The Analytic/Synthetic Distinction and Quine's Attack
On this page(5)
The empiricist has a problem with logic and mathematics, and for two centuries the same solution was offered. These truths are not discoveries about the world at all. They are true in virtue of what the words mean, which is why no observation could overturn them and why no observation is needed to establish them. The mystery of a priori knowledge dissolves, because there is nothing out there being known.
In 1951 Willard Van Orman Quine published a paper arguing that the solution cannot be stated. The paper is called "Two Dogmas of Empiricism", it is among the most cited in twentieth-century philosophy, and the question of whether it succeeded is still open.
The distinction under attack
Definition
Analytic truth
A proposition true solely in virtue of the meanings of its constituent terms, so that understanding it is sufficient for seeing that it is true. Standard examples are that all bachelors are unmarried and that no square is round.
Definition
Synthetic truth
A proposition whose truth depends on how the world is and not on meanings alone, so that understanding it leaves open whether it is true.
Definition
Conceptual truth
A truth held to follow from the content of the concepts involved rather than from empirical fact. Often used interchangeably with analytic truth, though writers who distinguish them treat conceptual truth as the broader notion, covering cases where the connection between concepts is not a matter of definition.
Kant's version defined analyticity as containment of the predicate concept in the subject concept, which is vivid and narrow. It applies to subject-predicate judgements and leaves the logical truths awkwardly placed. Frege improved it: a truth is analytic if it can be turned into a logical truth by substituting synonyms for synonyms. This is the version Quine attacks, and it is worth noticing that his attack begins by granting the logical truths and asking only about the rest.
That everything unmarried is unmarried is a logical truth outright. That every bachelor is unmarried is not, and becomes one only when bachelor is replaced by unmarried man. The whole account therefore rests on the claim that these are synonyms, and Quine's question is what that claim amounts to.
The circle
Definition
Quine's circularity argument
The argument that every proposed explanation of analyticity relies on a notion, such as synonymy, definition, or necessity, that can itself be explained only by appeal to analyticity, so that the family of terms is never anchored outside itself.
Quine takes the candidates in turn.
Definition will not do. To say that bachelor is defined as unmarried man is to report a pre-existing synonymy that the lexicographer has observed, not to create one. Except in the case of deliberate abbreviation, definition presupposes the relation it was supposed to explain.
Interchangeability without change of truth value will not do either, at least not in an extensional language, since it makes creature with a heart and creature with a kidney synonymous, these having the same extension as a matter of fact. To get the right result you must require interchangeability in all contexts including necessarily, which brings in necessity. And necessity, in the tradition Quine is addressing, was explained as truth in virtue of meaning, which is analyticity again.
Semantical rules will not do. One can stipulate for an artificial language which sentences count as analytic, and Carnap did. But the stipulation defines analytic-in-L for a particular L, and says nothing about what property these sentences share that makes the label apt. Quine's complaint is that a list is not an explanation, and that appealing to a rule that assigns the label simply relocates the question.
Quine's argument against analyticity
- An explanation of analyticity must not itself presuppose analyticity or any notion definable only in terms of it.
- Definition presupposes synonymy, which is a member of the same family.
- Interchangeability yields the right results only when restricted by necessity, which is a member of the same family.
- Semantical rules supply a list for a chosen language without saying what the listed sentences have in common.
- ∴No explanation of analyticity is available that does not move within the circle, and the distinction is therefore an unempirical dogma.
Check your understanding
The positive replacement
The second half of Quine's paper is often skipped and is where the lasting influence lies.
Definition
Confirmation holism
The thesis that empirical evidence bears on bodies of belief as wholes rather than on individual statements, so that no single statement has its own confirming or disconfirming observations.
Quine's image is of knowledge as a fabric that meets experience only at the edges. A conflict with observation produces a redistribution of truth values somewhere in the interior, and logic and mathematics sit at the centre, not because they are immune to revision but because revising them disturbs everything else. Any statement can be held true come what may, if adjustments are made elsewhere, and by the same token no statement is immune from revision.
