Real goldfinches and robot cats

Daniel doesn't like what John Haugeland says in "Objective Perception." In the former's words: "The very idea of giving a "constitutive ideal" for "thinghood" strikes me as inadvisable." Yet it seems that we can always try to say what we mean by "thing," such that if (Ex)(x lacks some property p), then x isn't a "thing" after all. After giving some examples, Daniel admits that:

[Everything in this paragraph seems like an overwrought version of Austin's bit about the finches that suddenly explode etc., and what we should say about them. I can't recall where that passage is. I need to read more Austin.]

Well, everyone needs to read more Austin. Here's the quote from "Other Minds". Even in special cases (of deciding "whether it's real"), "two further conditions hold good": first, that it's not true that just "because I sometimes don't know or can't discover [e.g. because it flies away], I never can." The second is that "'Being sure it's real' is no more proof against miracles or outrages of nature than anything else is or, sub specie humanitatis, can be. If we have made sure it's a goldfinch, and then in the future it does something outrageous (explodes, quotes Mrs. Woolf, or what not), we don't say we were wrong to say it was a goldfinch, we don't know what to say." [Philosophical Papers, p. 88]

However, he also goes on to say that "It seems a serious mistake to suppose that language (or most language, language about real things) is 'predictive' in such a way that the future can always prove it wrong. What the future can always do, is to make us revise our ideas about goldfinches or real goldfinches or anything else." [88-89]

This is almost right, but it makes it sound like the case is asymmetrical: that the future can't always prove our beliefs false (rather than our "ideas"), but that our "ideas" are always vulnerable to (forced?) revision. What I would rather say is that both our beliefs and our meanings are corrigible, which nicely combines the ideas that a) beliefs are corrigible in the light of further experience, and b) the interconstitutive nature of belief and meaning implies that the same is true of meaning. While it may be natural in any one case to do one rather than the other, ultimately the choice is up to us. Neither "the world" on the one hand nor "language" on the other (as it seems some want to say) can determine our choice unilaterally.

I've always taken this to be the moral of Putnam's robot cat example. If those things turn out to be robots, then we have two choices: we can say either that a) the supposedly analytic and thus unrevisable sentence "cats are animals" has, mirabile dictu, turned out to be revisable after all, as cats have turned out not to be animals after all, but are in fact robots; or b) that "cats are animals" remains analytic, but that, mirabile dictu, it seems that there are no cats among us after all, as those things we thought were cats have turned out to be robots instead. I don't remember, but I think Putnam himself may have claimed that we must say (a) here, but it sounds better to say instead that we are not forced to say (b) (i.e. due to the incorrigible qua non-empirical analyticity of "cats are animals"), but can say what we like.

In general, my motto in such cases is that when something unutterably weird happens, it may be that whatever we say will sound unutterably weird, which means that examples like Swampman (or Twin Earth, or grue, or whatever) are nearly always not worth it – if there's really a point there (beyond what I just said), you can make it better some other way.

Mea culpa, mea methodologica culpa

In my post the other day, I made an interesting slip (if that's how you want to think of it): I suggested that Putnam's claim that analyticity and a priority come apart (so that the first four sections of "Two Dogmas" can be detached from the last two) might be of some use to defenders of analyticity. They might want to argue, I thought, that if your target (qua "metaphysics") is really the a priori/a posteriori distinction, then it might be better not to identify it with the analytic/synthetic one (and get rid of them at the same time), but to distinguish the two, so that we might not simply keep around the presumably now unoffensive (qua non-metaphysical, once so distinguished) notion of analyticity, but also employ it (for the project of linguistic analysis) to combat more metaphysical notions (like the a priori).

But (as I noted in a subsequent comment) that just assumes that the defenders of analyticity might see the a priori as unacceptably metaphysical where analyticity is not. As it turns out, Hacker at least does not. I'll get to all that in a minute. Let me first give a quick and dirty characterization of four similar concepts, not worrying for the moment about whether any one of them can be collapsed into any of the others, or whether there really are any such things.

