Quine’s be Quining

Ron Bloom sends along the above and writes, “The rest of the article is just as crackling as is this paragraph.”

OK, so I went and read the article (Two Dogmas of Empiricism, from 1951), and I don’t really get it. I like the above-quoted paragraph but I couldn’t get much out of the rest of it. Maybe these ideas have just been absorbed in our thinking so they don’t seem special any more?

P.S. After I wrote this post but before it appeared, Bob’s been recommending Quine pretty strongly. Bob claims that Quine made Lakatos obsolete. I still don’t get it—but, hey, there’s a lot of things I don’t get that are still important!

51 thoughts on “Quine’s be Quining

  1. Most of the essay is concerned with rejecting the analytic/synthetic distinction. If you weren’t too concerned with this already, then it won’t make a lot of sense. This is the idea that some of our beliefs are intrinsically definitional, and others are intrinsically substantive. He’s making the case that every belief plays a role in determining the substance of our beliefs collectively, as well as the meaning of the language within which our beliefs are expressed. In some reasoning scenarios, we might designate some of our beliefs as definitions, for the purpose of getting the discussion off the ground, but that’s a choice of convenience, while the beliefs themselves in general perform both semantic and substantive roles. This is important in philosophy because a lot of other ideas that were being debated around that time took this distinction for granted, and he convinced a lot of people with this essay that the distinction itself is entirely artificial.

    • > This is important in philosophy because a lot of other ideas that were being debated around that time took this distinction for granted

      Right, this *was* important in philosophy (in the sense of importance for the sociology of the discipline). And it remains important for the *history* of philosophy for the same reason. But, arguably, Quine’s article was so convincing that it made itself obsolete; I don’t think many philosophers spend much time arguing about this type of thing these days.

      OTOH, I do think society at large hasn’t quite caught up with philosophy on this point. It wasn’t so long ago that politicians were debating the nature (not just the truth-value) of the proposition “marriage is a union between a man and a woman”. A more recent example is “[some not necessarily proper subset of] lives matter”.

      • Yes and no. There was a resurgence of these ideas associated with Kripke and Lewis. Fodor wrote a nice overview in the LRB years ago:

        https://www.lrb.co.uk/the-paper/v26/n20/jerry-fodor/water-s-water-everywhere

        My take is that Quine persuaded a lot of would-be philosophers (myself included) to go into other fields, because in rejecting this distinction he undermined the dominant methodology of philosophy in the English speaking world—namely, conceptual analysis. The people who remained in philosophy have had to cleverly come up with contra Quine takes to justify going on as before (this is what the Fodor piece argues), so there remain many who think Quine was wrong, but not too many left arguing he was right, because those who would are doing something more productive.

        • To add to the history of philosophy of science: Quine was one of several philosophers of science in the 1950s and 1960s who challenged “the received view,” the specific way of doing philosophy of science that came to dominance during and shortly after WW2. Kuhn was another one of these challengers, with The Structure of Scientific Revolutions (first published in 1962), and his critique was taken up much more broadly than Quine’s, stimulating changes in history and sociology of science. Under the influence of these critiques, by 2000 philosophers of science had generally abandoned the method of conceptual analysis, in favor of a case study method more informed by history and sociology of science.

  2. Hes talking about this:
    https://en.m.wikipedia.org/wiki/Duhem%E2%80%93Quine_thesis

    Basically in science you have a theory T that entails an observation O. Then to check it you need to make auxilliary assumptions A (equipment functions correctly, atoms/cells/whatever exist, observer is truthful, etc)

    So then you deduce:

    If T & A therefore O

    If you observe O, that is consistent with T & A both being true. But of course this is affirming the consequent. There may be other theories that explain the observation just as well.

    If you observe not O, then the conjunction T & A is negated:

    !O -> !(T & A)

    The negation of a conjunction tells you only *at least one* element is incorrect. Ie,

    !O -> !T | !A

    The consequence of this is that the central theory is “protected” by all these auxilliary assumptions that can take the blame whenever !O happens. So essentially falsification is impossible but evidence does force the scientist to adjust some assumption somewhere, which is often also an assumption used for other reasons, thus there is a rippling effect.

