Quine’s ultimate aim is to argue against the analytic/synthetic distinction, and the real heart of his argument against the distinction comes only in the last two sections of the paper. That will be the focus of our discussion; but you will need to have some familiarity with the first part of the paper, where Quine ties the idea of analyticity to a variety of other concepts, so I’ve also assigned sections I and IV. You should ask yourself, at each stage, what it would be like to abandon the distinctions and the other concepts that Quine links to the idea of analyticity (a number of these distinctions and concepts are marked by italics below).
• In §I, Quine considers three fairly direct ways of defining analytic truth—as truth in all possible worlds (20), as truth in virtue of meaning (21), and as any result of substituting synonyms for synonyms in a logical truth (23). His discussion of the distinction between intension and extension and of essence (21f) is less directly tied to his argument but it will be important for us since it is relevant to issues raised in the last week of the course. (The final discussion of Carnap’s state descriptions (23f) is a historical artifact; as Quine notes, Carnap’s idea was not intended as a general account of analyticity; but it might have been thought to provide such an account by someone who read this paper when it was first published.)
• In §IV, doubts about analyticity are connected with doubts about the idea of semantical rules (31). Although the discussion will often seem to turn on technical issues, this section is broad in its import because Quine is in effect questioning the very notion of meaning. He questions it in a way which recalls earlier work in which he had cast doubt on the idea of truth by convention, and his doubts about analyticity can be seen to at least run parallel to broad doubts about that idea. The full implications of Quine’s doubts about analyticity start to be indicated in the last paragraph of §IV (34f), where he casts doubt on the distinction between matters of language and matters of fact.
• The heart of Quine’s argument against the analytic/synthetic distinction lies in an attack on the verification theory of meaning that occupies the bulk of §V.
• On p. 35, he presents the theory as he sees it and shows how, if sustainable, it would support the analytic/synthetic distinction.
• Quine’s main target is a strong form of the verification theory which he calls “radical reductionism.” He describes and attacks this on pp. 36. Quine’s references to Carnap’s Aufbau (his Der logische Aufbau der Welt, or Logical Construction of the World) are rather sketchy, but they are important. Notice especially the reference to maximizing or minimizing “certain over-all features,” or achieving “the laziest world compatible with our experience” (both on 37).
• Someone holding the verification theory need not be committed to radical reductionism and Quine indicates a less ambitious alternative beginning on p. 38. His argument against this “attenuated form” follows the same lines as his argument against the stronger version though it is necessarily less conclusive. He states it in the paragraph on the middle of p. 38.
Quine later added a footnote to this paragraph in which he referred to the physicist, philosopher, and historian of science Pierre Duhem (1861-1916), citing an argument of Duhem’s against the idea that experimental evidence could ever force anyone to reject a hypothesis. The reason Duhem denied this possibility was that further auxiliary assumptions are always required to tie a hypothesis to experimental tests—if only to predict the operation of experimental apparatus. And this means that, if the predicted result of an experiment fails, the hypothesis being tested can be retained if blame is cast on one of these other assumptions. Notice that this would mean that a hypothesis could not be tested by itself but only as a part of a “corporate body” including the further assumptions. In “Two Dogmas,” Quine strengthens Duhem’s claim, suggesting that Duhem’s point applies to all empirical statements and that no boundary can be drawn around the corporate body of assumptions relevant to any one of these statements.
• Later on p. 38, Quine turns his arguments against the verification theory into an argument against the analytic/synthetic distinction. His argument goes roughly as follows. Anyone who maintains the analytic/synthetic distinction is committed to a distinction between factual and linguistic components of the truth of any statement—i.e., between the facts that make it true and the meaning that ties its truth to these facts. An empiricist will be committed also to the verification theory as an account of any meaning component of truth. But verificationism fails to assign meaning to individual statements. Thus, according to Quine, no empiricist can accept the analytic/synthetic distinction.
Quine’s concluding comment in this section—that the smallest meaningful unit is “the whole of science” (p. 39)—can be seen as his topic in the final section of the paper.
• Quine’s alternative to the two “dogmas” that he has criticized centers on two ideas, each taking up about half of §VI.
• The first is the comparison of our beliefs to a field of force underdetermined by its boundary conditions in experience (pp. 39f). A standard concrete image for a field with boundary conditions is a flexible sheet (for a drumhead perhaps) being pulled at its edges. The field would be undetermined if the forces at its edges did not uniquely determine the positions of interior points—for an extreme case of this, imagine the sheet is made of bubble gum.
• Quine’s second way of filling out his alternative to the dogmas (pp. 41ff) lies in his idea of “posits” and his claim that various such posits—e.g., “brick houses on Elm Street” (pp. 40, 43)—differ from centaurs or Homer’s gods “only in degree and not in kind” (p. 41).
Quine points to some of the consequences of these ways of thinking about science and our knowledge generally, but you should try to explore them further still. Do they seem to provide an acceptable alternative to one based on a distinction between linguistic understanding and factual reality?