Saturday, July 16, 2011

The Open Question Argument

I'm sympathetic to the old-fashioned idea that reflection on normative concepts suffices to see that metaethical reductionism is a non-starter. I recently sketched a "normative Knowledge Argument" along these lines, but of course the most famous argument in this vein is Moore's Open Question Argument.

The OQA has taken some heat after Kripke's (arguably overblown) cleaving of metaphysical and conceptual questions. For example, in Slaves of the Passions, Schroeder writes (p.72):
Most discussions of whether any reductive view about any normative property could possibly be true focus on one or another version of Moore's Open Question argument. But this is hard to take seriously, once we clarify that reduction is a metaphysical thesis, rather than one about our normative concepts... Open Question arguments employ tests of cognitive significance--for example, they ask whether someone can believe that Ronnie has a reason without believing anything about Ronnie's desires, or conversely.... [T]his is a good test for whether the concepts are identical. But as I view reduction, it is not a view about our thoughts.... All of the Open Question tests for cognitive significance distinguish between Hesperus and Phosphorus... But no one concludes that Hesperus is not Phosphorus.

I used to be sympathetic to such dismissals of the Open Question argument, but now think them overly hasty. Standard counterexamples involve rigid designators whose associated (reference-fixing) sense clearly opens the possibility of their co-referring with some other (non-synonymous) term. 'Hesperus' refers to that celestial body we see in the evening, 'Phosphorus' to that celestial body we see in the morning, and it's clearly an open possibility that the celestial body we pointed to in the morning is one and the same as that which we pointed to in the evening.

No such story seems available in the case of normative language. The concept of goodness, for example, does not have an associated sense that makes clear how it might end up referring to a merely natural property. It isn't a 'gappy' concept (as Parfit puts it): it does not mean that natural property, whatever it is, that fills such-and-such a role.

So here's my argument against dismissing the OQA:

First, for ease of exposition, let's introduce some terminology. Say that a term like 'Hesperus' refers opaquely insofar as the sense of the term does not by itself settle its intension (reference across possible worlds). In order to know how to apply the term to imagined possible worlds, we first need to learn more about how the actual world is. (I take it that not all terms are like this. Descriptive terms, like 'red', are apt to be transparent, such that grasping the concept is sufficient to know - a priori - how to apply it to counterfactual worlds.) We can now claim:

(1) Only opaque terms are ineligible/unfit to feature in Open Question Arguments.

(2) The Bridge Principle: Whenever a term is opaque, there is a transparently associated (reference-fixing) functional property that serves as a 'bridge' between the sense and reference of the term. (See above examples of the properties associated with 'Hesperus' and 'Phosphorus'.)

(3) Conceptual analysis reveals that there is no such functional property transparently associated with normative concepts.

Hence, (4) Normative terms do not refer opaquely, in the manner of 'Hesperus' and 'Phosphorus'. Grasping the sense of the term (i.e. the concept) enables one to likewise grasp its referent.

Hence (5) Normative terms are eligible candidates to feature in successful Open Question Arguments.



  1. Hi Richard,
    Just a quick question about your argument. What do you think about examples such as Burge's arthritis case or examples involving the substitution of synonyms within belief-ascriptions? The speaker who seems to know that he has arthritis in the knee might sincerely ask whether it could spread to his thigh, but it's impossible to have arthritis in the thigh. Or, the speaker who knows that a fortnight is a fortnight and a period of fourteen days is a period of fourteen days might sincerely ask whether it would be worse to go without internet for a fortnight or a period of fourteen days. Are these questions open? If so, I worry that the openness of a question is not a reliable guide to conceptual truth or analytic truth and certainly an unreliable guide to metaphysical possibility. (If not, I worry that we don't have a firm grip on what an open question is.) Either way, I'm worried.

  2. Hi Clayton, 'arthritis' seems like a paradigm opaque term, so receives a similar treatment to 'Hesperus', 'water', etc. (So it's not transparent what actual medical condition the term refers to, but there's some -- possibly complex -- associated functional role that is transparent.)

    The fortnight case is less clear, but it seems most natural to me to understand the question as not open. Anyone who doesn't realize that a fortnight is fourteen days just doesn't know what 'fortnight' means. (No?)

  3. Hi,

    interesting argument. Reminds me of Horgan and Timmons' revived Open Question Argument against the New Wave Naturalists. There's a good presentation of this in Miller's textbook in section 8.8.

    I guess I worry about why the premise 3 would be true. So, take the set of conceptual analyses we have currently - Smith's, Jackson's, Gibbard's, Railton's, and so on. Why and how does conceptual analysis tell us that these analyses - that could, at first, be understood to be aiming at identifying a functional property - are false? I know Parfit says so but what is the reason here?

    I guess the worry is something akin to the classic Frankena's begging the question objection to the OQA. It cannot be a premise in an argument that is supposed to show that moral concepts or properties are not reducible that they are not reducible, and it seems like 3 is assuming this instead of the argument establishing it as a conclusion.

