Looking back on the origins of “Two Dogmas of Empiricism,” W.V. Quine notes that he had in 1940, with Alfred Tarski’s support, already challenged Rudolf Carnap’s thinking about analyticity, but “[b]ecause of its negativity, [his] repudiation of analyticity was nothing [he] felt impelled to write about.”Footnote 1 Only when commissioned to speak to the American Philosophical Association in December 1950 did he finally put pen to paper and write up his thoughts. Similarly in a slightly later autobiography that he wrote for a general audience, he devotes just a couple of lines to the paper and dispatches it in very short order—he merely repeats that he had discussed the matter with Tarski and Carnap and his doubts about analyticity were long-standing.Footnote 2 And some forty years after the appearance of the paper, in a retrospective discussion, he again plays down what is now generally regarded as his most important and influential article. He writes: “I had not thought to look on my strictures over analyticity as the stuff of revolution. It was mere criticism, a negative point with no suggestion of a bright replacement.”Footnote 3 This is baffling. Why, assuming Quine is to be believed, his reluctance to tout what he had written and how come he failed to anticipate the ensuing “spate of controversial literature”?Footnote 4
Quine’s nonchalance about the contribution of “Two Dogmas” to philosophy is rarely advertised, and the paper is usually taken to pose a major challenge to philosophical tradition. It is read as two-pronged with the bulk of the discussion devoted to attacking the distinction between truth by virtue of fact and truth by virtue of meaning and the rest given over to promoting the view that our knowledge, trivial or substantial, concrete or abstract, is all of a piece. More specifically, §§1-3 are interpreted as denying the alleged logical difference between the likes of “There are brick houses on Elm Street” and “No bachelor is married,” quibbles about “ages, divorce and bachelors of arts” aside.Footnote 5 Moreover, §6, the final section of the paper, is interpreted as defending the proposition that the sole difference between such sentences is their centrality to our system of beliefs. (§4 on Carnap’s attempt to clarify analyticity in terms of semantic rules and §5 on “the verification theory of meaning” and the second dogma, the conception of meaningful statements as expressible in terms referring to immediate experience, are accorded much less attention, and I shall mostly skip over these sections.)
Unsurprisingly, common belief about “Two Dogmas” squares well with a fair amount of the text. At the start of the paper, Quine declares that he will argue that the “dogma” of “a belief in a fundamental cleavage” between truths “grounded in meanings independently of matters of fact” and truths “grounded in fact” is “ill-founded” and adds that “abandoning” this dogma (along with the dogma of reductionism) results in “a blurring of the supposed boundary between speculative metaphysics and natural science” (20). Still, such a line of interpretation leaves much to be desired. Besides attributing to Quine views he does not express and arguments he does not advance, it has the unfortunate consequence that, as has been argued time and again, his discussion falls badly short and is easily answered. Quine does not come close to disposing of the analytic/synthetic distinction once and for all, nor does he show his conception of knowledge to be superior to mainstream conceptions. It is no small irony that a paper routinely judged to be singularly important is just as routinely deemed unconvincing, even self-evidently wrong. While many philosophers embrace what they take to be Quine’s epistemological vision, few find what they take to be his reasoning at all credible.
