"In Defense of a Dogma" -- first impressions

I've just finished "In Defense of a Dogma"; I've not yet read N.N.'s post. It seemed worthwhile to write up my marginal notes "fresh", so to speak. The organization is not my finest -- I have basically just gone through each note and expanded on it, and then moved to the next note when I was finished. I'd polish it up a bit more, but I leave town in the morning (visiting my sister for the weekend), and I'd like to get something tossed up before I head out. A quick glance leads me to believe this is also my longest post so far. You've been warned. The Batman panel is a jump.



First, something else. From The Journal of Symbolic Logic, Vol. 23, No. 1. (Mar., 1958), a review of "In Defense of a Dogma" by Alan Ross Anderson:

But 'means the same as,' though not peculiar to philosophers, is found by Quine equally obscure, which suggests that if Quine were asked whether 'Der Schnee ist weiss' is a correct translation of (or means the same as) 'Snow is white.' he would give some such reply as "Well, if one is true the other is, though I haven't the faintest idea whether they 'mean the same,' whatever that may mean."

I think this nicely captures a way not to read Quine: As rejecting "meaning" in the broadest sense of the term. (It is commonly noted that, ironically, Davidson takes up Quine's rejection of "meanings" while making "theory of meaning" central to philosophy.)

For we know perfectly well what Quine says to such questions: If you abstract as much as this author does, there's not an answer to the question of whether or not that's a good translation for "snow is white". If you don't tell us anything about what we're translating into (or from), then every contender for the title of "a good translation" is on all fours with the others. And just saying "I meant translating into German" doesn't entirely anull this -- there are still multiple fine ways to translate a sentence into another language. For which of the various translations we use (in Quine's terms, which of the various rival translation manuals we use to interpret utterances by the speaker we are aiming to understand) makes no difference to our practice. But the various translation manuals are rivals -- they give different translations for the interpretand's utterances, and this can extend to differing on which of the interpretand's sentencess are true and which are false, in some cases. (Imagine cases where the interpretand's utterance may be takes as either trivially false or trivially true without it mattering which; or highly abstract theoretical matters which have a minimal attachment to anything the interpretand says or does.) But if truth-values can shift, then so can "meanings" (Sinne). So to know the meaning of an utterance, one doesn't need to know which "meaning" it is related to -- that picture of "meanings" is a confused one.

From an interview with Davidson, reprinted in "Problems of Rationality", p. 257:

I started out as many people did in those days, reading Ogden and Richard's The Meaning of Meaning and Charles Morris. Now what looked like the central problem to them was to define the concept of meaning: x means y where x is a word or a phrase or a sentence and God knows what y was supposed to be -- and you wanted: iff what? That is how a lot of people were thinking about philosophy of language. Really smart people sought analyses of particular locutions, but never said anything about how you could tell whether you had come up with a correct solution or on what grounds you criticize these things apart from ad hoc arguments. So I think perhaps I felt more frustrated by this situation that I found the subject to be in than I think other people did. On the one hand, so many issues seemed rather sharp: What is meaning? How do you even think about it? Where do you start? And somewhere along the line I discovered Tarski, and I thought: you don't even want to ask the question what is meaning. It's the wrong question. It was a huge shift of perspective to get away from worrying about what it is to talk about the meaning of a predicate. Reading Tarski made me realize that there's a way to get around all that -- and somwhere along there Quine showed up...

And now to the topic of the post: In Defense of a Dogma.

Graham Priest, Two Dogmas of Quineanism, The Philosophical Quarterly Vol 29, No. 117, Oct. 1979.

