Fundamentalism and Knowledge

I spent a fun few days over the break at the NYU Causation Conference held at La Pietra just outside Florence. I’m not sure how widely disseminated anything said there was meant to be, but I think it isn’t too much of a violation of netiquette to talk about one of the big issues that came up.

In many areas of philosophy there is a debate between those who think a certain kind of feature of the world is fundamental, and those who think it is reducible to, or a feature of, other features. Consider, for instance, the debate between Descartes and Ryle. Descartes thinks that minds are fundamental features of reality, things that can (at least to a great extent) vary freely of physical facts. Ryle denies that minds are fundamental. Though there is some scholarly dispute over what his positive view is, we can all agree that he denies Cartesian fundamentalism about the mental.

Ryle makes an argument against Cartesian fundamentalism that is used by many other anti-fundamentalists. (Indeed, this style of argument came up at the conference.) The argument is as follows.

  1. If fundamentalism is true, then certain kinds of situations where the fundamental facts vary widely from other facts are possible. (E.g. the situation Ryle makes much of is that many people around us have no mental states at all.)
  2. If these situations are possible, we need to have a reason to rule them out before we know they are not actual.
  3. We have no such reasons.
  4. But we do know that such situations are not actual. (E.g. we know, for example, that Einstein was a genius, and hence had a mental life.)
  1. So fundamentalism is false.

Now there is a lot that can be said against premise 2. The best thing to say, as Brad Skow said at the conference, is that any such principle seems to lead to scepticism. We think that brain-in-vat situations are possible. And any sense of ‘reason’ strong enough to make premise 3 true will probably imply we have no reason to believe such situations are not actual. So we don’t, after all, know anything. (There are several gratuitous appeals to closure principles in both the argument and my reply, but I assume readers can fill in the details appropriately.)

At the conference there was a lot of discussion about whether this line of reply to the anti-fundamentalist argument worked in general, or on particular occasions. I wanted to make here one point that no one made there, and which for all I know hasn’t been made in the literature.

Assume for reductio that this epistemological argument really raises a problem for the fundamentalist. Then it seems there is just as much of a problem for the anti-fundamentalist. That is, assume Ryle is true that we don’t have a reason to rule out the kind of wacky situations the Cartesian says are possible, i.e. situations where physical behaviour and mental properties are totally out of whack. The point I want to press is that saying these situations are impossible doesn’t obviously help.

Most philosophers, not all but most, say that at least some of the time in order to know that p, we need to rule out certain impossible situations in which p is false. For instance, to know that Fermat’s Last Theorem is true, we need to rule out the situations, every one of them impossible, in which it is false. The obvious conclusion is that the impossibility of a situation is no reason to conclude we need not have a positive reason in order to know that it does not obtain.

There is more to be said here, and I might try and follow this up later, but I think that there is a big problem here for arguments like Ryle’s. If we have epistemological standards high enough that knowledge requires ruling out outlandish possibilities, such as the possibility of various properties that are always correlated in the actual world (and in most salient counterfactual worlds) coming apart, then we should also say that knowledge requires ruling out salient impossibilities. And then the Rylean argument that we can’t rule out those situations will imply an absence of knowledge, whether Cartesian fundamentalism is true or not.

Disclaimer: I don’t for a minute believe in Cartesian fundamentalism, or in most other kinds of fundamentalism that are philosophically popular. I just don’t think we can prove much metaphysically by going via epistemology, hence this little rant.

16 Replies to “Fundamentalism and Knowledge”

  1. If I understand you correctly, your point is this: One might fail to know p even though not-p is impossible, if one fails to know that not-p is impossible.

    In conjunction with the premises of the argument, this entails not only that fundamentalism is false but also that we know it to be false. Why can’t an advocate of the Ryle argument embrace this?

  2. Won’t the Rylean claim that he knows it’s impossible that P (i.e., the properties come apart)? If this is granted to him, he doesn’t need to provide further reasons for thinking that P is impossible (or irrelevant, or not a defeater of some sort).

    There is a sense in which, when we prove necessary truth S, we rule out certain impossibilities: none of the worlds in which S is false are possible. But this can just as well be understood as universal quantification over possibilities. We have shown that in all possible worlds (or at least in all worlds in which S has a truth-value), S is true. This, of course, entails that none of the worlds in which S is false are possible. But we don’t have to offer separate reasons showing that those worlds are impossible.