Quine was largely indifferent to the criticism levelled against what he says in “Two Dogmas.” Having discussed the distinction at length during the 1940s, he would doubtless have already heard most of the objections pressed against him, and in the years following the publication of the paper he accepted what he took to be the grains of truth in his critics’ arguments but did not retract or qualify his main point. In Word and Object, he recognises a “behaviourist ersatz,” a “vegetarian imitation” of analyticity, covering “No bachelor is married,” all the while rejecting any more general notion of analyticity.Footnote 6 And in Roots of Reference, he eschews the traditional notion in favour of the idea that “a sentence is analytic if everybody learns that it is true by learning its words.”Footnote 7 Here and in later work, he stands his ground and bends his energies to fleshing out his conception of a dogma-free epistemology. He cashes out the metaphors in “Two Dogmas” to obtain what he takes to be a more satisfactory conception of knowledge, proposing, for instance, the notion of an observation sentence as a replacement for the “flabby reference [in ‘Two Dogmas’] to ‘experience.’”Footnote 8
But if Quine’s paper is not to be regarded as badly flawed (and his reservations about its importance respected), how should it be read? Three points need noting right away. First, as Quine later emphasised, his aims are epistemological in that he is primarily concerned with the suggestion that our knowledge of the unmarried status of bachelors is categorically different from our knowledge of the type of house on Elm Street. Though the phrase “a priori” occurs just once in “Two Dogmas” and then only in the everyday sense of “initial” or “on first thought” (37), he targets the distinction between statements known true prior to experience and statements known true only as a result of it. §1 (“Background for analyticity”) opens with a reference to “Kant’s cleavage,” a cleavage Quine sees as “foreshadowed in Hume’s distinction between relations of ideas and matters of fact, and in Leibniz’s distinction between truths of reason and truths of fact” (20). Furthermore in “Carnap and Logical Truth,” a companion piece from the same period, Quine mentions Kant’s question about the possibility of synthetic a priori judgement when introducing the topic.Footnote 9 And in Word and Object he notes that the analytic/synthetic dichotomy is “reminiscent of Kant” and takes issue with the idea of “a sweeping epistemological dichotomy.”Footnote 10 As he still later stresses, “the philosophically important question about analyticity [is] the question of [its] relevance to epistemology.”Footnote 11
Secondly, while Quine is regularly understood as arguing in §§1-3 that there is no non-circular way of characterising the notion of analyticity, this is not what he actually argues.Footnote 12 He does not rest content with observing that “analytic” belongs to a circle of notions including “self-contradictory” (20), “synonymous” (23), “definable” (24), “necessary” (29) and “interchangeable in all contexts without change of truth value” (32).Footnote 13 The problem is that we “lack a proper characterization of ... analytic statements” (23), analyticity construed in terms of self-contradictoriness having “small explanatory value” (20), synonymy being “no less in need of clarification than analyticity itself” (23), definition in the relevant sense presupposing synonymy “rather than explaining it” (26) and “extensional agreement fall[ing] far short of cognitive synonymy of the type required for explaining analyticity” (31).Footnote 14 In fact neither the term “circle” nor any of its cognates appear in the discussion of the analytic/synthetic distinction, only incidentally to pre-empt the objection that the argument being rehearsed itself is “flatly circular” (30). For Quine analyticity is unexplanatory and appealing to it gets us no further ahead. One might say the problem is not that the terms form a circle but that the circle of terms provides no explanation of a priori knowledge.Footnote 15
Thirdly, and surprisingly infrequently mentioned, is the fact that Quine discusses the question of analyticity within the framework of empiricism. While he refers to Kant and Leibniz at the beginning of the paper, it should have been clear that he mentions them to get the ball rolling and to underline that the analytic/synthetic distinction has a long and illustrious pedigree. It is presumably no accident that the paper sports the title “Two Dogmas of Empiricism,” not “Two Dogmas of Philosophy,” and that it begins with the words: “Modern empiricism has been conditioned in large part by two dogmas [i.e. the dogma of analyticity and the dogma of reductionism]” (20). Nor can it be fortuitous that Quine refers to himself as an empiricist, writing in §6: “As an empiricist I continue to think of the conceptual scheme of science as a tool, ultimately, for predicting future experience in the light of past experience” (44). Though he does not explicitly state that experience is the only sure guide to the truth about how things are, he was strongly of the opinion that experience is all we have to go on (and would have believed this substantive philosophical assumption requires no special defence beyond what can be learned from science itself). When he speaks of himself at the end of the paper as espousing “a more thorough pragmatism” (43), he is not qualifying his empiricism. As he later noted, he was “not clear on what it takes to be a pragmatist,” merely “taking the word from Carnap and handing it back.”Footnote 16
Bearing in mind these three points when reviewing what Quine says in “Two Dogmas,” we find that he proceeds very differently from how he is normally held to proceed and his evaluation of the paper becomes much less puzzling. In §§1-3 he does not pretend to consider all conceivable, even all plausible, ways of defining analyticity (for one thing in §4 he discusses Carnap’s more subtle treatment). Rather he casts his discussion within a framework of empiricism and attempts to show that “modern empiricists,” i.e. empiricists who embrace the analytic/synthetic distinction, jump the gun. It is no argument against him, as assumed in some quarters, that he equates the a priori with the analytic.Footnote 17 To the contrary, he objects to the equation, his challenge being directed at the assumption, widely held by empiricists when he was writing, that the analytic/synthetic distinction is epistemologically important since it explains putative a priori knowledge. Moreover, mindful of the preliminary points just noted, we are much less likely to overlook that in §6 Quine is still working within the framework of empiricism, in particular that he is not arguing that philosophers of every stripe should accept the conception of epistemology he advocates, only that empiricists who believe the a priori can be retrieved by invoking analyticity should accept it—a much less easily refuted contention.