I n "Two Dogmas of Empiricism"1 Quine presents essentially a two-pronged argument against the existence of analytic truths. The prongs are:
(i) There is no non-circular definition of 'analytic'.
(ii) Under the pressure of recalcitrant experience there are no beliefs that cannot be held on to, and conversely, no belief that may not be revised.
The first point is not a condemnation of the notion of analyticity. Many important concepts are definable only in circular terms. The importance of (i) is that, if any point in the circle of definitions is attacked, it is no use trying to defend it by appealing to some other notion in the circle since that is itself just as much under attack. For example, in the reply to Quine by Grice and Strawson, "In Defense of a Dogma", analytic sentences are characterized as those whose truth value cannot be revised without a change of meaning. But this will not do. For synonymy, and its converse, difference of meaning (which is obviously required to make sense of the notion of meaning change), are parts of the very circle all of which is under attack. This is the function of (i).
It is (ii) that provides the direct attack on the notion of analyticity. For it seems to undercut the whole point of drawing the analytic/synthetic distinction. If any belief can be held on to come what may, then we can conventionally refuse to let experience speak for or against any sentence we wish. The notion of analyticity is therefore vacuous.
Even if, as Strawson and Grice claim, we can sort sentences into analytic and synthetic by using paradigm examples of each, there is no point in this activity. For no important theoretical difference underpins this division. Similarly we could use observational criteria to sort substances into those containing phlogiston and those not containing phlogiston (as Priestley did), but once the theoretical underpinning of this distinction disappears, this becomes pointless.

In addition to being a nice summary of Quine's article, I think that Priest nicely handles the first section of S&G's paper, which consists of trying to make it seem that of course there must be some sense to the analytic/synthetic distinction, because there is widespread agreeement on what is and is not analytic in many cases. I think this line of argument just doesn't work. I think Priest also nicely handles S&G's closing argument -- the sense of "sense" and of "concept or set of concepts" (p. 157) they presuppose is tied in with the concepts Quine is suggesting we can do without.

For instance, take this claim from the close of S&G's article:

If we can make sense of the idea that the same form of words, taken in one way (or bearing one sense), may express something true, and taken in another way (or bearing another sense) may express something false, then we can make sense of the idea of conceptual revision. And if we can make sense of this idea, then we can perfectly well preserve the distinction between the analytic and the synthetic, while conceding to Quine the revisability-in-principle of everything we say.

The first part seems clear enough: "Snow is white" is true in the sense that this is its usual color, but false in the sense in which "Snow is frozen water" is true. And we can speak of this, if we like, by saying that "different concepts" are involved in the two cases of "snow is white", or that "is" means different things in the two cases, or that the two sentences have different "senses". This is just the banal fact that there are homonyms -- that an utterance can constitute different sentences in different contexts. But this by itself is not sufficient to make sense of the notion that there are sentences which are true due to their meaning alone; more argument is needed if we are to "perfectly well preserve" the notion of analyticity. Here is S&G's argument for the second part: "Where such a shift in the sense of words is a necessary condition of the change in truth-value, then the adherent of the [analytic-synthetic] distinction will hold that the form of words in question changes from expressing an analytic statement to a synthetic statement." -- Yes, if we could make sense of the notion that only changing the meaning of a sentence could make it false, then we would have analytic sentences! This is clearly question-begging. And the same goes for trying to use "the distinction between that kind of giving up which consists in merely admitting falsity, and that kind of giving up which involves changing or dropping a concept or set of concepts." For the shift from this distinction to the analytic-synthetic distinction only comes via the (undefended, and Quine thinks false) notion that there might be sentences which can only be given up the latter way.

In general, S&G seem transfixed by the idea that for a word or sentence to "have a meaning" is for there to exist some thing, a meaning, which stands in some relation to that word or sentence, and which might be related to other words and sentences as well. And rejection of this picture is confused with rejection of meaningfulness generally.

A fair bit of S&G's article is devoted to attributing to Quine a thesis that they admit he does not claim, but which they think he must be committed to, and then attacking the thesis: An explanation of any term in a "family-circle" of terms which are interrelated (as analyticity, synonymy, semantic rule etc. are), it is supposed Quine demands, "must be of the same general character as those rejected explanations which do incorporate members of the family-circle (i.e., it must specify some feature common and peculiar to all cases to which, for example, the word "analytic" is to be applied; it must have the same general form as an explanation beginning "a statement is analytic if and only if...")." The only reason given for attributing to Quine this suspect doctrine is that "he does not even consider the question whether any other kind of explanation might be relevant." I think the explanation for this is simple enough: No other sort of explanation for the distinction had been offered; it was always just assumed that the cluster of related notions which Quine wants to jettison were a cluster we could make good sense of at least some members of, or which we couldn't help but employ. Quine doesn't have to consider any other sort of argument because none was on the market. So S&G seem to have been too quick to attribute to Quine the suspicious doctrine.

