SORITES, ISSN 1135-1349
Issue #14 -- October 2002. Pp. 110-116
Copyright © by SORITES and Wayne Wright
Fodor's Epistemic Intuitions of Analyticity
(1) can be derived from either a conceptual role theory of meaning or verificationism, and (2) is supposed to follow from the failure of the analytic/synthetic distinction.
Holism is often thought to have several highly counterintuitive consequences, particularly regarding the incommensurability both of different thinkers' concepts and of the concepts of an individual thinker at different times.Foot note 8_2 Granting that there are no formal problems with the holism argument, the anti-holist's options seem limited to denying either (1) or (2).Foot note 8_3 Jerry Fodor (1987, 1990, 1998) has responded to the holism argument by denying (1), which leads to his atomistic theory of concepts. According to Fodor, the meaning of a concept C is independent of the meaning of any other concept. It follows from this that concepts, contrary to much of philosophical tradition, are not complexes consisting of meaning-constitutive connections with other concepts. This claim about the semantics of concepts has an important consequence for the requirements on concept possession: according to atomism, possessing an individual concept C does not require that one possess any other concept.
In this paper, I will not directly address Fodor's atomism. Instead, I want to discuss what Fodor (1998) sees as the principal inducement for responding to the holism argument by denying (1) and thereby opting for atomism, the repeated failure of efforts to refute (2) by attempting to make the analytic/synthetic distinction principled. If (1) is maintained and (2) is denied, molecularism would result; according to the molecularist, the semantic properties of a concept C are dependent on the semantic properties of some (but not all) of a thinker's other concepts. If an analytic/synthetic distinction could be maintained, the molecularist could distinguish conceptual connections which are meaning-constitutive from those which are not by saying that a concept C's meaning depends on those concepts to which it is analytically related, and any concept to which C is not analytically related does not figure in determining C's meaning. Since many still cling to strong intuitions of analyticity, there is at least some reason for supposing that there are meaning-constitutive interconceptual connections, despite the implausibility of holism.
Fodor sympathizes with such intuitions of analyticity to a certain extent, but denies that they have anything to do with a semantic property of concepts. Instead, he claims that such intuitions are really informing us as to an epistemic property of concepts, which is not sufficient for doing the work required to reject (2) of the holism argument. Since holism is unacceptable, the only remaining option is to deny (1) and embrace atomism. In what follows, I will argue that Fodor fails to make a convincing case against analyticity, and thus he cannot rule out molecularism, which is the primary anti-holistic competitor to his own theory.
When inquiring into whether there are any analyticities, one is often struck by the following sort of consideration: BACHELOR and UNMARRIED MALE are connected in a way which BACHELOR and LIVES ALONE are not.Foot note 8_4 Put another way, bachelors are unmarried males looks to be a different kind of truth than bachelors often live alone. Such cases often lead philosophers to harbor intuitions on behalf of analyticity, despite the strong influence Quine's attack on analyticity continues to exert. The pro-analyticity sentiment is bolstered by the suspicion, often felt even in quarters sympathetic to Quine, that it is unclear both what Quine's arguments really were and how successful his arguments should be taken to be. Pairing strong intuitions on behalf of analyticity with the apparent lack of a decisive argument against analyticity might lead some to suppose that we are justified in appealing to analyticity to explain the difference between `bachelors are unmarried males' and `bachelors often live alone'; the former, and not the latter, is an analytic truth.
Fodor agrees that the intuitions on behalf of analyticity are strong. However, he claims that the alleged confusion over Quine's arguments is mistaken.