As Quine sees it, “modern empiricists” do not live up to their own standards of concept construction. Taking philosophers who nail their flag to the empiricist mast at their word, he argues that the a priori cannot be accounted for by appealing to analyticity construed in terms of self-contradictoriness, synonymy, definability, necessity or “interchangeability.” These notions do not explain the a priori, a fact that Quine takes to provide good reason to believe that the analytic/synthetic distinction is “a metaphysical article of faith” (37). (By “metaphysical” Quine means, I take it, “empirically unjustified.”) Since self-respecting empiricists cannot tolerate unempirical dogmas, they must resign themselves to the conception of epistemology adumbrated in §6 (or come up with an adequate response to Quine’s challenge). They cannot but agree that “no statement is immune to revision” and there is no “difference ... in principle” between the shift in “the logical law of the excluded middle [that] has been proposed as a means of simplifying quantum mechanics” and “the shift whereby Kepler superseded Ptolemy, or Einstein Newton, or Darwin Aristotle” (43).
More precisely, Quine is out to establish that the (modern) empiricist has not justified the introduction of an explanatorily relevant epistemological distinction between analytic and synthetic statements, i.e. has not shown that a distinction explaining a priori knowledge can be devised. Despite the seeming plausibility of treating the a priori in terms of analyticity, he writes: “[A] boundary between analytic and synthetic statements simply has not been drawn. That there is such a distinction to be drawn at all is an unempirical dogma of empiricists” (37). Put otherwise, since empiricists cannot draw the distinction in any of the ways they are apt to suppose, they should not weigh their empiricism down with it. It is not good enough to assume it possible to introduce a distinction between analytic and synthetic statements, one that captures and hence explains the distinction between so-called a priori and so-called a posteriori knowledge. This has to be demonstrated. The fundamental difficulty is that the “modern empiricist” wrongly assumes there is no need to show (or thinks it can readily be shown) that there exists a line between analytic and synthetic statements that explains the a priori/a posteriori distinction. No such line has been drawn and—as the considerations canvassed in §§1-3 (and §4) strongly suggest—the prospects of drawing such a line are exceedingly dim.
To appreciate Quine’s discussion in “Two Dogmas,” it helps to consider how he proceeded in the early 1930s when he was closest to Carnap. In “Lectures on Carnap,” lectures he gave at Harvard University in 1934 when just 26, he accepts the analytic/synthetic distinction and sketches a method for introducing it, his thought being that a set of sentences can be delineated, each of which possesses what, following Kant, he calls “inward necessity.”Footnote 18 Though he does not say it in so many words, he plainly takes modern empiricists, Carnap in particular, to be on firm ground just insofar as they can show sentences generally held to be analytic possess such necessity. Within the year, however, he realised that the prescription he had provided for introducing the distinction does not work, and he came to question the possibility of concocting a prescription that does the trick. Thus, in “Truth by Convention,” a paper published in 1935, he announces that the process of turning sentences into analytic truths fails, logic being required to show that logic has “inward necessity.”Footnote 19 Since there does not seem to be a process of “elevating a statement from putative to conventional truth,” he concludes that it is doubtful that saying a statement is “true by convention” (and analytic) adds anything to saying it is “firmly accepted.”Footnote 20
Especially striking and important for understanding Quine on analyticity is that in both “Lectures on Carnap” and “Truth by Convention” he considers the possibility of dividing the body of sentences accepted as true into two with sentences that possess “inward necessity” on one side and sentences that do not possess it on the other. (In the “Lectures” Quine starts with “the whole range of admittedly true sentences” and “our whole range of accepted sentences,” while in “Truth by Convention” he starts with “more or less firmly accepted statements.”Footnote 21) His central assumption is that only given a method for turning accepted truths into definitions can it be justifiably concluded that a line can be drawn coinciding with the line dividing a priori from a posteriori knowledge. The primary difference between the lectures and the paper is not what Quine demands, but what he thinks can be achieved. Whereas in the lectures he believes the notion of “inward necessity” can be captured, in the paper he doubts it.Footnote 22 In both 1934 and 1935, he takes it to go without saying that it has to be demonstrated that a line can be drawn in a principled way—at least “nearly enough”—across the body of firmly established statements dividing a priori statements from a posteriori statements.Footnote 23