    Similarly for Ryle’s style of argument. Assume Ryle knows that it is impossible for certain mental and physical properties to come apart. There are certain impossible cases where such properties do come apart, but, by hypothesis, we know they’re impossible (although without anything like a proof). Thus, we need not offer any reason to think they’re impossible. Of course, everything turns on the claim that we know that it’s impossible for the properties in question to come apart.

  3. My problem is that I’m not sure what to make of this part of your post:

    “That is, assume Ryle is true that we don’t have a reason to rule out the kind of wacky situations the Cartesian says are possible, i.e. situations where physical behaviour and mental properties are totally out of whack. The point I want to press is that saying these situations are impossible doesn’t obviously help.”

    Does the last part (the saying) grant to Ryle that he knows the wacky situation to be impossible (although without reasons)? Or did you mean to be denying that knowledge to him by your description of Ryle’s position (that he doesn’t think we have reason to rule out the wacky scenarios)?

  4. Welcome back, Brian!

    I think the main issue for knowledge isn’t metaphysical possibility but epistemic possibility. The question is whether the ‘wacky situations’ still count as epistemically possible for the Rylean (in the way that either result for Fermat’s Last Theorem is epistemically possible). If these situations remain epistemically possible, no epistemic advance has been made.

    This same sort of issue arises in discussions of the relations between properties and powers. Those who hold that properties have their powers necessarily argue that if the relation were contingent we would get a skeptical result, due to wacky possibilities in which properties swap powers.

    Here’s a self-indulgent quote from my paper “Quiddistic Knowledge,” where I am arguing that the nomic necessitarian does no better in avoiding this seemingly skeptical result:

    “The second route to reinstating skepticism about properties is via the lingering epistemic possibilities. Consider the scenario that charge obeys an anti-Coulombic law, and that all of our current evidence otherwise is due to a massive practical joke. The nomic necessitarian will decree such a scenario to be metaphysically impossible. She will say that it is a merely epistemic possibility (cf the Kripkean reply to conceivability). But epistemic possibility is all the skeptic ever needed! If the scenario is epistemically possible, then it seems that one does not know that it is false, and so by closure one cannot know the falsity of anything it entails. Morale: downgrading a scenario from metaphysically to epistemically possible cannot help our epistemic position.”

  5. I wonder if Ryle has a response to the worry that Jonathan and Brian are mentioning. Maybe he should reformulate his argument like this:

    (1) Either Cartesian Fundamentalism (CF) or Ryle’s view (whatever that is) is true.
    (2) If CF is true, then crazy situations are metaphysically possible.
    (3) If crazy situations are metaphysically possible, then crazy situations are epistemically possible.
    (4) Crazy situations are not epistemically possible.
    (5) So Ryle’s view is true.

    The key move then would be to defend (4). I don’t know how Ryle would do this; the idea that since the crazy situations are, well, crazy, they are not epistemically possible doesn’t strike me as indefensible, but I don’t know how to push the point further. (“No, that’s absurd — we know that a crazy situation like that couldn’t happen.”)

    Presumably, Ryle should hold that anyone who knows that Ryle’s view is true knows that the crazy situation is not epistemically possible. (Ryle’s view entails that they are not metaphysically possible, and anyone who knows what the view entails should find that the crazy situations are epistimically impossible as well — hand-waiving at closure principles…)

  6. First, I should apologise to Jonathan for ripping off an argument of his without proper citation. I think the point I’m trying to make here just is the one Jonathan was making in the quiddities paper, and I should have acknowledged this. My bad.

    On Rob’s point, I think that Ryle doesn’t really have an argument that he knows P to be impossible. The argument is that if P is possible, bad epistemological things happen. But I think (a) they don’t happen, and (b) if they did happen they’d happen whether P is possible.

    On Kris’s argument, I think that premise 3 looks pretty hard to support. For comparison, we Humeans think that there is a possible world where the past is just like in the actual world, but in which it rains beer all next week. But that’s not an epistemic possibility; I know it isn’t going to rain beer next week. The metaphysical possibility of crazy Cartesian situations doesn’t make them any more epistemically possible.

  7. Kris: I agree with Brian that 3 looks problematic. Take any contingency that we know to be false. It will be metaphysically possible but epistemically impossible. Brian’s example of it raining beer next week is of this sort. Perhaps an even more dramatic example: “I do not exist.”

    What I think emerges from Rob and Brian’s discussion is that the issue of whether Ryle helps avoid any ‘skeptical’ problem turns on whether his theory cuts down the epistemic possibilities. Here is an argument that it doesn’t (same strategy as in the argument from my quote above on quiddities). Suppose behaviorism is true but all your evidence about dispositions has been skewed by a little practical joke. To be in pain really is to be disposed to laugh and sing—we’ve just been hiding that from you for now. If this practical joke scenario is epistemically possible, then the behaviorist is in just as much (or as little) difficulty as the dualist in knowing what mental state term applies to a third person.

