In the following, I want to revisit the criticism of the analytic-synthetic distinction brought forth in Putnam’s “The Analytic and the Synthetic” and later writings and compare it with Quine’s familiar attack on the same distinction in the last two sections of “Two Dogmas of Empiricism.” My purpose in this comparison will be to ﬁnd the features of Putnam’s argument that prepare the ground for his increasingly explicit acknowledgment of the utility of a successor distinction, which sharply separates his pragmatic, induction-based epistemological commitments from Quine’s naturalistic rejection of any such project. The following brief review of some doctrinal agreements vis-à-vis logical empiricism will set the stage for specifying the issue between Putnam and Quine regarding the epistemic analytic-synthetic distinction. Following Quine, Putnam’s early writings on the analytic-synthetic dis-
tinction criticize the attempt undertaken by logical empiricism (especially Carnap and Ayer) to use a precise and general distinction between analytic and synthetic statements for empirically contentful systems of statements for the purpose of giving a “demystiﬁed” account of the special status, interpretation, empirical meaningfulness and justiﬁability of statements (such as the laws of logic and mathematical truths, statements of semantics, and theoretical statements in scientiﬁc theories) that had traditionally resisted – since supposedly independent of experience in content and justiﬁcation – ready integration into a broadly empiricist methodological outlook. The key to solving all three of the mentioned problems was to ﬁrst identify apriority, necessity and analyticity, interpret the latter as “truth in virtue of meaning,” and then to specify for any empirically contentful language a
precise class of analytic statements that gave as much traditional apriorities and necessities (which linguistiﬁed the a priori) as would be strong enough to explain the interpretability and acceptability-conditions of theoretical statements in terms of observation-statements.2 Given the underdetermined character of the latter and the multiple options in the choice of linguistic frameworks of empirical justiﬁcation, conventionalism had the remaining task of explaining the apparent necessity of logic and mathematics (as the choice to retain statements) and the acceptance and meaningfulness of nonobservational statements (as the choice of theoretical frameworks for systematizing empirical knowledge). According to conventionalism and linguistiﬁed apriority, such choices concern the presuppositions of empirical argument. Since, as presuppositions, they cannot be forced on us by empirical arguments the success of which would have to already presuppose them, their acceptability and actual acceptance have to be due to other sources. Conventionalists overcome the psychologism of their predecessors by attributing these facts of selective, empirically underdetermined acceptance to social agreements on the analyticities to adopt. Likewise, and in parallel fashion, Carnap, Lewis and Ayer believed that the acceptance and empirical meaningfulness of observationally underdetermined theories could be explained as reﬂecting collective choices as much about connecting principles between observation and theory as about rivals that, given such principles, equally well accommodate the available empirical data. Regarding the explanation of apriority and necessity, it is now our commitment to certain agreed-upon norms of inquiry that explains why the statements expressing these norms are stable under all circumstances as long as the language and framework we use remain the same. This is the thrust of a conventionalistlinguistic explanation of mathematical and logical necessity.3 Quine and Putnam both reject the idea of taking some statements (many of which are extra-logical) in science as true on merely communal ﬁat instead of empirical argument as incompatible with the empirical nature of the requirements on scientiﬁc judgment. Consequently, both oppose the linguistic conception of the substantive a priori4 as much as conventionalism. As to the ﬁrst, Putnam is as critical as Quine of the proposal to understand
epistemic special statuses, insofar as they are present in systems of empirical belief, on the model of linguistic stipulations (analyticities), changing or endorsing which is rather a matter of linguistic legislation than of empirical information and argument. On well-known holistic grounds, Quine and Putnam portray this project as an artifact of the requirements imposed by positivistic methodology that has no reasonable basis in the way statements are (epistemically or semantically) assessed in actual scientiﬁc practice. The main reasons for this stem from the inseparability of linguistic and empirical information within an empirically contentful system of knowledge. Such inseparability puts into question the idea of there being an identiﬁable subset of speciﬁc statements of the system that merely codiﬁes a special,
non-empirical kind of “knowledge of meaning.” The particular lines of thought undergirding these critical considerations are three. First, in virtue of the multiple inferential relations of beliefs within these systems, it is futile to non-arbitrarily isolate some of the system’s accepted statements as “truths in virtue of language” if they are logically candidates for revision in case of conﬂicts between the system and incoming statements. Second, it is futile to try to trace the empirical signiﬁcance or testability of statements that employ certain expressions to a given class of statements that allegedly have the sole function of specifying the application conditions of these expressions, since the application of any statement (also the meaning-specifying conditions) requires many stated and tacit auxiliary empirical assumptions which are, since required for application, equally part of the application conditions of the expressions. Third, given this inseparability and auxiliarydependence of empirical applicability, the explanation of conventional-linguistic constraints on empirical knowledge would have to portray “knowledge of meaning” as strong enough to constitute empirically signiﬁcant (hence applicable) constraints, and thus would always also articulate ordinary empirical beliefs. But then such knowledge is vulnerable to changes in empirical belief like ordinary empirical knowledge. For these reasons, we fail to obtain the supposed explanans for epistemic priority, necessity and the empirical signiﬁcance of empirically underdetermined theoretical statements. Consequently, in these writings, Putnam also agrees with Quine that it is an error to regard the ability to draw a ﬁxed dichotomy between analytic and synthetic statements as important for our understanding of the empirical justiﬁcation of hypotheses. This endorsement of the mentioned doctrines reﬂects Putnam’s commitment to two main methodological tenets that he shares with Quine, holism and the acknowledgement of the indeﬁnite extent to which even deeply entrenched (“a priori”) belief is revisable in generalized scientiﬁc terms under changes in empirical knowledge, which we could call “fallibilism” or “(indeﬁnitely) general revisability”. Both are customarily taken to undermine by themselves the analytic-synthetic distinction. But, as I mentioned above, these agreements in doctrine do not issue
in an agreement of the lessons, and it is in their positions vis-à-vis conventionalism that the diﬀerences take shape. While Quine concludes that drawing any kind of analytic-synthetic distinction within the body of our empirical knowledge is “folly,”5 and that it “has been given no tolerably clear meaning even as a methodological ideal” (Quine 1956: 132, emphasis added) Putnam claims that drawing a successor distinction “is of logical and methodological signiﬁcance” (Putnam 1962a: 249) and warns against the “danger of denying its existence altogether” (Putnam 1962: 33). At issue is the methodological signiﬁcance of an epistemic analytic-synthetic distinction, i.e. the signiﬁcance of an apriori-aposteriori distinction, given holism and general revisability.