1. Tautologies are "truths of logic": P or not-P (in classical logic).

2. Analytic sentences are "truths (by virtue) of meaning": That bachelor over there is unmarried.

3. Truths are known a priori when we don't have to go out and look, but can confirm them from the proverbial armchair.

4. Truths are necessary when it is impossible for them to be false (they're "true in all possible worlds").

If you like these concepts, you can supply your own examples for the last two. (The SEP article on "A Priori Justification and Knowledge" has as an example of a necessary proposition this one: "all brothers are male," which is not one I would have chosen if I were trying to distinguish necessity from analyticity). Anyways, my point is that however the categories do or do not overlap, the characterization of each has its own typical angle: tautologies have to do with logic, analyticity with meaning, a priority with knowledge (and justification), necessity with ontology (or modality, or in any case metaphysics).

A lot of us want, in some sense or other, to rule out "metaphysics" as nonsense, e.g. a) Ryle, Hacker, etc.; b) Wittgenstein (early and late, on most interpretations); and c) some but not all naturalists. So necessity (or, redundantly, "metaphysical necessity") looks fishy to us. But (as I started to talk about before) in order to combat metaphysics (including but not limited to "necessity"), some of us think we need to hold on to analyticity – a concept which deals, the thought goes, not with the world (i.e., on the other side, qua the object of a "metaphysical" statement, of the "bounds of sense"), but with meaning (which is safely on "this" side). Or so I read Grice & Strawson (I'm trying not to make a straw man here!). For G & S, then, analyticity is both unobjectionable and necessary uh, required for the project of finally exorcising our metaphysical demons. (I assume, if perhaps I shouldn't, that no one has a problem with (the very idea of) tautologies.)

Where does that leave the a priori? If we assimilate it to necessity (on the one side), then it's a metaphysical notion, worthy of dismissal; and if analyticity is the "least metaphysical" of the three (on this quick and dirty characterization), then if Quine's attack on analyticity goes through, it seems that a fortiori (so to speak) the others go as well. But for "analysis" to be possible, G & S believe, there need to be such things as "analytic" truths. So again, my off-the-cuff suggestion was that if we drew the line between analyticity (needed for the method of "analysis") and the a priori, we could use the former to dismiss the latter (along with the more overtly metaphysical notion of necessity).

But Hacker at least is clear that he does not want to do this. For Hacker, the a priori is the central concept he wants to defend: not as a possibly unacceptably metaphysical subject (i.e. object) of philosophical speculation, but as its constitutive method. It is this and this alone which distinguishes philosophy from empirical inquiry. I should have realized this, as the notion is (as in the SEP article) characteristically applied to the manner in which knowledge is acquired rather than its (semantic) form or (ontological) object, and the main contention of the "conceptual analysis" folks is that, again, philosophy is a matter of the clarification of our concepts as specifically opposed to empirical inquiry; so of course they want to defend the a priori as well as analyticity. (My excuse is that I didn't want to assume the naturalist characterization of the a priori (i.e. as hopelessly unempirical) from the beginning, even, or perhaps especially, because I too am not too keen on the notion, if for somewhat different reasons.)

For a interesting account of Hacker's attitude toward Quine, I recommend his paper "Passing By the Naturalistic Turn: On Quine's Cul-de-sac" (available on his website). The main target is Quine's "naturalized epistemology" (so some of what Hacker says is perfectly congenial), and in attacking it Hacker commits himself hook, line, and sinker wholeheartedly to a full-on Manichean dualism of pure a priori conceptual analysis, on the one hand, and not-at-all-philosophical empirical inquiry on the other. Picking up Quine's gauntlet, he begins:

There has been a naturalistic turn away from the a priori methods of traditional philosophy to a conception of philosophy as continuous with natural science.

and ends:

This imaginary science [naturalized epistemology] is no substitute for epistemology – it is a philosophical cul-de-sac. It could shed no light on the nature of knowledge, its possible extent, its categorially distinct kinds, its relation to belief and justification, and its forms of certainty. [...] For philosophy is neither continuous with existing science, nor continuous with an imaginary future science. Whatever the post-Quinean status of analyticity may be, the status of philosophy as an a priori conceptual discipline concerned with the elucidation of our conceptual scheme and the resolution of conceptual confusions is in no way affected by Quine's philosophy.

Snap! That last sentence answers our (my) question about priorities (no pun intended) pretty clearly, I'd say. Let's come back to this article; it's got a nice mix of right and wrong, uh, agreement and disagreement between Hacker and me (and Quine with both of us).