Quine, I want to suggest, broaches the problem of analyticity in 1950 the same way as he broached it in 1934/1935. In “Two Dogmas,” he again proceeds within the framework of empiricism, assumes we are apprised of a body of accepted truths and considers the possibility of retrieving the notion of the a priori by appealing to the notion of analyticity. What changed between 1935 and 1950 is mainly the strength of his doubts regarding the possibility of devising a recipe for separating the analytic from the synthetic (and hence the a priori from the a posteriori). Whereas in 1935 he was unsure of success, by 1950 he had come to think nothing will work. He had satisfied himself that analyticity is an ill-begotten notion, there being no satisfactory alternative to the flawed recipe proposed in “Lectures on Carnap” and explored in “Truth by Convention.”Footnote 24 Though the phrase “inward necessity” is dropped (and synonymy becomes his main focus of attention), Quine continues, on the account of the paper I am defending, to question the possibility of introducing an empirically acceptable notion of analyticity that captures the philosophical conception of the a priori. In “Two Dogmas” he suggests that some seemingly plausible ways of filling the bill are no such thing and implies this is no great loss.
It is worth noting that Quine’s approach to analyticity is very different from Carnap’s, and it is easy to see why they never saw eye to eye. It is one thing to hold, as Quine did, that the existence of an empirically acceptable distinction between analytic and synthetic statements has to be shown to be acceptable, not merely explicated in related terms, quite another to hold, as Carnap did, that there is a distinction, one that only requires explication. Contrary to Quine’s later characterisation of his 1934 lectures, they are not “abjectly sequacious.”Footnote 25 Carnap does not start out assuming we are apprised of a range of accepted sentences and go on to seek a method for turning sentences in the range into analytic truths, let alone suspend judgement concerning the existence of the distinction. He assumes there is a distinction and sets out to render it precise.Footnote 26 The “suitable setting” for considering the problem of analyticity that Quine provides in his lectures on Carnap makes all the difference.Footnote 27 Given Quine’s starting point, he was all but bound—once he realised that the argument in the lectures is erroneous—to end up concluding that the distinction is a dogma of empiricism. When the error is removed, the sceptical conclusion follows as a matter of course (pending the discovery of a better method of introducing the distinction, an exceedingly tall order).
The singular nature of Quine’s approach is especially apparent in “Notes on the Theory of Reference,” a generally overlooked work from the same period as “Two Dogmas.”Footnote 28 In this work, Quine takes up the objection pressed by Carnap and R.M. Martin that analyticity is in no worse shape than truth.Footnote 29 The notion of “truth-in-L” (where L stands for a particular language) is conceded to be scientifically irreproachable and the empiricist is allowed to have—subsequent to Tarski’s discussion of truth—as much right to the notion as anyone. But “analyticity-in-L” is still different since it lacks justification of the sort that Tarski provides for “truth-in-L.” The snag is that while we have an acceptable criterion of adequacy for truth—“p” is true-in-L (for each sentence p of L) if and only if p—we have no such criterion for analyticity. Whereas Tarski’s paradigm endows “is true-in-L” with “every bit as much clarity, in any particular application, as is enjoyed by the particular expressions of L to which we apply [the predicate],” Quine writes, “we have no clue comparable in value [for ’is analytic-in-L’].”Footnote 30 In other words “analytic-in-L” has to go because it cannot be disquoted and there is no substitute for disquotation in the offing.Footnote 31
Quine is not averse to analysis and explication, and it is no argument against him that he disparages the notion of analyticity while providing accounts of comparable notions. Thus in Word and Object he deems the analysis of ordered pairs in terms of unordered sets a “philosophical paradigm.”Footnote 32 And he has no quarrel with Carnap’s attempt to explicate various notions, only with his attempts to explicate of the notion of analyticity. He himself explicates when possible and where necessary, neither of which is the case, he thinks, for analyticity. He does not work with a double standard and should not be criticised for attempting to explicate observationality while rejecting the possibility of explicating analyticity. In his view observationality, unlike analyticity, is central to science and while there is a need for a clear notion of “observation sentence,” there is no need for a clear notion of “analytic sentence”. Not unreasonably he takes it to make good sense to “foster perspicuity by fancying boundaries” in the one case but not in the other.Footnote 33 His “reservations over analyticity ... concern the tracing of any demarcation, even a vague and approximate one, across the domain of sentences in general,” and he is unpersuaded that there is a notion of analyticity, the boundaries of which require fancying.Footnote 34