    (Brian: You are being very generous [as usual], but no need to apologize for us coming to a similar idea.)

  8. Hi Jonathan and Brian,

    I agree that (3) shouldn’t be supported by the following sort of claim:

    (*) For all P, if P is metaphysically possible, then P is epistemically possible.

    • looks pretty clearly false, and your examples show that it is.

    My thought was that (3) was a way of reformulating Brian’s original (2) and (3), and would be supported by whatever was taken to support those premises. (Which haven’t been explicitly stated by me or Brian.) I had hoped that this reformulation at least dodged the original worry. But now I’m thinking that when it comes time to spell out the rationale for my premise (3), the same problems will arise as they did for Brian’s view.

    I’m also still worried about how someone like Ryle could support (4)- and Jonathan’s newest post worries me that they can’t.

  9. I just wanted to comment on one more point, from Brian’s disclaimer:

    “Disclaimer: I don’t for a minute believe in Cartesian fundamentalism, or in most other kinds of fundamentalism that are philosophically popular. I just don’t think we can prove much metaphysically by going via epistemology, hence this little rant.”

    I think it is ok in principle to use epistemic arguments in metaphysics. I mean if you can show something like:
    1. We know that p
    2. If metaphysical position x were true then we would not know that p.
    Then of course you can conclude:
    3. Metaphysical position x is false.
    This is just modus tollens. The presence of epistemic vocabulary (contextualist concerns to the side) should not invalidate modus tollens!

    But where I am in agreement with Brian is that I think it is very, very difficult to establish 2. What one typically sees are arguments that a metaphysical position renders a certain class of entities either (i) causally inert (eg possible worlds, epiphenomenalist qualia), or (ii) swappable in an indiscriminable way (eg souls, quiddities, substrata, etc.) The problem is that neither inertness nor swappability seem barriers to knowledge generally. Abstract entities seem causally inert but knowable. The external world seems swappable (for an indiscriminable demon scenario) but knowable as well. So while I think it is ok in principle to use epistemological arguments in metaphysics, I fear that many of the arguments one actually sees in practice presuppose unsustainable (/overly skeptical) epistemic principles.

  10. Do most philosophers really believe that knowing something just consists in ruling out possibilities? Because that seems to be what’s needed to think that the Fermat’s Last Theorem case requires ruling out impossibilities. Can’t there be some way to talk about this stuff in a more syntactically-based way?

    Anyway, it’s nice to see the distinction between epistemic and metaphysical possibilities coming up. It certainly causes a lot of confusion in probability sometimes, when people forget just what the elements of these sets are to which they assign probabilities.

  11. Brian,
    Welcome back!

    One (perhaps minor) quibble or factual error: Andrew Wiles (Math, Princeton) has proven Fermat’s Last Theorem (see here, and here). I think you could equally well substitute Goldbach’s Conjecture as an example of an unproven mathematical expression, which mathematicians have not ruled out all the situations in which it would turn out false.

  12. Alan,

    Brian’s argument (the one you’re referring to) doesn’t require an example of an unproven mathematical conjecture. On one, weaker, reading, his argument works regardless of whether any or all mathematical claims are known to be true. On another, stronger reading, it requires that there be mathematical questions known to be true, and if FLT is the example, the argument in fact presupposes that FLT has been proved.

    His argument is really just this:

    (1) Given any claim p, if p is known to be true, then we have ruled out all the situations in which p is false.

    (2) Given any mathematical claim p, if p is true, then all the situations in which p is false are impossible.

    (3) There are mathematical claims (e.g. Fermat’s Last Theorem.)

    (4) Therefore, there are claims such that if they are known to be true, we must have had to rule out impossible situations (those in which p is false).

    In this argument, it is not necessary that FLT be undecided.

    If we take Brian to be arguing for the stronger conclusion:

    (4’) Therefore, there are claims such that they are known to be true, and knowledge of which requires ruling out impossible situations.

    the argument requires the stronger premise

    (3’) there are mathematical claims known to be true

    and if FLT is the example, the argument requires that FLT is known to be true, not that it is undecided.

    I should note that the argument appears to be depend on the unstated premise (unstated in my formulation above): for some true mathematical claims p, there are situations in which p is false (otherwise the “all” in (2) is vacuous and does not yield the implied “some” in (4)/(4’)). And this seems to me one contentious point in the argument.