Not John Gielgud

I was preparing a post on the analytic/synthetic business we have been discussing (okay, so far it's other people, here, here, and here), and (curious as ever) I followed a trail of links to Wikipedia's article on "Two Dogmas," which I basically just glanced at (looks okay), but there's an interesting bit at the end which no-one has said anything about yet:

In his book Philosophical Analysis in the Twentieth Century, Volume 1 : The Dawn of Analysis Scott Soames (pp 360-361) has pointed out that Quine's circularity argument needs two of the logical positivists' central theses to be effective:

All necessary (and all a priori) truths are analytic.

Analyticity is needed to explain and legitimate necessity.

It is only when these two theses are accepted that Quine's argument holds. It is not a problem that the notion of necessity is presupposed by the notion of analyticity if necessity can be explained without analyticity. According to Soames, both theses were accepted by most philosophers when Quine published Two Dogmas. Today however, Soames holds both statements to be antiquated.

Upon reading this, I had two thoughts in quick succession, and wouldn't you know, they're in tension with each other. The first one was: I hardly think the defenders of analyticity (that is, those who, like our friend N. N., see Quine's attack as threatening the philosophical project of conceptual analysis, whether or not they see the latter as constitutive of philosophy itself), or anyone else unimpressed by Kripke for that matter, should welcome criticism of Quine's argument along these lines. I can't see any such philosopher saying: "see, you can too have analyticity – all you have to do is explain it in terms of an independently established notion of metaphysical necessity!" Surely the whole point of "conceptual analysis" was to put "metaphysics" out of business. So, no help there, right?

The second thought I had was this. Of course the contemporary naturalist/empiricist line of thought, in which "Two Dogmas" was an important early move, is also determined to put metaphysics out of business. But in so doing, it seems to assimilate philosophy into the empirical sciences, not as itself an empirical discipline, but as concerned solely with making sure that science dots the i's and crosses the t's in the proper way (once out of the lab and writing up the results). So if you put all of your anti-metaphysical eggs into the naturalist basket, by rejecting the distinction underlying the competing strategy of conceptual analysis, that means that when the naturalists (if not the empiricists) then turn around and reinstate metaphysics, you have no recourse.

Naturally they'll put a doily on that monstrosity by calling it a "scientific" metaphysics (whatever that means); but when it's accompanied, even justified, by a swipe at "linguistic philosophy" for neglecting metaphysics – well, that's going to be pretty galling. The reason my two thoughts are in (mild) tension with each other is that while the first implies that Soames's criticism of the argument of "Two Dogmas" is of no help to the linguistic analyst, the second thought leads to a different conclusion. For now that philosopher can resist the naturalistic line of thought right at the beginning: if the point of "Two Dogmas" was to deprive metaphysical pseudo-inquiry of the purely non-empirical conceptual space in which it was supposed to operate, well then the naturalistic revival of metaphysics shows that it failed to follow through on its promises. This means that (given the original choice between naturalism and conceptual analysis) as far as unmasking metaphysics as nonsense is concerned, the linguistic strategy is the only game in town after all.

These (quick) thoughts, you will notice, elided two complications, which I should at least mention. First, I exempted properly empiricist naturalism from the accusation of reversion to metaphysics. But it's not clear to me that they will be able to fend off such accusations when coming from fellow naturalists. (My own objections to these positions are of another order entirely, so when naturalists trade accusations of "reversion to/neglect of metaphysics," I don't need to take sides.) For the second elision, let's return to Wikipedia's article:

In "'Two Dogmas' revisited", Hilary Putnam argues that Quine is attacking two different notions. Analytic truth defined as a true statement derivable from a tautology by putting synonyms for synonyms [is] near Kant's account of analytic truth as a truth whose negation is a contradistinction. Analytic truth defined as a truth confirmed no matter what[,] however, is closer to one of the traditional accounts of a prioricity. While the first four sections of Quine's paper concern analyticity, the last two concern a priority. Putnam considers the argument in the two last sections as independent of the first four, and at the same time as Putnam criticizes Quine, he also emphasizes his historical importance as the first top rank philosopher to both reject the notion of apriority and sketch a methodology without it.

It does seem that the a priori, rather than analyticity, is the key notion here, and perhaps the defenders of Grice and Strawson would like to argue that the way to debunk the former (as I think we may construe their project) is to keep the latter rather than running the two together and discarding both.