I am, myself, inclined to think it's pretty clear after all how Quine's main argument against analyticity is supposed to run: namely, that nobody has been able to draw a serious and unquestion begging distinction between conceptual connections that are reliable because they are intrinsic/constitutive and conceptual connections that are reliable although they aren't; and that it would explain the collapse of this project if there were, in fact, no such distinction. (Fodor (1998), p.71)Foot note 8_5
It should be noted that it is unclear to what extent Fodor actually endorses the argument he attributes to Quine. Fodor says that his informational semantics will show why Quine was right, but he does not clearly indicate that he finds Quine's argument at all convincing. There are places, however, where Fodor seems to not only agree with Quine's negative conclusion regarding analyticity, but also to think that Quine's argument is successful. In Section IV, I will argue that even if it turned out that Fodor wanted to endorse Quine's argument, he cannot do so and remain consistent with what he says about a prioricity and necessity.
A deeper concern about this passage is that the argument attributed to Quine does not constitute a genuine refutation of the analytic/synthetic distinction. Simply put, the fact that we cannot (or, at least, have not yet) draw(n) a principled analytic/synthetic distinction should not by itself be taken to entail that there is, in fact, no analytic/synthetic distinction. That we are unable to draw a principled analytic/synthetic distinction may very well be a necessary condition of there being no real analytic/synthetic distinction, but it is far from obvious that it is a sufficient condition. Further, why suppose that distinguishing between conceptual connections which are meaning-constitutive and those which are not requires that we have on hand a principled analytic/synthetic distinction? Considering the strength of the intuitions of analyticity, there seems to be no prima facie reason why we could not take paradigm cases of intrinsic conceptual connection, such as seems to hold between BACHELOR and UNMARRIED MAN, as meaning-constitutive and paradigm cases of conceptual independence, such as seems to hold between BACHELOR and LIVES ALONE, as not meaning-constitutive. At most, Fodor's argument manages to shift the burden of proof back on to the supporters of analyticity. However, granting for the moment that the burden-shifting is successful, Fodor still needs some further argument if he is to have a knockdown refutation of the analytic/synthetic distinction, which is what he seems to think he has and what his endorsement of informational semantics requires.
Fodor claims that his atomistic informational semantics is a fallback position from the perceived failure of inferential role semantics (p.146). The repeated failure of attempts to make the analytic/synthetic distinction principled is supposed to lead us to despair over there ever being an acceptable molecularist version of inferential role semantics; according to Fodor, any inferential role semantics will lead directly to holism.Foot note 8_6 As already noted, however, Fodor's problem is that, despite the problems surrounding the analytic/synthetic distinction, there still remain strong intuitions of analyticity. If informational semantics truly is just a fallback position, it has to be shown that there is something which needs falling back from; mere hand-waving about there being difficulties for one's competition will not suffice. Thus Fodor needs to account for there being intuitions of analyticity in a way which shows them to offer no evidence for supposing that there really are meaning-constitutive interconceptual connections.
The standard, Quinean way of deflating analyticity intuitions is to say that cases which appear to exhibit analyticity are instead exhibiting «theoretical centrality.» This might work for `F = MA' and the like, which could appear to be analytic simply because a great deal of the rest of physical theory hangs on it. However, as Fodor points out, `bachelors are unmarried males' and `Tuesdays come before Wednesdays' do not exhibit any such theoretical centrality. As a matter of fact, such truths seem to rest at the outermost edges of the web; they are so highly conventionalized that nothing of theoretical import rides on them at all. Moreover, that they are so isolated from nearly everything else seems to make `bachelors are unmarried males' and `Tuesdays come before Wednesdays' better candidates for analyticity than `F = MA'. They look to be mere conceptual truths. So, theoretical centrality, by itself at least, cannot be all that there is to a deflationary account of analyticity.