    Also, I'm not sure why the fact that there is a functional property built into the associated sense of a term should be transparent to competent speakers in reflection. One view to take would be that senses are not propositional descriptions but rather know-how of being able to classify things in a certain way. On this view, that there is a functional property included in the sense could be a posteriori - a best explanation of how we use the concept. This would undermine the Bridge Principle. So, it seems like the same response as to the original open question is available against the third premise.

    All the best,
    Jussi Suikkanen

  4. Hi Richard,

    Thanks for your response. I think I have two questions (well, three, actually). The first has to do with Burgeability. Is your view that all Burgeable terms are opaque? (If so, are there really many transparent terms we competently use?)

    I think the issues having to do with knowledge of meaning are important, but I also worry that knowledge of meaning is a very wooly notion. In the case of fortnight/14 days, I take it that the following is the case--even those who don't know that a fortnight is a period of exactly fourteen days might have sufficient grasp of the concept to entertain thoughts with that concept, make knowledgeable judgments involving that concept, etc... Bracket that because I think that even with knowledge of meaning, we can still generate something like an open question that cannot have the significance OQs are supposed to have.

    (i) Gil knows the meaning of 'bachelor' and believes that bachelors are unmarried adult males, but it's an open question for him whether the Pope is a bachelor.

    If (i) is true, are we compelled to say that it's a (non-epistemic) possibility for the Pope to be a bachelor?

  5. Hello

    The conclusion (5) doesn't seem to follow from the premises, since the term "successful" in (5) comes out of nowhere. You could rephrase (1) to say that only opaque terms are unfit for successful Open Question Arguments, but that would seem to assume that there are non-opaque terms that are fit for successful Open Question Arguments, which is very close to the conclusion you want to reach.

    A safer but less powerful version of (1) would be that opaque terms are unfit for Open Question arguments. Of course, that won't get you to the conclusion that non-opaque terms are fit for such arguments. It seems that you need an independent argument for some sufficiently strong version of (1) in order for the argument to work.

    Regards, Larry Franz

  6. The conclusion might be that that if an OQA is to be successful, it must rely on non-opaque terms like normative ones, leaving aside the question whether an OQA can be successful. That would still be an interesting conclusion.

  7. But since it's hard to think of a way to show that there are successful Open Question Arguments, other than by pointing to successful ones, it seems that the best we could do is argue that a successful OQA will have certain characteristics, as you've done, and then offer examples.

    Ok, I think my soliloquy is done now.

  8. Hi Jussi - thanks for the reference, I'll have to check that out.

    Dialectically, I was imagining an opponent who grants that the OQA rules out analytical reductionism, but takes synthetic reductionist views to be immune. So, within this dialectic, I see the bridge principle as doing most of the work.

    I'm not sure I have much to say to those who deny premise 3, except that I sure don't mean to be talking about any mere functional property when I call something morally bad. Maybe they're talking about a different concept from me, I dunno...

    "One view to take would be that senses are not propositional descriptions but rather know-how of being able to classify things in a certain way."

    This is an interesting proposal. But can such dispositional accounts adequately capture the cognitive significance of our concepts? There seems to be an important categorical aspect to the meaning of our thoughts, to which we have direct ('a priori') access. And the bridge principle does seem very plausible to me -- it seems to capture the standard Hesperus/Phosporus cases very nicely, for example -- so I guess I'd want to hear more about the reasons for denying it.

    (This does seem like the right point in the argument to be applying pressure though, I agree.)

  9. Larry - right, feel free to insert "successful" into premise (1). I think it should be uncontroversial that referential opacity (in the specified sense) is what general objections to OQAs hinge upon, such that there would be no such general objection to an OQA that relied upon only non-opaque terms. (There would, of course, still be questions to raise about the details in any given case -- whether the allegedly open question really was open, etc.)

  10. Clayton - I'm pretty skeptical of the significance of Burgeability. Maybe I'm unusually self-centered or something, but when I'm doing philosophy I really don't much care about the linguistic dispositions of other people. (Speaker meaning rather than semantic meaning, may be one way to put it?) I think we can bracket Burgean issues by just stipulating that everyone else in my linguistic community shares my linguistic dispositions, as that shouldn't change the answers to any interesting or important philosophical questions, right?

    So to answer your first question: Yes, I think all Burgeable terms are partly opaque (insofar as they partly involve deference to others' linguistic dispositions), but we can bracket that and still be left with plenty of intrinsically-determined meaning. (I confess I've never understood why some people are skeptical of narrow content.)