    • Yes, this is related to the second dogma he rejects. It’s a two-part claim, and he rejects the second part:

      (1) The meaning of our beliefs is their empirical content (in statistics we’d say the substance of our model is the predictions it makes about observable random variables).
      (2) The unit of meaning is the individual belief.

      He rejects (2) but retains (1), arguing that the meaning of our entire belief system is the predictions it entails, but that this meaning is imbued throughout the system, in such a way that no individual belief faces the test of confirmation/disconfirmation. The Duhem-Quine thesis is that idea taken to the extreme, saying that if you want to preserve any given belief in the face of arbitrarily bad predictions you can always do so by revising the rest of your model sufficiently.

      • Thanks for the informed and thoughtful comments.

        Having read mostly CS Peirce, I see much of him in Quine as say Aristotle in Peirce – no one is an island on their own. And many of their ideas do infect many researchers without a sense of where they originated or why.

        For statisticians there are insurmountable opportunities learning directly from philosophers. At least for Peirce, he has to be read widely (over years?) before much comes across. Coming to movie having missed the first 2000+ years.

        Part of the issue may be, as one philosophy colleague explained the field – we ask wildly deep almost childish seeming questions and vigorously argue over the answers like lawyers.

    • Regarding the Duhem-Quine thesis, I would recommend reading Duhem’s writings on the topic. He is a crystal clear writer, even if his focus is narrower than Quine’s. Quine I find entirely incomprehensible.

      • Can you hit on the keypoints to take away from it besides the above? I was introduced to all this via Paul Meehl and haven’t seen anything to think his take was off, so my take pretty much follows his. Eg:

        (1997) The problem is epistemology, not statistics: Replace significance tests by confidence intervals and quantify accuracy of risky numerical predictions. In L. L. Harlow, S. A. Mulaik, & J.H. Steiger (Eds.), What if there were no significance tests? (pp. 393-425). Mahwah, NJ: Erlbaum.
        http://meehl.umn.edu/sites/meehl.dl.umn.edu/files/169problemisepistemology.pdf

  3. I suggested to Andrew that he read Quine if he wanted to understand why philosophers abandoned empiricism. It has the beauty of all great philosophy, which is dissolving a problem on its own terms. I believe I told Andrew that Quine hoisted the empiricists by their own petard. As Sam aptly summarized above,

    Quine’s article was so convincing that it made itself obsolete; I don’t think many philosophers spend much time arguing about this type of thing these days.

    Exactly. It was that influential. It dissolved a major branch of philosophy and paved the way for a more pragmatic philosophy of science. Where in the world is Keith on this one? To me, the line of thinking goes throgh Peirce and the later Wittgenstein to Quine.

    I also suggested that most practicing scientists would tell you they were logical positivists but would behave like pragmatists. I think this is very much the same way they would tell you they were frequentists but would behave like Bayesians.

    I’m afraid I don’t know Lakatos beyond what I read in the Stanford Encyclopedia of philosophy article. From what I gathered there, he seems to be trying to maintain elements of Popperian philosophy of science while admitting Kuhn may have had a point. That’s a rather tough row to hoe.

    Andrew’s contributions on philosophy of stats have been of a more methodological nature and not really something that looks like philosophy of science to me. As I told Andrew, it’s the same kind of theoretical/applied divide we see in stats, or in applied math or physics vs. theoretical math. When you move from calc to real analysis, you learn that all those definitions you worked with had tricky side conditions and didn’t quite generalize well enough. Further, you learn all kinds of counterxamples to things that feel like they should be right. In logic and set theory, where I’m more comfortable, that’d be things like Russell dissolving Frege’s concept of set comprehension with the liar’s paradox, Cantor’s illustration of multiplie infinities, or Gödel’s proof that a logic strong enough to state arithmetic is necessarily incomplete, or Church and Turing’s reformulation of Gödel in terms of the halting problem. All of these things show that our simple working notions are wrong. But in the end, when I wax philosophical about how to write computer code, I’m not worrying about Cantor or Gödel—I’m doing applied philosophy or what I might call methodology.

    I also told Andrew that I don’t know this philosophy of science material well enough to participate in a public workshop on it! Philosophy is really hard.

    P.S. My own background on all of this is through the analytic philosophy tradition in language—I used to work on natural language semantics and non-standard logics and co-teach the philosophy of language class at Carnegie Mellon with Teddy Seidenfeld. My appointment was in philosophy because that’s where linguistics was.