  13. I should note that the argument appears to be depend on the unstated premise (unstated in my formulation above): for some true mathematical claims p, there are situations in which p is false

    That’s right, I am taking something like to be a premise. The motivation (which may not be perfect) is that at least some of the time, mathematical knowledge is some kind of achievement. It seems to me the most natural way to represent that achievement is by saying we are ruling out impossible situations, but that too is controversial.

    Perhaps I would have been better off going for the following more complicated argument. The following is obviously very hand-wavy, but I think the idea behind it is sound.

    (1) Either knowledge should be understood on a broadly eliminative model (coming to know that p requires ruling out situations in which ~p) or it should not be.
    (2) If we adopt the eliminative model, then we have to say that some of the relevant situations are impossibilities, or else we cheapen mathematical knowledge.
    (3) If we adopt a non-eliminative model, then there is no reason to say that our actual empirical knowledge required us to rule out (in any sense where ruling something out is a kind of cognitive achievement) the possibilities countenanced by the fundamentalist.
    (4) So, either way, there is no epistemological argument against fundamentalism.

    Premise (2) is still controversial – Stalnaker for one controverts it – but it is weaker than what I had in the text. And of course premise (3) is controversial too. But I think this kind of argument makes it clearer what I was getting at.

    Of course the version Jonathan used – that what matters for knowledge are epistemic possibilties not (as the anti-foundationalist assumes) metaphysical possibilities, and we have nearly 40 years of work now on the idea that these are distinct – is perhaps clearer still and more to the point.

  14. Brian,

    I’ve been thinking about posting something about this for a while, but first your blog was hijacked and then I wondered whether I should interrupt the philosophical flow with a (at least at first sight) historical question. But the question continues to bug me, and the flow of discussion seems to have died down, so I’ll ask it now. Where does Ryle give an argument like the one you attribute to him, and in particular where does he commit himself to something like your controversial premise (2) (in the original argument)? The passage that immediately comes to my mind is The Concept of Mind, pp. 20-21, but you may have been thinking of another passage. Anyway I don’t see Ryle in that place as giving the argument you attribute to him, or not obviously as doing that anyway. And I find it interesting how in the discussion thread people just go about talking about the “Rylean argument” and “Ryle’s style of argument” when no one has actually showed how Ryle argued in that way.

    I do think there is something of philosophical, not merely historical, interest here. But I confess I am still trying to figure out exactly what it is. In any case I would like to think about this further with the relevant bit of Ryle in front of me — am I looking in the right place? If I can then come up with some way of making the historical and the philosophical come together in a clear way, I’ll let you know.

  15. Michael,

    I was thinking of the passage in pages 20-21. Or, at least, I was thinking of one of the arguments there. In particular, I was thinking about why Ryle thinks that we can’t “tell the difference between a man and a Robot” if Descartes’s theory is correct. I’m really not sure why Descartes couldn’t just reply that surface colour usually suffices to let us know.

    To be fair, I think Ryle has a few considerations going on simultaneously. And the argument I’m giving is perhaps one of the least charitable readings of the text. At least one of the considerations is that we don’t have direct access to these minds, apart from our own. But I suppose I still think that argument is vulnerable to the criticism I made above.

    As some support for the claim that Ryle is making some fairly strong epistemological assumptions here, note the line at the top of page 21. Ryle says that “the [Cartesian] explanation given presupposed that one person could in principle never recognise the difference between the rational and irrational utterances issuing from other bodies, since he could never get access to the postulated immaterial causes of some of their utterances.”

    Now perhaps this isn’t best read as the argument I gave, though it’s a little hard to see how to get a good argument out of it. Usually we can come to know about Fs by observing their causal consequences. (That’s how we learn about black holes, for instance.) But maybe Ryle’s point here is not the one I was giving, but an implausible claim like “If fundamentalism is true, then we don’t have the right kind of access to the fundamental entities to know about them,” where it is taken as given that observing their causal consequences is not the right kind of access.

    This is a bit rambly, but it’s late and I need to think more about this. By way of clearing things up, I should say that if it’s right to attribute to Ryle a more directly causal argument (rather than the modal argument I was discussing), that too is discussed (and I think soundly dismissed) in Jonathan Schaffer’s paper mentioned above.

  16. Brian,

    Thanks. I thought that was the passage you had in mind. I have been trying to think through
    what Ryle intends in that passage for some time. Your inclination seems to be to involve Ryle in
    some sort of generalizable argument involving general principles — in your response to Soames’s
    book you put this in terms of Ryle’s being guilty of “abductophobia” or an aversion to postulated
    causes; in your post here, you attribute to him a very general principle about having to rule out
    alternative possibilities in order to have knowledge; in your last response you suggest that Ryle
    “is making some fairly strong epistemological assumptions” on the basis of the bit you quote
    from p. 21.

