the words, you know it's true; you don't have to interview any bachelors. The predicate (“unmarried”) is already contained in the subject (“bachelor” means “unmarried adult man”). Kant called this analytic: the predicate analyses out of the subject.

Now: “Snow is white.” Understanding the words isn't enough — you have to actually look at snow. The concept of snow doesn't logically contain the colour white; it could have been grey, like the snow of industrial Manchester. Kant called this synthetic: the predicate synthesises new information onto the subject.

The standard tests philosophers use:

1. Negation test. Negate the sentence. Is the result a contradiction (“there exists a married bachelor”) or merely an empirical falsehood (“snow is purple”)? Contradictions point to analyticity.
2. Synonymy substitution. Replace each word with a synonym. If the result is still trivially true, the original was probably analytic.
3. Definition test. Could a competent speaker recognise the truth purely from the dictionary entry? If yes, it's a candidate for analyticity.

None of these is watertight — that's the heart of Quine's complaint, below — but together they pick out a recognisable family of cases.

Why Kant needed the distinction

Kant's Critique of Pure Reason (1781) faced a crisis. Hume had argued that all knowledge divides into “relations of ideas” (trivial, like definitions) and “matters of fact” (empirical, never certain). Mathematics seemed to fit neither: 7 + 5 = 12 isn't a definition (the concept of 12 isn't contained in the concepts 7, +, 5), but it also doesn't depend on experience — you don't verify it by counting beans.

Kant's solution was a four-cell grid: analytic vs synthetic, crossed with a priori (knowable independently of experience) vs a posteriori (knowable only through experience). Three cells were uncontroversial; the explosive cell was synthetic a priori — knowledge that says something new about the world but is justified without experiment. Kant put mathematics, Euclidean geometry, and the principle that every event has a cause in this cell.

	A priori (justified without experience)	A posteriori (justified through experience)
Analytic	“All bachelors are unmarried” — uncontroversial	Empty cell — Kant says impossible
Synthetic	“7 + 5 = 12” — Kant's signature category	“Snow is white” — uncontroversial

Analytic vs synthetic — at a glance

	Analytic	Synthetic
Source of truth	Meaning of terms	Way the world is
Negation	Self-contradictory	Merely false
Information added	None — predicate in subject	Genuine new content
Verification	Conceptual analysis	Observation or experiment
Defeasible by experience?	No (Kant) / yes (Quine)	Yes
Typical examples	Definitions, logical truths, “triangles have three sides”	Empirical claims, scientific laws, history
Disputed cases	Math, “every event has a cause”, ethical truths	Necessary identities (“water is H₂O”)

A worked example: “Whales are mammals”

Is “whales are mammals” analytic or synthetic? It looks analytic — “mammal” is part of the definition of “whale” in any modern biology textbook. Yet for centuries Europeans classified whales as fish; Linnaeus reclassified them as mammals only in 1758. The truth was discovered, not unpacked from the meaning of “whale” in ordinary 17th-century English.

Quine uses cases like this to argue that what counts as “part of the meaning” is itself a moving theoretical decision. We chose to define whales by reproductive biology rather than by habitat. That choice was driven by empirical theory. So even paradigm “analytic” truths are disguised theoretical commitments — exactly Quine's holism.

Quine's attack: “Two Dogmas of Empiricism”

W. V. O. Quine's 1951 paper is one of the most cited articles in 20th-century philosophy. The argument runs:

1. Definitional analyticity presupposes synonymy: “bachelor” and “unmarried adult man” mean the same thing.
2. But what is synonymy? Two terms are synonymous if they can be substituted in any sentence without changing truth value — except in opaque contexts like belief reports, where the substitution fails.
3. The fix: synonymy holds for “cognitive” substitutions only — i.e., where the result is analytic. We're back where we started.
4. Every alternative explanation Quine surveys (semantic rules, intension, conceptual containment) bottoms out in another notion that itself needs analyticity to be defined.
5. Therefore: there is no principled, non-circular distinction. Beliefs face the “tribunal of experience” as a connected web; any belief — even logic — could in principle be revised.

Quine's positive picture is semantic holism: the unit of meaning isn't a sentence but a whole theory. Some sentences sit near the periphery (easy to revise — “there's milk in the fridge”) and some near the centre (hard to revise — laws of logic). The difference is degree, not kind.

Responses to Quine

• Grice and Strawson, “In Defence of a Dogma” (1956). Quine demands a reductive definition; but ordinary distinctions don't usually need one to be real. We distinguish red from orange without a non-circular reduction.
• Carnap's response. Quine's argument works only if we deny stipulative definitions. In a formal language we just stipulate which sentences count as analytic via meaning postulates; circularity is no objection because we're constructing the distinction, not finding it.
• Boghossian (1996). Distinguishes “metaphysical” analyticity (true in virtue of meaning alone — Quine kills this) from “epistemic” analyticity (justified purely on the basis of meaning — survives). Logic and math may be analytic in the second sense.
• Kripke's twist (1980). Kripke's Naming and Necessity introduced necessary a posteriori truths (“water is H₂O”) and contingent a priori truths (“the standard metre is one metre long”), shattering the assumption that analytic = necessary = a priori. The grid is messier than Kant thought, but the categories don't collapse.

Why the distinction matters

• Philosophy of mathematics. If math is analytic, it's about logic; if synthetic, about a special abstract subject matter. Frege, Russell, Hilbert, Gödel all stake out positions here.
• Philosophy of science. Are theoretical postulates (“force = mass × acceleration”) definitional or empirical? The Duhem–Quine thesis says you can always shift the answer by relocating where you absorb anomalies.
• Conceptual engineering. Modern philosophers (Cappelen, Plunkett) ask whether we should change our concepts. The project assumes some claims are true by meaning — i.e., analyticity is doing real work, even if blurry at the edges.
• AI and language. Large language models trained on text learn statistical co-occurrence; they don't draw the analytic/synthetic line. Whether they can is an open empirical question with theoretical stakes.

Common confusions

• Analytic vs a priori. Different distinctions. Analytic is about what makes a sentence true; a priori is about how we know it. Kripke showed they don't coincide.
• Analytic vs necessary. Necessity is metaphysical (could it have been otherwise?), analyticity is semantic. “Water is H₂O” is necessary but not analytic.
• Analytic vs trivial. “Trivial” is psychological. Analytic truths can be deeply non-obvious — set-theoretic theorems are analytic in the logicist sense but took centuries to prove.
• Synthetic = empirical. Not in Kant's framework. Synthetic a priori is precisely synthetic-but-not-empirical.
• Quine refuted the distinction. Quine refuted the sharp, principled distinction. A graded one is alive and well.