    • Lakatos basically had a bayesian philosophy of science. You can’t prove something true, or falsify it.

      What you can do is make otherwise surprising predictions that come true. That means the predictions must be precise enough to distinguish one theory from any others (this is found in the denominator of Bayes rule).

    • “I suggested to Andrew that he read Quine if he wanted to understand why philosophers abandoned empiricism.”

      Philosophers abandoned empiricism? Do you mean they abandoned logical positivism? Surely, plenty of contemporary philosophers have an empiricist outlook.

      • “Philosophers abandoned empiricism? Do you mean they abandoned logical positivism?”

        To be precise, by “empiricism” in “Two Dogmas of Empiricism” Quine meant Carnap’s views in Logische Aufbau der Welt. Other members of the Vienna Circle already held views very much like what Quine was arguing for in “Two Dogmas,” most notably Otto Neurath. (Readers of Quine might remember “Neurath’s boat.” That was a metaphor that Neurath was using before Quine visited Vienna.) But Neurath died suddenly in December 1945, and his outspoken Marxism meant that his views didn’t survive the depoliticization of philosophy during the McCarthy era.

        I would expect that almost all philosophers of science working today would consider themselves “empiricists,” though being philosophers they would mean many different things by that term.

    • It seems to me that very often when people say something is “behaving like Bayesian” (such as Bob or Anoneuid above) they just mean that models are not believed uncritically and ignoring the background. This is no more Bayesian than common sense, and it certainly is not “non-frequentist”. No frequentist is committed to blindly believing their models let alone ignoring background information, and they don’t have to be “Bayesian” to question the models and to act “pragmatic” in Bob’s sense. As long as you’re not computing posterior probabilities in a Bayesian manner and interpret them accordingly, you’re not a Bayesian; claiming anything else is just what I call “Bayesian propaganda”.

      • Lakatos’ main idea is that science progresses when new theories make otherwise surprising predictions that come true. This is literally what Bayes’ rule tells us to do. Compare what is predicted when H[0] is true vs if it was not.

        Bayes Rule (H[i] = hypothesis i, D = data):

        p(H[0]|D) = p(H[0]) * p(D|H[0]) / p(D)

        Where:

        p(D) = p(H[0]) * p(D|H[0]) + sum( p(H[1:n]) * p(D|H[1:n]) )

        If the data is much more likely under H[0] than H[1:n] then the second term in the denominator becomes negligible and can be dropped.

        Also, precise predictions are much more convincing because the the likelihood will be much more concentrated. If you hypothesis is “god works in mysterious ways” then that explains pretty much any data and the likelihood is low everywhere.

  4. Quine’s Two Dogmas piece has to be understood in the context of Carnap (in Quine’s own words, his most important mentor) and the Vienna Circle’s project of creating a Scientific World view. Carnap’s project was to reconstruct the language of science into a formal language in which agreed with the principle of verification, i.e. all meaning consists of verifiable results. People often, as have commenters here, take Quine’s essay to be a refutation of empiricism. That may be true of some versions of empiricism, for instance, Berkeley’s empiricism. But, Quine was engaged in the same project that Carnap was, and he follows up this essay with a book, Word and Object that is his attempt to accomplish just what Carnap is also trying to accomplish. Both are using mathematical logic to reconstruct a language of science. In Quine’s ultimate attempt observation sentences are those sentences that speakers of the language invariably affirm when faced with the same stimulus. In rejecting the analytic/synthetic distinction, Quine admits that there is no ultimate distinction between the logical inference rules of the language and the so-called empirical content. I always have the “choice” to change the rules of inference in the face of an unexpected observation instead revising my other beliefs. Carnap did think that he could meaningfully define analyticity in a metalanguage in terms of logical truth in the object language. Tarski, Godel, and Quine each expressed skepticism that Carnap’s approach to analyticity, but I am not aware that any of the three wrote a detailed refutation of Carnap on that point. Still, the difference between Carnap and Quine is not large, and both can be seen as fulfilling the Vienna Circle’s project of upholding Verificationism. Often the criticisms of the Vienna Circle miss the point (at least with respect to Carnap), the point is to have a logical reconstruction of language in which the meaning of our language is rooted in verifiable observation. For Quine, the meaning of the language is rooted in observation although the ontology of the language may surpass the observational content. To the extent it does, the ontology is indeterminate. For Carnap, the ontology of the system can be discussed in a metalanguage in terms of the logical syntax of the language, but the choice of those rules is ultimately a choice of framework, which itself is not subject to any empirical test. So, contrary to the conception that Quine refuted Verificationism of the Vienna Circle, his work supports it. As does Carnap’s. But, it is important to see that is both Verificationism is a choice. Quine’s system rejects analyticity, and renders meaning beyond the observational content indeterminate. Carnap preserves meaning determinacy, but makes the choice of a framework a pragmatic consideration. Neither clearly articulates exactly why Verificationism is the choice that should be made. For that I think we have to go back to Pierce, who unlike Quine and Carnap is willing to engage in moral discourse. For him, I believe, verificationism is a moral choice. You cannot be logical unless you identify your interests with the interests of all humanity. We should choose a scientific world view because it is the only one capable of inevitably making progress over the long course of history.