It is worth noting that S&G fall into the same confusion that the reviewer I quoted at the beginning does (p. 146): "Is all talk of correct or incorrect translation of sentences of one language into sentences of another meaningless? It is hard to believe that it is." Indeed it is. Nobody would last a week in a foreign-language course if they held otherwise. But understanding sentences of another language need not be conceived of as a matter of figuring out which sentences of the target language "fit" the various senses I am acquainted with, and so need not stand or fall with the suspect notion of synonymy; elsewhere in the article, S&G admit that terms like "means the same as", "is self-contradictory", and "is inconsistent" do not, in their ordinary use, qualify for membership in the "family-circle" which Quine is attacking, but this does not keep them from making silly arguments like this one.

Now to S&G's examples.

To explain the distinction between "logical impossibility" and some other kind (they say "natural (or causal) impossibility", but this strikes me as a poor name for it), they consider two possible exchanges.
One:

X: My neighbor's three-year-old understands Russell's Theory of Types.
Y: You mean the child is a particularly bright lad.
X: No, I mean what I say -- he really does understand it.
Y: I don't believe you -- the thing's impossible.

At which point X is supposed to present the child, whom Y can then quiz etc., and it might so happen that Y can see that what X said is true. In this case X's initial claim is just supposed to have been taken to be "naturally impossible", such that an extremely odd child might be presented who would verify it.
Two:

X: My neighbor's three-year-old is an adult.
Y: You mean he's uncommonly sensible or very advanced for his age.
X: No, I mean what I say -- he really is an adult.
Y: Perhaps you mean that he won't grow anymore, or that he's a sort of freak, that he's already fully-developed.
X: No, he's not a freak, he's just an adult.

At which point Y is supposed to give up the notion that Y understands what X is talking about. For "whatever kind of creature is ultimately produced for our inspection, it will not lead us to say that what Y said was literally true, but at most to say that we now see what he meant." My initial reaction to this was "It should not be a logical truth that Freaky Friday is a work of fiction." It is not particularly difficult to come up with science-fiction scenarios in which a person is simultaneously a child and an adult. And a character from any of those being presented to Y would seem to show that what X said was simply true. (It is no argument that the science-fiction scenarios are all in fact fictional, since the same charge would take down S&G's first example, too -- I am confident that there never has been, nor never will be, a three-year-old who reads Russell with comprehension. Or, I am at least as confident of this as I am that no three-year-old will simultaneously be an adult.)

To summarize the difference that S&G claim to be illustrating, they claim that, upon X's incessant claims that he is speaking the literal truth, the "appropriate" response would be, in the first case, to say we did not believe him; in the second case, to say we did not understand him. As far as I can tell, there's no argument at all for this. And I have no intuition that this is in fact the proper way to talk, if that is supposed to be the support for the claim. If someone wanted to maintain that they didn't believe that a three-year-old was an adult, and that they didn't understand what could be meant by a three-year-old understanding the "Theory of Types", I don't have the slighest idea why anyone would object that they'd gotten the two backwards. I would personally be inclined to ask to see the kid, and if I don't see what I was promised to then dismiss X as crazy, in both cases.

(I think S&G might be rigging the deck on the second one with the final line: "He's not a freak, he's just an adult." This is ambiguous: Either X is denying that the three-year-old is just fully-developed, or X is now saying his neighbor's three-year-old is both an adult and not a freak. Which seems even weirder, and so might be coloring intuitions. But the response can be the same: One merely needs to imagine that not only is some three-year-old simultaneously an adult, but also that this is as common as morning dew. But why this should be supposed to be "logically impossible", rather than merely false, I don't know.)

S&G conclude their use of these examples with the claim that "If, like Pascal, we thought it prudent to prepare against very long chances, we should in the first case know what to prepare for; in the second, we have no idea." But this does not tell us anything interesting about the two cases. For the fact that I cannot imagine something does not make it impossible. Consider a third scenario:

X: My neighbor's three-year-old understands the Voynich manuscript.
Y: Now you are just making stuff up.
X: No, I mean what I say -- he really does understand it.