This sort of consideration forces Fodor to supplement the appeal to theoretical centrality, and he chooses to do so with a version of Putnam's notion of a «one-criterion» concept. A one-criterion concept is one for which «there is just one way to tell that [it applies]» (p.80). For Putnam, one-criterion concepts, and not those which exhibit theoretical centrality, are the sorts of concepts which figure in analytic inferences. Putnam's reason for cutting things up this way looks to be grounded in the already noted difference between truths such as `bachelors are unmarried males' and `F = MA'. Take BACHELOR as an example of a one-criterion concept: there's just one way to tell whether Smith is a bachelor, which is by checking whether he is an unmarried male, so `bachelors are unmarried males' is analytic. Thus on Putnam's story, the good news is that it turns out that there really are analyticities, but the bad news is that analyticity is confined to a group of concepts which are of little, if any, philosophical interest. It will turn out that there are conceptual truths involving BACHELOR and TUESDAY, but not involving JUSTICE, GOD, or TRIANGLE.
Unfortunately for Putnam, put to use as an account of analyticity itself, one-criterionhood fails. Fodor points out that the `just one way to tell' test for analyticity works only if you already have in place some means of individuating criteria.Foot note 8_7 Whether BACHELOR has only one criterion (viz., `unmarried man') «depends, inter alia, on whether `unmarried man' and `not married man' are synonyms» (p.82). But, following Quine, `synonymy' and `analyticity' are both members of the same troublesome circle of interdefinable terms, so any appeal to synonymy in the explication of analyticity is bound to be circular. Thus Putnam's one-criterionhood does nothing to remove the standard Quinean worries about analyticity.
Since Fodor's project is not to give an account of the metaphysics of analyticity, but instead is to explain our intuitions of analyticity consistent with there being no real analyticities, he feels that he can live with the sort of circularity involved in the notion of a one-criterion concept which dooms Putnam's use of it (p.82). For Fodor, the reason that `bachelors are unmarried males' and `Tuesdays come before Wednesdays' seem analytic is that concepts such as BACHELOR and TUESDAY are so highly conventionalized that there are few ways to introduce them to someone who does not already have them, and we come to think of those few ways as being particularly intrinsically connected to that concept. Thus analyticity is not a semantic feature of concepts, as the molecularist insists it is, but is instead epistemic. Since the only apparent way of determining whether someone is a bachelor is to check whether he is an unmarried male, `bachelors are unmarried males' seems to be analytic, whereas since there are other ways of finding out whether something is a dog besides checking whether it is an animal, `dogs are animals' seems less, or perhaps not at all, analytic.
One concern which Fodor's Auntie has about this strategy is that if there's one way to tell that something is an X, then there are bound to be lots of ways to tell that something is an X. Thus it turns out that there are no one-criterion concepts after all, and Fodor has done nothing to account for intuitions of analyticity. If I can tell that Smith is a bachelor by checking to see if he is an unmarried male, I can tell that Smith is a bachelor by having Jones check whether he is an unmarried male. And, if I can tell that Smith is a bachelor by having Jones check whether he is an unmarried male, I can tell that Smith is a bachelor by having Jones ask Williams to check whether he is an unmarried male. This can go on indefinitely. Fodor responds to this worry by saying that «some ways of telling pretty clearly depend on others. It's ...the pretty clearly independent ones ...that you are supposed to count when you decide whether something's a one-criterion concept» (pp.82-3). According to Fodor, whatever way I devise of determining whether Smith is a bachelor will depend (in some way) on determining whether he is an unmarried man, and whatever way I devise of determining whether today is Tuesday will depend on determining whether today is the second day of the week.
Although I agree to some extent with Fodor regarding the particular example Auntie brings up, his response to the general point raised by Auntie's objection fails (Auntie never was one for details, I suppose). Contrary to what Fodor says, I can think of ways in which I might determine whether today is a Tuesday which do not depend on my performing inferences involving SECOND DAY OF THE WEEK. If I know that today is 9 December, and I know that 2 December was a Tuesday, and I know that a week contains seven days, then I can determine that today is Tuesday without ever knowing what place Tuesday (or any other day, for that matter) has in the order of the days of the week; SECOND DAY OF THE WEEK does not at all figure in my inference-process. If I'm working as a telemarketer, and I'm given a list of people to call, and I'm told all of the people I am to call are bachelors, and I find Smith's name on that list, then I can determine that Smith is a bachelor without ever deploying UNMARRIED MAN in my inference-process. Just as there is a «plethora» of reliable ways of determining that something is water besides checking whether it's H2O, the only apparent limit on the number of reliable ways one might determine whether today is a Tuesday or Smith is a bachelor is one's imagination.
