Monday, June 16, 2014

How Not to Read/Teach “Two Dogmas”
In 1976, when I delivered the John Locke Lectures at Oxford, I often spent time with Peter Strawson, and one day at lunch he made a remark I have never been able to forget. He said, "Surely half the pleasure of life is sardonic comment on the passing show".  This blog is devoted to comments, not all of them sardonic, on the passing philosophical show.
Hilary Putnam

Surfing the web, I find that two very wrong ideas about Quine’s “Two Dogmas of Empiricism” are (almost) ubiquitous. Many philosophers, including some who do not make serious mistakes about any other twentieth century philosopher, tell their readers that “Two Dogmas” defends  two forms of holism: “meaning holism” and “confirmation holism”.  This post has two purposes: to explain why these are misinterpretations, and to speculate about the possible reasons that many philosophers (including, no doubt, many teachers of courses that discuss “Two Dogmas”) impose these misreadings on that famous text.
But first a disclaimer: I do not think that “only Hilary Putnam understands Quine, and everyone else gets him wrong.”  The best books on Quine (as opposed to many articles and PDFs I see on the web) do not make these mistakes. For example, Peter Hylton’s indispensable book, Quine, does not mention “confirmation”—which is as it should be, because, as I pointed out in a previous blog, Quine mentions “confirmation” in “Two Dogmas” only to say that, “apart from prefabricated examples of black and white balls in an urn”, we no more have a theory of confirmation than we do of analyticity!  Since the whole point of “Two Dogmas” is that “analytic” is a term we should stop using in philosophy, the reason Quine compares the two problems, the problem of explaining analyticity and the problem of explaining confirmation, has to be that we should no more employ the notion of confirmation than we should the notion of analyticity. Quine is not a  confirmation holist; he is a confirmation rejectionist. (See also my post on Quine and Popper). 
As for “meaning holism”, I have noticed something peculiar: I have noticed that whenever a passage from “Two Dogmas” is quoted to show that Quine is a “meaning holist”, the passage isn't ever about meaning. What is cited are passages that display Quine’s “Duhemianism”, that is, the view that our bodies of scientific theory confront “recalcitrant” experiences as wholes. This is a form of holism—call it “evidential holism”, but it doesn’t concern meaning. (Why the evidence relation isn’t a justification relation for Quine, I shall come to shortly.) This is peculiar, because no one, to my knowledge, thinks that Duhem was a meaning holist!  Indeed, since the demise of the early logical positivist claim that empirical statements can be verified (and not just confirmed or disconfirmed) by sense-data [a claim that Carnap abandoned in Testability and Meaning,  and arguably already in Logical Syntax of Language], it has become hard—perhaps impossible—to find a philosopher who does not agree with what Quine sometimes called “moderate holism” with respect to the relation between theory and evidence. But no one says that all these philosophers are “meaning holists”.  So why is Quine, and Quine alone, read in such a way that if he is a Duhemian, he must be a “meaning holist”? Just as he is a confirmation-rejectionist  and not a confirmation holist, Quine is a meaning-rejectionist and not a meaning holist.
 The source of the misreading.
I believe that the source of both misreadings is a refusal to believe that Quine’s views are as radically different from those of traditional epistemologists and traditional philosophers of language as they are. Philosophers who have written serious books defending or criticizing Quine (or partially defending and partially criticizing him), in particular Peter Hylton and Christopher Hookway stress this radicalism; so it seems too me that people who make these mistakes must have failed to avail themselves of the insights of these major Quine scholars. And at the heart of what am calling Quine’s philosophical “radicalism” is this: the evidential relation, for Quine, is not a justificatory relation. When Quine tells us that the “evidence” for scientific theories comes from experience, be means that the causal starting points in the human construction of science are  impacts on our senses (after Word and Object: impacts  on our sensory receptors). Some beliefs are “argued for probabilistically” [Quine in Pursuit of Truth—see my post on Quine and Popper], and sometimes even deductively, given a background theory (“the lore of our fathers”)—for example, given a lot of theory, a scientist can say that the BICEP2 [] experiment’s significance level is 5.2 sigma—but there is no such thing as our scientific theory as a whole being probabilized by observations alone. If you want a “foundation” forget it.
If one makes the mistake of taking Quine’s talk of evidence and of accepting statements and of revision as talk about confirmation and disconfirmation, it easy to see how one will misread him as a “confirmation holist”. The mistake of reading him as a meaning holist is somewhat stranger, but here is my conjecture: “Two Dogmas” criticizes “the verification theory of meaning”, by arguing that there is no such thing as the method of verification (or confirmation/disconfirmation) of an individual sentence.  The philosophers I am referring to must be assuming that Quine cannot be so radical as to reject the notion of meaning altogether; so, they must think, he must mean that it is sufficiently large bodies of theory (perhaps total science) that have meaning and methods of verification together!  But this is a double error.
I have already explained why bodies of theory are not confirmed by observation alone, for Quine: because the notion of confirmation is one he wants us to throw in the trash barrel along with the notion of analyticity. (Quine does tell us that given background beliefs, we can justify many statements deductively, and given probability theory and the like, sometimes probabilistically.) And similarly with the notion of “meaning”: one of the main claims of “Two Dogmas” (and of Word and Object and subsequent publications) is that there are no acceptable [to Quine] identity-conditions for “meanings”.  Yes, there are “translation manuals” (Word and Object), and the purpose of Word and Object is to show how communication (speaking with members of one’s community as well as translation of alien languages) is possible without positing such entities as “meanings”, indeed, without going beyond Fred Skinner’s behaviorist account of “verbal behavior”.  But there are no “meanings”, neither of single utterances nor of whole theories. In short, Quine was already practicing “naturalized epistemology” (and language theory) long before he wrote “Epistemology Naturalized”.
In conclusion
I agree with Hookway [Language, Experience, and Reality] that Quine’s three fundamental commitments are “empiricism, scientism, and physicalism”. Although I do not share those commitments, each of them is important, and Quine, like David Hume, was a great philosopher because he was willing to point out the radical consequences of his own commitments, rather than hide them or deny them. One does no favor to either Quine or philosophy by making him less radical than he is.