Quine’s misgivings about analyticity are reminiscent of nothing so much as Albert Einstein’s misgivings about simultaneity.Footnote 35 Like many philosophers, he was struck by Einstein’s work on the theory of relativity and apparently had it in mind when he first focused on the possibility of introducing the notion of analyticity to explain a priority. In both “Lectures on Carnap” and “Truth by Convention,” he refers to Einstein’s definition of simultaneity in terms of light emissions and treats the Einsteinian strategy of raising an accepted truth to the status of a linguistic convention as a model. Thus in his lectures he writes: “Einstein found it important, in enhancing the rigor of physics, to define ‘simultaneity’—rather than simply using the word [assuming] everyone concerned knew well enough what it meant,” and in the paper he writes: “The concept of simultaneity at a distance affords a stock example of [the replacement of a law by a definition].”Footnote 36 Though hard to prove, it is not farfetched to picture Quine as coming to believe in 1935 that the notion of analyticity (understood as doing service for “inward necessity”) is in no better shape than the classical notion of simultaneity. Certainly in 1950 he was of the opinion that philosophers should forgo such a notion of analyticity no less than Einstein was of the opinion that physicists should forgo the notion of absolute simultaneity.
The similarity between Quine’s and Einstein’s strategies even extends to the arguments they rehearse. In much the same way that Einstein questions the uncritical acceptance of absolute simultaneity, Quine questions the uncritical acceptance of analyticity. He judges the notion of absolute analyticity suspect in point of its empirical soundness no less than Einstein judges the notion of absolute simultaneity suspect in point of its physical soundness. And, like Einstein, he considers whether a certain generally assumed notion meets strict standards of scientific acceptability, concludes that it does not and refuses to accept promissory notes. It is, he thinks, incumbent on empiricists to show that analyticity pays its way, not presume it meets the scientific standards they favour, a demand that parallels Einstein’s demand on physicists regarding simultaneity. For him, the only useful notion that comes up to empiricist standards is his own “behavioural ersatz,” a notion comparable to Einstein’s notion of relative simultaneity. Whereas Einstein is intolerant of notions of simultaneity stronger than “simultaneous-in-a-frame-of-reference.” Quine is intolerant of notions of analyticity stronger than “analytic-in-L.”Footnote 37
It is striking too that objections to Quine’s discussion echo objections to Einstein’s discussion. It is often argued against Quine, as against Einstein, that he simply discounts the possibility of an underlying reality (in Quine’s case, a realm of meanings; in Einstein’s, absolute time). In Quine’s eyes, this objection cuts no more ice than the corresponding objection to Einstein cut ice in Einstein’s eyes. He holds that the assumption of underlying meanings is an untenable metaphysical prejudice, one that no self-respecting empiricist should give in to. And for much the same reason it is not a problem that he argues from an epistemological premise to a metaphysical conclusion (and, even worse, presupposes verificationism, a doctrine that nobody nowadays accepts). Exactly the same objection could be levelled against Einstein, and exactly the same response is in order. Quine, no less than Einstein, repudiates the belief that what is not known cannot be excluded, it being as central to his thinking as to Einstein’s—indeed to scientific method—that only well-founded notions are permissible.Footnote 38 A scientific philosopher if ever there was one, Quine thinks philosophers should shy away from notions that have not been shown scientifically sound.Footnote 39
And §6? How should it be read? Did Quine regard the account of epistemology he provides here “a bright replacement” or did he include it for want of anything better? There is, it has to be said, little of substance in §6 that is new. The positive view Quine sketches in this section is one he had long espoused. As early as 1934, he expressed himself in terms close to those he expresses himself in in 1951, and there are more than a few passages in “Lectures on Carnap” easily mistaken for passages in “Two Dogmas.” For instance Quine states in his lectures that revisions in our system of beliefs are “guided largely by the tendency to dislodge as little of previous doctrine as we can compatibly with the ideal of unity and simplicity in the resulting doctrine” and declares that “[t]he more firmly accepted sentences [will be modified] last, if at all, in the course of evolving and revamping our sciences.”Footnote 40 Still the discussion of §6 contributes significantly to philosophy. The importance of the paper lies less in Quine’s conception of epistemology than in his suggestion that there is no richer conception to be had. One could say that his bright idea is that there is no bright replacement available to empiricists and they do not need one. To treat the analytic/synthetic distinction as surplus to scientific requirements is not a minor departure from philosophical orthodoxy. It means retreating to a view of philosophy and logic as branches of—and on a par with—science.