    Your main challenge (with the Pope/bachelor example) is very interesting. When I consider the question whether the pope is a bachelor, I can see that it might have a kind of 'openness' to it, but it seems importantly different in kind from the openness of whether pleasure is good. In particular, I don't have the sense in the pope case that anything substantive is at issue; it is at most a sense that my concept of bachelorhood might not be fully determinate, and that the pope may be a borderline case. (Actually I think he's not, but I'm imagining someone with a slightly more lenient concept in this vicinity than my concept of bachelorhood.) In other words, the question has at most the 'open' feel of an unsettled terminological question.

    What's interesting about the OQA for normative concepts is that the 'openness' does not have this terminological feel to it. Indeed, it may not even seem particularly epistemically open, for I may know that pleasure is good. The real issue is that it seems substantive to claim that pleasure is good; I'm not merely affirming a concealed tautology.

    To avoid this ambiguity (in the different senses of 'openness'), perhaps a better name for the argument would really be the Substantive Question Argument.

  11. Hi Richard,

    "Your main challenge (with the Pope/bachelor example) is very interesting. When I consider the question whether the pope is a bachelor, I can see that it might have a kind of 'openness' to it, but it seems importantly different in kind from the openness of whether pleasure is good."

    I'm sympathetic to this, but my worry about this area of research is that so many important parts of the discussion are theoretically infected. (At any rate, the literature on the paradox of analysis is really what got me worried back in graduate school about the OQA. I think there's a fun Benson Mates paper on this you might enjoy, but it's been ages.)

  12. Hi Richard,

    thanks for the response. Just few quick comments:
    "except that I sure don't mean to be talking about any mere functional property when I call something morally bad"

    Well, when I call something a thermostat, I don't mean to be talking about a mere functional property but rather a mechanism that reacts to cold air by heating and warm air by cooling. In the same way, when I call something morally bad, I don't mean to be talking about a mere functional property but rather that thing being something not to be avoided, to be felt guilty about, to be warned against and criticed for, and so on.

    If the talk about a functional property in the former case is not transparent to me, it's not clear why it should be in the former case.

    "But can such dispositional accounts adequately capture the cognitive significance of our concepts? There seems to be an important categorical aspect to the meaning of our thoughts, to which we have direct ('a priori') access"

    I'm not quite sure what you mean by 'cognitive significance' and 'categorical aspect' in this argument. Could you help me a bit here before I say anything silly? If you mean whether this kind of views can deal with Frege's puzzle, then that should not be a problem as long as there can be different classificatory practices that pick out the same things.

    In any case, here's one case against the bridge principle. I take it that names are opaque - the sense does not determine reference across the worlds (Hesperus and Phosperus after all are names).

    Now, with names, if we believe Kripke, it cannot be required of individual speakers that there that the speaker associates a transparent reference-fixing functional property to the name. If you are in the chain of use, you can talk about Richard Feynman even if you don't know where you picked up the name. You might know that he is a famous phycisist but that doesn't pick him out uniquely either.

    The same probably goes for animal species. I can talk about Jaguars without having a functional properties in mind that would pick them out uniquely. So, the bridge principle cannot be true of names and natural kinds for all speakers, and so it cannot be general principle on which to rely.

    You might think that the original namers of a person need to have a functional property in mind. This wouldn't help with the moral properties case, and it's not clear why mere interaction with the name object would not be enough.

  13. Hi Jussi, I was thinking Frank Jackson's causal descriptivism undermined Kripke's reasons for doubting that names are transparently associated with functional properties. (They might be properties like "the person dubbed 'Feynman' at the end of the causal chain that led to this tokening".) Importantly, normative terms do not seem to be deferential or opaque in this way. We can fully grasp what we mean by calling a state of affairs intrinsically good, for example. If one were willing to grant the transparency of these terms directly, it could substitute for the second and third premises of my original argument.

    P.S. By 'cognitive significance' here I'm partly thinking of phenomenological significance -- there's something that it's like to think about intrinsic goodness, and I take it that the phenomenology here is influenced by (our grasp of) the meaning of our thoughts.

    I took your talk of 'know how' to indicate a kind of dispositional account (is this what you intended?), whereby the meaning of a concept is given by what we're disposed to apply it to. I spoke of 'categorical' aspects of meaning to contrast with this: there is an aspect of meaning that is directly present to our minds, we do not have to wait and see how we use our words before we know what they mean.

  14. I think this ultimately comes down to your view of metaphysics.

    If you believe in a robust metaphysics which allows necessary truths whose truth is not a consequence of the meaning of the words, e.g., you think that it's a necessary truth that Lewisian possible worlds exist, then your hole argument really doesn't accomplish much.

    After all there was no 'hole' in 'If Napolean had been a better general he would have conquered Europe' that asks for a universe just as real as ours where Napolean was a better general and did conquer Europe yet if you think Lewisian possible worlds exist necessarily the two claims are necessarily equivalent.

    On the other hand if you have a sparse view of metaphysics as I do and think necessary truth and analytic coincide (it's complicated to fully spell this out) then your argument is more convincing.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.