This is what dissolves the distinction from the other direction. If confirmation is holistic, there is no such thing as the observations relevant to a given sentence, and so no sentence divides cleanly into the ones confirmed by experience and the ones confirmed by meaning alone. The two dogmas turn out to be one: the analytic and synthetic distinction stands or falls with the reductionist assumption that each statement has its own empirical content.
Quine's own example is that the logical law of excluded middle has been proposed for revision in response to quantum mechanics. Whether that proposal is a good one is beside his point, which is that it is intelligible as a proposal, which it would not be if the law were true by meaning.
Check your understanding
Where the argument stands
Quine's conclusion is not widely accepted now, and his framework is not widely rejected either, which is an unusual combination.
The revival of the distinction owes most to developments Quine did not anticipate. Kripke's separation of necessity from apriority removed one of the pressures that had made analyticity carry too much weight. Work in semantics has given synonymy a theoretical home in linguistics, where it is treated as an empirically tractable relation rather than a philosopher's primitive. And the two-dimensional framework developed by David Chalmers and Frank Jackson offers a way of reconstructing something distinction-like out of how a term's reference is fixed.
The holism, meanwhile, has become close to orthodoxy. Few now believe that individual sentences face the tribunal of experience alone, and the picture of belief as a web under simultaneous adjustment is the background of contemporary epistemology, including the Bayesian machinery two articles from now.
What Quine permanently changed is the burden of proof. Before 1951 the analytic and synthetic distinction was a starting point. After it, anyone who wants to use the distinction is expected to say what it comes to.
Objection
Quine's argument is self-undermining. His conclusion is that no statement is immune from revision, including the laws of logic. But the argument for that conclusion uses logic at every step, and if logic might be revised then the argument might not be valid. Worse, the conclusion is offered as something we should accept on the basis of reasoning, which requires that some inferential standards be in force, and those standards are exactly what the conclusion says could go.
Reply
The charge is old and the standard reply is available, though it costs Quine something.
Quine is not claiming that logic is revisable in the sense that the argument's validity is currently in doubt. He is claiming that revisability is a matter of degree and that nothing sits at degree zero. Using logic while holding that it might in principle be revised is no more incoherent than using a bridge while holding that it might in principle need replacing. The argument is offered from within the web, as all arguments are, and the position explicitly denies that there is anywhere else to stand.
There is a sharper version of the objection that survives this. If holding a statement true come what may is always possible, and if any statement can be given up, then the notion of evidence bearing on belief threatens to become empty, since any body of evidence is compatible with any statement being retained. Quine's answer is that the constraint is not logical but practical, a preference for minimal mutilation and for simplicity, and that this is all the constraint that empirical method ever had.
Whether that is enough is the real question, and it is not one the self-refutation charge reaches. The worry is not that Quine's argument is invalid. It is that his replacement for the discarded distinction leaves the rational force of evidence resting on nothing firmer than a preference for tidiness.
Check your understanding
Next
The oldest problem about empirical justification, which holism does nothing to solve: why anything observed should tell us about anything not yet observed.
Further reading
- W. V. Quine, "Two Dogmas of Empiricism", The Philosophical Review, volume 60, number 1 (1951), pages 20 to 43.
- H. P. Grice and P. F. Strawson, "In Defense of a Dogma", The Philosophical Review, volume 65, number 2 (1956), pages 141 to 158.
- Rudolf Carnap, "Meaning Postulates", Philosophical Studies, volume 3, number 5 (1952), pages 65 to 73.
- Gillian Russell, Truth in Virtue of Meaning (Oxford University Press, 2008).
- Stanford Encyclopedia of Philosophy, "The Analytic/Synthetic Distinction", by Georges Rey.
Quiz yourself on this article
10 questions. Miss one and it comes back until you get it right.
Reached the end?
Mark it, and the sidebar keeps track of where you have been.