[quickphilosophy] Re: Fodor on Concepts III: Other Arguments Against BCP

  • From: wittrsl@xxxxxxxxxxxxx
  • To: quickphilosophy@xxxxxxxxxxxxxxx
  • Date: Wed, 01 Sep 2010 14:20:50 -0000

--- In quickphilosophy@xxxxxxxxxxxxxxx, Ron Allen <wavelets@...> wrote:
> Hi Walt:


> Well, OK, there is a lot hanging on that "most" modifying "concepts" in your 
> point (1).

Yeah.  (1) is an empirical claim and I wish Fodor had gotten into it in more 
detail.  It ought to be clear that language can work/be generative without ALL 
concepts being compositional, so the question is, how many must be?  I was just 
trying to help him out with the "most" there, though I have to admit that (1) 
doesn't seem implausible to me with the "most" in there.

> But, the argument you reconstruct from Fodor does not depend on a certain 
> percentage of concepts obeying the rule of composition. In fact, if I 
> understand the argument aright, there need be only one concept that is 
> compositional for it to work. 

I don't see that.  The argument depends on the empirical premise that human 
language is generative and understanding happens because of that.  So, the 
argument needs just so many concepts to be compositional as is required for 
that premise to be true.

>It's an argument by example. Because, the alleged BCP proponent asserts that 
>all concepts are given exclusively by a capacity to sort accordingly and a 
>capacity to infer accordingly. Fodor contrives things so that BCP asserts 
>universalities, whereas he, Fodor, only asserts an example.
> But, I still don't see where you have accurately captured Fodor's notion of 
> compositional. He represents it as a permutation of terms: Bob loves Alice; 
> Alice loves Bob. 

Here again we have the problem of quickphilosophy not moving quickly enough to 
suit my short-term memory deficiencies.  But, as I recall (and this could be 
totally wrong), the stuff about Bob and Mary and permutability are just more 
examples of what it takes for language terms to be both finite and generative.  
That is, he thinks that kind of permutation requires compositionality and 
compositionality (according to his argument) is inconsistent with BCP.

>But, you represent it as a decomposition. Is the permutational argument wrong, 
>and the later argument about the night-flying bluebird the right one? 

I don't think they're the same argument.  Permutability is provided as an 
argument for compositionality.  The night-flying bluebird example is to show 
that compositionality is inconsistent with BCP.

>It could be that I am too simplistic.

It's not that, I don't think, it's that you're skipping a step (which, if it's 
any consolation, is something I invariably do when someone, foolishly, asks me 
for directions).

> So, maybe there are two aspects to Fodor's compositionalism: one is 
> permutational and one is decompositional. Thus, composition is really, in 
> some cases that are important, decompositional. Hmm...your reformulation does 
> not address the permutational variant, which seems to me to have been the one 
> by which Fodor defined the concept of composition. So, let me have a second 
> look at the composition = decomposition option.

I don't address the permutation biz, because I don't think it's essential to 
the main argument.  It's just an additional example he gives in support of the 
need for compositionality.  I already agree with him that language must be 
compositional (at least in great part), so I don't care about the Bob loves 
Mary thing. It's just additional dicta I don't think the paper actually needs 
(not that I think it's wrong, however).

> There's nothing in the argument that depends on a quantification of how many 
> compositional arguments there are. If the argument works, then it works by an 
> example. And the example is an A+N example of a night-flying bluebird.
> Which is not an A+N example, but an A + B + C + N example, where A = night, B 
> = flying, C = blue, and N = bird. So, already the example is not a simple 
> composition, but a complex composition, and we'd have to wonder if the 
> concept would not be more appropriate if it were argued as a night-bird, or a 
> flying-bird, or a blue-bird. Ah, but those concepts would allow their 
> extensions be more easily culled, and this would undercut Fodor's argument.

Hmmmm. The bluebird is just an example, of course, and maybe he should have 
picked a simpler one.  I agree that it need not follow that because I can sort 
both "As" and "Bs" in favorable conditions, that I necessarily can sort "AB"s 
in favorable conditions (which his example is intended to show).  
But......couldn't one argue that whenever that is the case the concepts AREN'T 
really composed of nothing but Aness and Bness.  His own nightflying bluebird 
example suggests that.

> So, we're going to play along with Fodor, because he needs it, and say that A 
> = night-flying-blue and N = bird. Thus, we have an example of a concept that 
> decomposes into two lesser concepts (night-flying-blue and bird), and we hold 
> this up to BCPism to see what happens.

Right, but now I'm beginning to wonder whether the more complex concept really 
IS nothing more than what can be composed of those constituents.  I mean, 
couldn't somebody object that whenever one CAN sort As and Bs but CAN'T sort 
ABs we have good evidence that AB ISN'T just made up of Aness and Bness?  I 
suppose that in that case Fodor would just go back to his claim that many 
(most? all?) concepts simply MUST be compositional, or language understanding 
wouldn't be possible.  But, again, why couldn't there just be a handful of 
AB-type concepts that aren't compositional in that way without language being 
completely stymied?

> Just as an aside: This is not a strong argument. Fodor is arguing from an 
> example, and he is arguing from an example of some thing that does not exist. 
> He may as well say, "But, there might be an A-example and an N-example, such 
> that A+N is composed of A and N, and yet there is no way to sort A+N or there 
> is no way to infer from A+N." But, that is nonsense, because, in general, I 
> can sort A+N if and only if I can affirm A and I can affirm N, and I can 
> infer from A+N, precisely whenever I can infer A and I can infer N. This is 
> not a strong argument: It is no argument.

I agree that that is a weakness.  It gets back to the empirical nature of 
language "working" and just how much compositionality that requires. He doesn't 
give a sense of that, and I don't think he (or maybe anybody else), really 
knows the answer to that so no close approximation can be provided.  But, 
nevertheless, I agree with his premises: they all seem plausible to me.

> Fodor could have picked a real example, the Epidonax Flycatchers, for 
> example. It is really difficult to sort them into their five species. But 
> that is just to restrict the realm of exemplars to casual visual inspection. 
> These are not "good conditions" for Epidonax Flycatchers. Good conditions 
> vary with the concept. There is no reason to suggest that things might be 
> otherwise. You can tell an A+B+C+N from others by going out at night, killing 
> everything that is flying, seeing whether it is a bird, and then seeing 
> whether it is blue. Those are "good conditions" for the A+B+C+N concept. They 
> are not the same conditions as for a Bald Eagle, but who, except for perhaps 
> Fodor, suggested that the conditions for one concept had to be the same as 
> those for sorting the extension of another concept.
> What Fodor really needs for his argument to work is an example of a concept 
> that can be described, but not sorted. That's ridiculous. 

Right.  He wants us to forget that BCPers (at least the non-behaviorist ones) 
only want to make sorting a necessary condition.

>Fodor is demanding of the BCPers that they understand the specificity of his 
>example and then not understand the specificity of the example by making 
>mistakes in sortation.
> I'm all for an effort to discredit behaviorism and this caricature of 
> BareBones Concept Pragmatism as well, but Fodor's argument just isn't getting 
> me there. OK, it is entertaining; I'll grant that much! It seems to me that a 
> Chinese Room Argument (CRA) against BCP is effective for showing that the BCP 
> criteria are at most necessary conditions for concept possession, and not 
> quite sufficient in most, if not all, circumstances. I would be hard put to 
> come up with circumstances for which BCP provides sufficient conditions for 
> concept possession of any concept, given the CRA.

I agree with you.


Other related posts: