Remedying Some Defects in the History of Analyticity

To a first approximation, analytic truths are those sentences that are true solely based on, or capable of being known by, semantic facts. The bulk of this dissertation is an investigation into how analyticity relates to the philosophical works of Descartes, Frege, Carnap, Quine, and Williamson. I also discuss some prima facie constraints on analyticity, and provide a survey of the relevant literature on the concept. Erde suggests that Descartes took "cogito ergo sum" to be analytically true on the basis of conceptual containment. I argue that Erde's thesis is false for several reasons, one of which is that if Erde's interpretation were correct, "cogito ergo sum" would have no more import than "I am walking, therefore I exist," but clearly Descartes thought that it does. Katz argued that though Descartes had a sense of the fact that "cogito ergo sum" is analytic, since Descartes did not have access to the true linguistic theory in which its justification can be formulated, Descartes's epistemology was provided as an ersatz alternative. I argue that Descartes's epistemology leaves no room for any amendments from linguistics. In short, Descartes held that "cogito ergo sum" is (1) true solely because it was made true by God, and (2) can be known to be true solely because God gave us the ability to intuit truths; however, Descartes did not appreciate that this account could not convince a skeptic. Finally, for the same reasons that undermined Katz' view, Díaz failed to show that Descartes's ontological argument involves analyticity. Nevertheless, I argue that since Descartes held that (1) the idea of God contains necessary existence and that (2) the understanding of words commits us to having the ideas signified by those words, we can know that "God exists" is true simply based on understanding those words. Though Kant's subject-predicate containment account of analyticity is superficially different from Frege's account involving only logic and definitions, I argue that the two are relevantly similar for the following reasons: Kant held a liberal view on what counts as subject-predicate sentences, maintained that analytic judgments ought to increase our knowledge, and advocated logical analysis as the way to determine which judgments are analytic. The substantive difference between Kant and Frege is whether analyticity is grounded in logical analysis based on concepts or definitions. Unlike Kant, who seems to have assumed that none of our concepts are vague, Frege called attention to the vagueness in ordinary language, and recommended that in formal contexts, vagueness be removed by "fruitful" stipulative definitions. In contrast with the later work of Wittgenstein, Frege thought that ordinary language was flawed for various reasons, the major one of which is that ordinary words, with their vague senses, rarely have perspicuous referents, and hence fail to be capable of capturing the definiteness of truth. Stipulations can ensure perspicuous referents, but such definitions must not invalidate any propositions already commonly accepted or affirm any that are already commonly not accepted, on pain of changing the legitimate subject associated with those words. Carnap and Quine initially agreed on the characterization of analyticity as only involving rule-governed symbol manipulation, but even this characterization tacitly imported the semantics of logical terms, and Carnap subsequently was convinced that widespread toleration of semantics could generate fruitful results. Quine argued that accepting semantics in this fashion leads to an incorrect account of ontology. Carnap's view on ontology was "pluralistic" in the sense that ontology varied by language, and languages varied by which sentences were taken to be analytic by the language formulators. Quine's famous case against analyticity amounts to the insistence that (1) "analytic" is in need of a scientifically respectable explanation, since it is a term of art, and that (2) there is no antecedently clear sense of "analytic" on the basis of which Carnap could use the word, or provide an explication for it. However, I show that Quine eventually gave this objection up and gave an explication of his own. The lasting disagreement between Carnap and Quine was over the usefulness of analyticity. Williamson's two main arguments against analytic and synthetic truths being true in different senses are reductios. According to him, if "true" were ambiguous, then the standard disquotational principles for true and false sentences, and the truth-tables for sentential logical connectives, would have to incorporate these distinct senses, but it is not possible to do so while keeping the alleged senses of "true" distinct. I argue that in the case of the disquotational principles, Williamson offers an inadequate disambiguation, and when appropriately disambiguated, the correct statuses of the disquotational principles can be demonstrated. In the case of compositional semantics, while Williamson is correct that characteristic truth-tables cannot be generally disambiguated, what is important is that the same problem does not apply to particular applications of truth-tables.