Knowing that there are lots of ways of determining that something is a bachelor or a Tuesday, however, doesn't sway one's intuitions of the analyticity of `bachelors are unmarried males' and `Tuesdays comes before Wednesdays' (or, at least, it doesn't sway my intuitions). For that matter, that there are plenty of ways to determine whether something is a dog which do not involve inferences in which ANIMAL is deployed doesn't make `dogs are animals' seems any less analytic, despite Fodor's claims otherwise. Fodor's appeal to one-criterionhood to account for intuitions of analyticity seems quite unsuccessful. And, as already noted, if Fodor cannot come up with a way to explain away intuitions of analyticity by showing them to be of something other than analyticity, then there is simply no reason to follow him in rejecting analyticity and thereby opting for his atomism.
Getting back to a point raised in Section II about what to make of Fodor's discussion of Quine's argument against analyticity, it seems as though any appeal to Quine by Fodor would be quite strange, because it would undercut other important parts of his own theory. Fodor, like Quine, is clearly opposed to analyticity, but, unlike Quine, Fodor is congenial towards a prioricity and necessity. The oddness of Fodor's even mentioning Quine in his own discussion of analyticity is brought out in the following passage:
I think it's reasonable, on the philosophical evidence, to suppose that such conceptual connections don't exist. Quine was likely right about conceptual connections, even though he was wrong about necessity and a prioricity, both of which are, so I suppose, very important and perfectly real. (p.87)
The problem with this sort of thinking is that `necessary' and `a priori' put you directly in the troublesome circle of interdefinable terms, which also includes `analyticity', that Quine criticized in «Two Dogmas».Foot note 8_8 In fact, Quine showed how, if you had a prioricity and necessity, you could get analyticity. Thus if necessity and a prioricity are real, and Fodor's argument is Quine's argument, it is fair to ask: What's so unreal about analyticity? So, it is clear that Fodor is not relying on Quine to make his case against analyticity. One has to wonder what Fodor could be relying on.
I certainly do not mean to say that the general position Fodor takes is an impossible one to maintain. One could follow Kripke (1980), who goes to some length to show that necessity, analyticity, and a prioricity are very different notions and that we would be wise to keep them separate. So, there's no prima facie reason to doubt that necessity and a prioricity are okay and analyticity is not. But, to hold that, one of the things you absolutely need to have is a convincing argument against analyticity. Well, if Fodor's anti-analyticity argument isn't Quine's anti-analyticity argument, then what is his anti-analyticity argument?
Fodor rightly points out that if his informational semantics turns out to be true, then analyticity is ipso facto ruled out. But, since atomism is only a fallback position from the perceived failure of the analytic/synthetic distinction, a satisfactory deflationary account of our strong intuitions of analyticity is needed before Fodor's position should be accepted. Otherwise, Fodor begs the question; because the viability of informational semantics rides on the success of the anti-analyticity argument, it seems to me clearly illegitimate to use informational semantics as a premise in an argument against analyticity. As was noted at the end of Section III, the appeal to one-criterionhood will not provide the needed deflationary account of analyticity. Thus all Fodor really seems to have established is that informational semantics and analyticity are incompatible. However, that does not in any way constitute a substantial conclusion; we knew that from the start. Thus unless Fodor has tucked up his sleeve an anti-analyticity argument of his own (one which does not have informational semantics as a premise), he has done nothing to show that molecularism, the anti-holistic position competing with his atomism, is false. Given Fodor's failure to decisively rule out the primary competitor to his theory and the fact that his informational semantics is burdened with problems of its own, it appears very much as though molecularism is still an open option and that there is no reason at all to buy into Fodor's atomism.