    What I’d like to suggest, though, is that Ryle’s argument turns on aspects of the view that he’s
    attributed to Descartes and others (whether fairly or not) rather than on very general
    epistemological principles. As I see it, for Ryle, it is supposed to be part of the view under
    discussion — or implied by the view, or perhaps implied by the motivation for the view — that
    one can’t have access to the thoughts of others and that there is no real basis for any inference for
    their thoughts, and that this is quite different from the case of postulated unobserved physical
    causes/explanations of observed physical phenomena (such as sub -atomic particles). The
    difference is something like this: the postulation of sub-atomic particles is ultimately the
    postulation of further elements within the physical world, subject to the same laws of physics as
    the items they are postulated to explain. The postulated mental realm on the other hand, is
    supposed to be subject to its own laws, distinct from the laws of the physical realm. This is
    necessary because the motivation for the postulated mental realm is to preserve such things as
    human freedom from the “bogy of mechanism.” If the postulated mental causes could in the end
    be brought under the same general laws as the phenomena they were meant to account for, the
    fear that mechanistic explanations would extrude from our lives the mental (which Descartes
    wanted to avoid as a “religious and moral man,” p. 19) would reinstate itself, since everything
    would still be governed by the same problematic mechanical laws.

    Thus consider Ryle’s later argument in chapter III on The Will. Speaking about volitions as
    postulated inner causes of external actions, Ryle says that “the episodes supposed to constitute
    the careers of minds are assumed to have one sort of existence, while those constituting the
    careers of bodies have another sort; and no bridge-status is allowed. Transactions betweeen
    minds and bodies involve links where no links can be.” This is because “minds, as the legend
    describes them, live on a floor of existence defined as being outside the causal system to which
    bodies belong.” Because of this, Ryle claims, “the connection between volition and movement
    is admitted to be a mystery … not of the unsolved but soluble type, like the problem of the cause
    of cancer, but of quite another type.” (p. 66)

    Here and in the earlier argument, Ryle admits that Descartes begins with the idea that we know
    how to distinguish between the rational and the non-rational, between the voluntary and the
    involuntary, etc. Descartes is then supposed to have assumed that there must be a causal
    principle grounding these distinctions that we make – the rational and the voluntary must have a
    different sort of cause from the non-rational and the involuntary. Mental episodes are then
    postulated as this cause. And the desire to secure rationality and freedom from the threat of
    mechanism leads to the postulation of these causes as of a radically different sort from physical
    causes. The problem, as I see it, is that the supposed theory postulating these causes is no theory
    at all – according to Ryle the postulated mental acts which are supposed to cause our behavior,
    and so on, are merely postulated, without being any part of a real explanatory theory. So, in the
    paragraphs leading up to the passage you cite from The Concept of Mind Ryle first tells us that
    on the Cartesian view, we have to postulate “extra centres of causal processes” (p. 19) but that
    their “workings” “had to be described by the mere negatives of the specific descriptions given to
    bodies” and that the mind is postulated as a non-material engine governing the material engine
    of the body, which is “invisible, inaudible, and … has no size or weight … cannot be taken to bits
    and the laws it obeys are not those known to ordinary engineers” so that “nothing is known of
    how it governs the bodily engine.” (p. 20) The difficulty then is that we don’t (according to Ryle)
    have anything like a real theoretical postulation here. On the Cartesian view as he has described
    it, the only way we have of knowing about any mental states is by direct acquaintance, since
    there are no principles by which to infer from anything else to their presence, no real theoretical
    context in which they do their work, etc. Recall that “no bridge-status is allowed,” as Ryle says
    in the chapter on The Will; later in that chapter he says that “the thrusts postulated are screened
    from scientific observation,” so that no correlation of volitions and actions could ever be
    “scientifically established” (p. 68) – I think not because the mental thrusts were postulated as
    unobserved causes but because they were postulated as outside the physical realm altogether.

    Now all this may be unfair to Descartes, but the main point is that Ryle’s argument does not turn
    on generalized inductophobia or general principles about what’s required for knowledge, but on
    specific features of the position he’s criticizing (whether straw-man or not is another issue).
    And I do think there is a philosophical moral here, which is that arguments need to be examined
    carefully before they are claimed as instances of general patterns of argument or relying on
    general principles, and then criticized on that basis.

Leave a Reply