  1. Very important topic.

    Hilary Putnam: "I have already explained why large bodies of theory are not confirmed by observation alone, ... Quine does tell us that given background beliefs, we can justify many statements deductively, and given probability theory and the like, sometimes probabilistically."

    Amen! Two observed facts can produce a ‘line’ of truths. Many observed facts can form a ‘network’ of truths. Every truth (observed or not) which is produced by the ‘network’ can be the ‘evidence’ to support or reject a new theory. A single truth is only a member of a huge truth-‘network’. The science is judged by the network. Thus, most of the ‘evidence’ needs not to be an observed fact.

    Hilary Putnam: "… communication (speaking with members of one’s community as well as translation of alien languages) is possible without positing such entities as “meanings”, ..."

    I have a major problem with this statement. What does ‘meanings’ mean? What is ‘communication’? Is making love a ‘communication’? What is the ‘meaning’ of making love? Enjoyable act? A ripe? Will produce a baby?
    The above statement is nonsense. Rejecting ‘meanings’ in communication (of any kind) is ever worse nonsense.

  2. Dear Professor Putnam.
    Isn't there one quote which at least suggests meaning holism rather than a peculiar kind of meaning eliminativism?' 'The unit of empirical significance is the whole of science'. If empirical significance comes in units then presumably there is such a thing. And if the unit in question is the whole of science doesn't this suggest holism, at least about empirical significance, if not about other kinds of meaning (which are perhaps ready for the chop)?
    Charles Pigden

  3. PS. Another thought.

    Of course one might reason as follows.
    1) Meaning has to be molecular to account for our capacity to manufacture new but meaningful expressions.
    2) Epistemology has to be holistic because of the Quine/Duhem thesis.
    3) Therefore epistemic theories of sentence-meaning are radically mistaken.

    This seems fine to me but it presumably was NOT what Quine had in mind.

    Could Quine have been thinking something along the following lines?

    1) Meaning to be a respectable a concept would have to have an empirical analysis.
    2) An empirical analysis of meaning would have to be holistic. (Quine/Duhem)
    3) But meaning has to be molecular to account for our capacity to manufacture new but meaningful expressions.
    4) Therefore meaning cannot be given an empirical analysis.
    5) Therefore meaning is not a respectable concept and ought to be dispensed with.

    This reconstruction would make meaning holism (or holistic verifcationism) a way-station on the road to meaning eliminativism, but not Quine’s final destination.


    Charles Pigden (That Popper/Quine post was really interesting!)

  4. PPS

    And another thought which perhaps makes sense of both Quine’s quote from ‘Two dogmas’ and his aged paean to Popper. Here’s a way to reconstruct Quine’s argument.

    1) For anything to have a meaning in any serious sense it must have an empirical meaning, that is to say verification-conditions.
    2) But individual sentences do not have verification conditions (Quine/ Duhem)
    3) So if anything were to have meaning in a serious sense, it would have to be an entire theory. (’The unit of empirical significance is the whole of science')
    4) But experience cannot confirm – that is verify – entire theories (Popper)
    5) So NOTHING has a meaning in any serious sense.

    This gives us meaning-eliminativism via 1) verifcationism wrt meanings, 2) Quine/Duhem wrt to the non-existent verification-conditions of sentences, 3) holistic verificationism as to-be-dismissed fall-back position for the reality of meanings and 4) the Popperian rejection of theory confirmation.

    Since meaning-eliminativism is absurd (some things mean something) and points 2) & 4) are correct, the way to reject the conclusion is to reject premise 1). Utterances don’t have to have verification-conditions in order to be meaningful in some serious sense.

  5. Meanings, insofar as they can be expressed as definitions, for Quine seem to be partial translation manuals from a language to itself, and the poor identity conditions of meaning are a result of the fact that "incompatible" translation manuals can agree on all stimulus-synonymous equivalences. So the conditions under which one observes a rabbit are identical with the conditions under which one observes an undetached rabbit part. Quine's tendency to eliminate talk of meanings is ultimately pragmatic: to reweave the web in light of new observations we can improve our powers of prediction and control by shifting from an old auto-translation manual to another, and insofar as this is the case, such a shift is rational. If meanings exist as determinate entities which are fixed and unalterable, such a shift would not be possible, much less rational.