§6 is often judged to be weak and easily challenged. Quine’s starting point has occasioned little dispute, it being generally conceded that “[t]he totality of our so-called knowledge or beliefs ... is a man-made fabric which impinges on experience only along the edges” and “total science is a field of force whose boundary conditions are experience” (42). And there has not been much criticism of his point that “the total field is so underdetermined by its boundary conditions, experience, that there is much latitude of choice as to what statements to reëvaluate in the light of any single contrary experience” (42-43). What has not been granted is that, as he goes on to say, “[n]o particular experiences are linked with any particular statements in the interior of the field, except indirectly through considerations of equilibrium affecting the field as a whole,” to say nothing of his further conclusion that “[a]ny statement can be held true ... if we make drastic enough adjustments elsewhere in the system” and “[c]onversely, by the same token, no statement is immune to revision” (43). Not without reason, it has been frequently noted that taking our knowledge to be all of a piece because theories are underdetermined by the evidence is a gross non sequitur. But does he fall for it? Only when the assumptions informing his discussion are overlooked.
Again, it is essential to remember that Quine is working within the framework of empiricism and that §6 is entitled “Empiricism without the Dogmas,” not “Epistemology without the dogmas.”Footnote 41 He does not suppose that whoever accepts the underdetermination of theory by evidence must regard our knowledge as an undifferentiated whole, only urges the far less objectionable proposition that empiricists must regard it this way given their view of evidence and empirical content. As in the 1934 lectures, i.e. before he uncovered the shortcomings of the method for elevating accepted sentences to analytic truths, he considers how the serious empiricist should view our knowledge. It is not a strike against him that non-empiricists will find what he says unsatisfying and old-school empiricists who have no truck with the idea of a special class of a priori truths will regard him as belabouring the obvious. He is addressing the “modern empiricist” who supplements the defensible core of empiricism with the assumption that the a priori is explicable in terms of analyticity. His claim is that empiricism rigorously construed has a consequence that tends to be overlooked (and was doubtless largely unrecognised in the 1930s and 1940s). In arguing that modern empiricism is no advance on empiricism traditionally understood, he is being reactionary. But in insisting that modern empiricists should, given their empiricist assumptions, stick with an empiricism that, as he puts it, takes there to be nothing in the mind not already in the senses, he is being revolutionary.Footnote 42
A number of other points about §6 are worth mentioning by way of conclusion. One is that the section cannot, despite how it is sometimes anthologised, be separated without loss from the rest of the paper. Quine does not set out a self-contained argument and claim to demonstrate from scratch that statements are epistemologically all of a single kind, still less does he purport to show that philosophers of a more rationalistic bent are wrong to assume the existence of two different kinds of epistemological statement. Starting from what he believes he has already shown, he proceeds, taking the empiricist’s emphasis on the primacy of experience as given, to spell out what empiricism comes to given the success of his critique of the analytic/synthetic distinction. After noting that statements in “the interior of the field” are not pinned down by “experience ... indirectly,” he declares, good empiricist that he is, that “it is misleading to speak of the empirical content of an individual statement—especially if the statement is at all remote from [experience]” (43). Then without further ado he declares it to be “folly to seek a boundary between synthetic statements, which hold contingently, on experience, and analytic statements, which hold come what may.” This, evidently, follows only given what Quine has already argued. It is, after all, hardly foolish to seek a boundary between analytic and synthetic statements while holding that single sentences have empirical content only in the context of a body of theory.