In this situation, I should have no idea what to expect. I can imagine what having the Voynich manuscript explained would be like in the sense that I can explain what reading a proof that there is a largest prime would be like, or looking at a picture of an invisible green sheep. (Parts of the manuscript would be pointed at, and ways to translate the apparently meaningless text would be given; various arguments would be given that anything larger than such-and-such would be a multiple of numbers other than one and itself; I would look at something and it would appear to be both invisible and green.) Now, this final comment is not central to S&G's claims, but it seems dumb enough to draw attention to.

On to the next example I found disagreeable: The green point. In "Two Dogmas" Quine writes

I do not know whether the statement "Everything green is extended" is analytic. Now does my indecision over this example really betray an incomplete understanding, an incomplete grasp, of the "meanings" of "green" and "extended"? I think not. The trouble is not with "green" or "extended", but with "analytic".

S&G claim that the falsity of this claim can be shown by replacing "analytic" with "true". But surely this is false: It is true that anything green is extended. S&G suppose that our hesitation is due to the fact that "the boundaries of application of words are not determined in all directions" -- we might not be sure whether or not to count a "green point of light" as extended or not; it seems to me obvious to respond that while there may be green dots, there are not green points in the mathematical sense. An unextended object is an invisible object. To suppose that there might be unextended green objects would then be to suppose that there might be invisible green objects. That there are no such objects does not strike me as controversial; at the least, I doubt Strawson or Grice would disagree with it. So Quine's example is one in which "we should hesitate between 'analytic' and synthetic, and have few qualms about 'true'."

S&G claim that in these sorts of examples, "no more than in the sample case does the hesitancy necessarily imply any obscurity in the notion of analyticity; since the hesitation would be sufficiently accounted for by the same or a similar kind of indeterminancy in the relations between the words occuring within the statement about which the question, whether it is analytic or not, is raised." This seems to just be false. I am fully confident that anything which is green is extended. But this is just to say that the extension of "green" is a subset of the extension of "extended". S&G would have me believe that there is some further problem troubling me that is preventing me from seeing whether or not "Everything green is extended" is analytic or not. This would be news to me. I'm not aware of any problems with "green" or "extended".

Quine, then, remains in need of an answer: Is "Everything green is extended" analytic or not? (I suspect Kant would count it as part of the synthetic a priori, claiming that the concept of "green" by itself cannot give up the manner in which that concept is applied to intuitions, and so not that it only applies to spatial (extended) ones. Whereas for Kant, analytic judgements don't need intuitions -- those is what is "synthesized" in "synthetic" judgements. "This looks like the synthetic a priori" is not a good sign that we have a clear view of things.)

(As an aside, I do think, as McDowell suggests in his first afterword to Mind and World (the long one on "Davidson in Context", p. 158), that we can make some sense of the analytic/synthetic distinction by taking analytic sentences to be those that express the "necessary structure" of any conceptual scheme (taken in a non-scheme/content sense). But this "necessity" should not mislead us -- our beliefs about what is necessary and what is not are as revisable as any others. (I take this to be part of McDowell's point when he warns that his talk of "necessity" should not be used to give us a "reassurrance" that our thought is on the right track.) To say that some sentence is "necessarily true" is just to refuse the revisions to my web of belief that would be necessary to countenance that sentence's being false (even counterfactually); this may be due to my lack of an ability to imagine what those revisions would encompass, or a simple confidence that many of the needed revisions would be bad ones, or some other reason. And any attempt to say whether or not a sentence is "germane" to revision in light of experience will depend on one's beliefs about what could constitute a possible experience -- and these beliefs, too, are revisable. So in a sense, McDowell's defense of analyticity doesn't do better than S&G's. There doesn't seem to be a point to demarcating true sentences into analytic ones and non-analytic ones by referring to the "necessary structure" of a conceptual scheme -- the distinction between the sentences which are candidates for revision and the sentences which are not is neither fixed, nor firm, but is a matter of degree and is open to debate (supposing someone can make a compelling case for it, which we can't rule out in advance -- who knows what the future holds?). So this reinterpretation of what is going on in the distinction between analytic and synthetic distinction does not rehabilitate the distinction, but merely makes it easier to see why it seems a compelling distinction: It's tied in with the fact that some of our commitments are very strongly held, others are not so strongly. But this is all sketchy and dashed-off; the defense of analyticity/necessity is a minor point in a generally excellent Afterword.)