    • We should choose a scientific world view because it is the only one capable of inevitably making progress over the long course of history.

      Hard to read the large block of text but brute force trial and error (eg, “evidence based medicine”) can also accomplish this. It proceeds frustratingly slowly, but it will progress eventually.

    • It’s one thing to ask how philosophers really meant something, in what tradition they stood etc., and to ask what the consequences of their work actually were, including interpretation by people who read just what they wrote without engaging with in what discussion they were, whom they referenced even where not explicitly referencing them etc. I think both ways of reading have their justification, and as far as you are right, one could maybe say that Quine helped much criticism of verificationism and empiricism even if that may not have been his intention. But this doesn’t mean such reading is wrong; Quine may very well have inspired valuable thought different from what he wanted to inspire, but so be it. For readers like me who are more interested in how they can learn from these authors for their “personal philosophy” than in what tradition they stood and what they “really meant”, being inspired in this way may be the best thing to get from the authors.

  5. “The rest … is just as crackling …” I thought that was sarcasm. Didn’t Socrates say it much more clearly and briefly? The only thing we know is that we know almost nothing for certain.

    • Well, not quite as crackling as the preamble. Here’s what I get from the article: the statement that there is in fact an analytic/synthetic distinction is itself *synthetic* — i.e. an empirical observation. I attempted once to write up such an argument in a Philosophy 101 paper and the damn TA gave me a C-. I will never forget the scrawled comment, “You write much and say little!” Man did that burn!

  6. It seems to me that Quine is important if you think of physical laws (science) as representing truth and being real; and that he’s “antiquated” if you think of physics/science as an incomplete model (or even a set of models) of the world that is nonetheless very useful in certain respects.

    On an individual level, it’s also psychologically true that a person’s fabric of belief isn’t determined by experience; I’ve spent a few months debating Flat Earthers, and there is no experience you can show them that’ll “force” them to revise their central belief, even though it firmly rests in the realm of physics and geography, and they embrace empiricism.
    The same goes for the “stop the steal” people. The philosophical nature of truth is suddenly politically relevant.

    • On an individual level, it’s also psychologically true that a person’s fabric of belief isn’t determined by experience; I’ve spent a few months debating Flat Earthers, and there is no experience you can show them that’ll “force” them to revise their central belief, even though it firmly rests in the realm of physics and geography, and they embrace empiricism.

      The funny thing to me is that technically the Earth is flat according to General Relativity.

        • Eg:

          Point #1 is actually straightforward to explain: objects simply travel on the straightest possible paths through spacetime, called geodesics. The paths only seem curved because of the warping of spacetime.
          […]
          Now, I mentioned that spacetime needs to be warped in order for objects’ trajectories to appear curved to us despite them actually being “straight.”

          https://physics.stackexchange.com/questions/3009/how-exactly-does-curved-space-time-describe-the-force-of-gravity

          So, according to GR a satellite orbiting the Earth is travelling on a straight path through spacetime. Hence the earth it is travelling over must be flat.

        • You’re taking this too far for a joke.
          Put a gyroscope on a satellite, you’ll find the orbit is curved and not straight.
          “Inertial” does not mean “straight” any more than hydrostatic “level” means “flat”.

        • It’s not a joke. Either you can accept the Earth is flat or you can reject General Relativity (which makes you some kind of “science denier” I guess).