A related point concerns Quine’s so-called “holism,” the view that science is a unified whole, the sentences of which are—observation sentences aside—intrinsically interrelated. The conception of our knowledge sketched in §6 does not stand and fall with the Duhemian observation that “[o]ur statements about the external world face the tribunal of sense experience not individually but only as a corporate body” (41), and it is no objection to this conception that Carnap accepted Duhem’s observation.Footnote 43 In §6 Quine treats the Duhemian observation in tandem with what he takes himself to have already shown, namely that the analytic/synthetic distinction is “an unempirical dogma of empiricists.” Whereas Carnap takes it for granted that there exists such a distinction and regards statements within a theory (or theoretical framework) to face the tribunal as a corporate body, Quine eschews the distinction and takes “the [t]otality of our so-called knowledge or beliefs” to face the tribunal. Again the difference comes down to the question of the possibility of introducing the distinction. For Carnap, it is introducible because it figures in science and Duhem’s point applies subsequent to its introduction. For Quine, on the other hand, the distinction cannot be introduced and Duhem’s point applies across the board.
Also, it is a mistake to criticise Quine for failing to recognize other constraints on theory acceptance. He would not need to be told that many philosophers believe the underdetermination of theory by evidence can be eliminated (and the analytic/synthetic distinction grounded) by appealing to a priori principles and that it is commonly believed that the gap can be closed by invoking self-evidence, pure intuition, rational insight or the grasping of meanings.Footnote 44 He is not speaking to such philosophers in “Two Dogmas” (though he includes some paragraphs on the idea of meanings as entities in §1). He is attempting to convince “modern empiricists” for whom scientific theories are underdetermined by evidence that there is no reducing completely the distance between periphery and interior solely by appealing to empirical considerations.Footnote 45 The argument of §6 is that empiricists cannot justifiably deny that our “latitude of choice” is unlimited (as opposed to Carnap, who holds it is limited to the language of science). Accepting the underdetermination of theory by evidence and the folly of attempting to provide an explanatory account of analyticity, Quine concludes that the consistent empiricist must agree that our knowledge comprises an epistemologically undifferentiated whole.Footnote 46
Nor does Quine maintain that the account of a dogma-free epistemology provided in §6 constitutes the best argument against the analytic/synthetic distinction.Footnote 47 Subsequent to writing “Two Dogmas” in 1950, he may well have come to think his positive remarks are more compelling than the negative considerations he advances in §§1-3 (he does not return to the argument of these sections). But he does not say, or even intimate, in “Two Dogmas” itself that the distinction should be shunned because it is dispensable. Just the reverse, he argues that the distinction should be dispensed with because it is “ill-founded”. Far from claiming the distinction is uncalled for, he explicitly says: “This is a distinction I reject” (46). In his view the argument of §§1-3 (and §4) is very much to the point since it undermines the (linguistic) account of the a priori favoured by the “modern empiricist” and serves as a premise for the argument of §6. For better or worse, his argument in “Two Dogmas” is that empiricists must agree—given their conception of evidence and standards of scientific explanation—that the dogma of analyticity does not pass muster and embrace the account of our knowledge he advances in §6.Footnote 48
In retrospect it is not hard to see why Quine thought a discussion of analyticity suitable for a symposium on “Main Trends in Recent Philosophy” (the other panellists spoke on “Speculative Philosophy” and “Moral Philosophy at Mid-Century”). He would have reckoned his remarks appropriate since they concern the acceptability of a widely-defended view. Moreover I hope it is less puzzling why he suggested he was not saying anything especially questionable or shocking. From his point of view “modern empiricism” plainly does not meet the standards of explanation to which empiricists are themselves committed. On the account of “Two Dogmas” I have defended, he did not intend to mount an argument independent of substantive philosophical presuppositions, only to undercut modern empiricism from within. Throughout he accepts the insight, central to empiricism, that our knowledge is a creative construction only partly pinned down by observation and experiment (and that there is no evidence other than empirical evidence). A committed empiricist—he seems to have believed no other view deserves attention—he confines himself in “Two Dogmas” to disabusing fellow empiricists of their errors and to clarifying what empiricism rigorously understood amounts to. In 1950/1951 and for almost another fifty years, he would hold that empiricism disentangled from the dogma of analyticity (and the dogma of reductionism) is as good as it gets.Footnote 49