          Personally I think GR is wrong and believe in a round Earth.

        • The object travels on a “geodesic” in spacetime. If we want to call a geodesic a “generalized straight path” then it is indeed “straight”. But only insofar as it is clear that the term “straight” is being used in an extended sense!

        • Even if we take your notion of the meaning of the word flat to be true, the fact is that different people experience the universe in different ways due to their frame of reference, both location and speed matter, and all are equally valid. This is the meaning of relativity…

          So “the earth is flat per Anoneuoids definition” is a true statement exclusively for those people at the surface of the earth traveling at orbital speed. Which is zero people ever in the history of the universe.

          So, no the earth is not flat to any actual person

        • Actually the meaning of relativity is that you can choose your reference frame. It is a matter of preference.

          You can choose for the Earth to be the center of the universe if you want. It just makes the calculations commonly done by humans more complicated, so no one does it.

        • GR has nothing to say about the “shape” of the earth as a distribution of mass as we perceive it with our ordinary 3D vision at a snapshot in time. It could be a ball or a doughnut or a plate. Perhaps it is a doughnut after all. But GR is indifferent to that. What GR attends to is the geometry of the path in 4D space time which may be identified with the “history” of the earth, as it makes its traverse through the starry void. The earth as a distribution of matter traces out a world-line in 4D spacetime; and so far as GR is concerned that world-line (excuse the pun) *is* the earth — in the representation provided by the GR model. The GR model expresses the geometry of such world-lines in general by a principle of least-action, consistent with the GR field equations; and the intrinsic geometry of of these “paths” (in the GR theory) is an emergent property of the overall distribution of mass in the neighborhood locally. The curvature of these paths (which are submanifolds of 4D space-time) can be thought of “intrinsically” or “extrinsically”. The “intrinsic” view speaks of the curvature of space time; the extrinsic view speaks of the influence of the gross distribution of mass insofar as the curvature of space-time is an epi-phenomenon thereof.

        • GR has nothing to say about the “shape” of the earth as a distribution of mass as we perceive it with our ordinary 3D vision at a snapshot in time.

          Correct. In reality, ie 4D spacetime (according to GR), the earth must be flat since objects orbiting it are following straight lines through spacetime.

          Just because it appears curved to humans with their limited senses, does not make it curved.

        • Anon, you say “4D spacetime (according to GR), the earth must be flat since objects orbiting it are following straight lines through spacetime.” This is wrong. If the earth were shaped like a pyramid, objects orbiting it would still follow geodesics (which you are defining to be ‘straight lines’).

  7. It’s not often that everyone’s favorite part of a piece of writing is the last section: Camus’ mom died in the first sentence, where we also learn that all families are miserable in their own way and to call him Ismael; Amendments 1-10 get all the attention and the Cogito is only the second Mditation, but does anyone even remember how many Amendments or Meditations there are total? Has anyone ever gotten past the first volume of Das Kapital? Anyway….

    The majority of that essay, focused as it is on the “analytic/synthetic” distinction, is a little boring and feels weird, probably for the reasons Bob describes (it just obliterated the thought process, so it is gone now – which, crazy!). But this last section is about something else – after tearing down what came before (logical positivism, falsificationism, historicism), he wants to provide something new. And that is what I think endures about this paper today – that little glimmer of insight or unconcealment about the nature of knowledge and knowing.

    I really love this “wholism” as an over-arching metaphor for knowledge. A big web, in which statements are linked together, and the more statements a logic or a fact hold in place, the more central it is to our understanding; the more it is supporting, the harder it is to cast off without losing a lot of other things we thought we knew. And so get three improtant things: you have a realization that much “knowledge” is contingent on other things (which can’t ever be proved or known simply empirically); a sociology or metaphsyics or knowledge that determines which statements will get adjusted when facts conflict with things we thought we knew (the least important ones, the ones which support the fewest nodes of the knowledge-web); and a rich metaphor which allows the concepts of “knowledge” to include and encompass everything from Ptolemaic astronomy in its time/place to quantum theory in it’s. Very pragmatic, as Bob says.

  8. Andrew, you are right, Quine’s essay is stupid from beginning to end.

    Most of it is devoted to silliness such as this: Bachelor means unmarried man, and “Bachelor has fewer than ten letters” is a true statement, but substituting “unmarried man” for “Bachelor” gives a false statement.

    This tells us nothing about empiricism. It only tells us something about the field of Philosophy that this garbage was taken seriously.

    • Roger:

      Quine’s statement may appear trivial at first blush, but the implications for inference in the social sciences is quite important. He is telling us that the substitution property allowed by the equivalence of two numerical variables [A = B] such that any function including A could replace all instances of A with B and produce the exact same result. This property cannot be extended to variables containing semantic meaning, such as those used in many of the social sciences.

    • I’m not sure you are agreeing with Andrew. Although Andrew may think Quine’s essay is stupid, that’s not what he said. What he said was that he didn’t get it, which makes sense. It’s unreasonable to read a (pretty advanced) paper in a discipline you’re not that familiar with and expect to understand very much of it. To properly understand Quine’s paper I think it definitely helps to have a solid foundation in formal logic and the history of (analytic) philosophy. I’ve assigned this paper for my classes several times, but I’m still discovering new things about it every time I read it. That’s the hallmark of a good paper. However, the first time I read it, I didn’t understand very much of what was going on or what the big deal was supposed to be.

      • Roger, Olav:

        I don’t think Quine’s article is stupid! I don’t get Quine’s article, but I’ll defer to Bob and others and just say that the Quine article is part of a whole literature that I haven’t read. Lots of things are hard to understand and appreciate in isolation but can be valuable when understood in context.

      • No, Quine’s is not advanced, and does not require formal logic. The logic parts are not at all interesting.

        The paper has been described as “sometimes regarded as the most important in all of twentieth-century philosophy”! Andrew cannot read the paper and see why it is important. I think that is because there was a trend for philosophers to deny objective knowledge. Statements like 7 + 5 = 12 and “a man is a creature with a kidney” seem to be objectively true, but philosophers like to try to explain them away somehow. Quine was considered a hero to those philosophers because he attacked objective knowledge. That is why the paper is praised.

        But the paper itself is a silly straw man attack on some philosophical opinions. It does not really refute anything about logic or empiricism or reductionism or anything else.

        • “Quine was considered a hero to those philosophers because he attacked objective knowledge. That is why the paper is praised.”

          That doesn’t land a punch anywhere even in the ring; the target isn’t Quine that’s standing there; it’s the shadow of some other academic bugbear, decorated with a Quine caricature, like those stupid grinning Charles I masks that were fashionable in the crime-thriller movies in the 80s.

        • This is what I get from the article.

          [1] Our understandings of logic (its internal relations, its tautologies, theorems) are on par with our understandings of all things
          [2] Put another way, the fact of the analytic/synthetic distinction is really a “synthetic” judgement. The fact that there is such a distinction is not a truth of logic. (One might say it can be re-cast as a truth of meta-logic; but then *that* fact of the matter becomes a truth of one order higher yet).

        • Quine’s paper has nothing to do with denying that there are objective truths. Quine is arguing against the analytic/synthetic distinction, but that’s completely consistent with there being objective truths since both analytic propositions and synthetic propositions can be objectively true (i.e. true in a way that is mind-independent, not determined by social convention, etc.)

          And Quine is certainly not trying to refute empiricism in his paper—quite the contrary. Nor is he trying to refute logic.

        • Where I do lose patience (or concentration) with the paper is in the section on the problems of “synonymy”. All the smoke and steam about the pitfalls of mapping equivalent truths about the world to equivalent sentences (or not). The evening star vs. the morning star. Scott and the author of Waverly and so on. I think all that is going on here is that the obsession with precision is pushing up against the limits of the informal language (English) especially in respect to the unbounded universe of “things”. When the two come together there is a clash; the equivalence class of things denoted by different expressions is hard (or impossible) to define within the object language — especially when the object language is informal, and the universe of things to be denoted is neither finite nor recursively denumerable. I believe the whole set of paradoxes alleged to inhere in ‘synonymy’ collapses when the exercise is done carefully and tediously in an artificial universe of ‘objects’ and a computable schema of descriptive predicates which cover some or all of the objects in a many-to-one fashion (thus ‘synonymy’ emerges). But I was never persuaded by all that stuff about ‘the morning star’ and ‘the evening star’ that synonymy was out-of-our-reach!

Leave a Reply to Christian Hennig Cancel reply

Your email address will not be published. Required fields are marked *