In her Carus lectures, Judith Thomson mentioned that ought seems to be ambiguous between a reading meaning, very roughly, probably and a more normative reading. So, for instance, if Jack is a regular party attender, and Jill has promised to attend this party, (1) is true on the first reading, and (2) on the second. But we have an ambiguity here, as evidenced by the fact that (3) is rather odd.

(1) Jack ought to be dropping by soon.

(2) Jill ought to be dropping by soon.

(3) ??Jack and Jill each ought to be dropping by soon.

I thought this was all terribly interesting, and pretty convincing, but I’m told (by Chris Kane) that it has been discussed a bit in the epistemic deontology literature. (Since I heard the Carus lectures, not read them, I haven’t seen the footnotes where this is probably mentioned. And it might have been mentioned in the lecture too, but I’m not the most attentive of listeners, especially when trying to listen and deal with a mild case of black death or whatever I was suffering from in Cleveland.) So rather than write up all of my thoughts, which included ripping off some of Thomson’s jokes about people falling from the Empire State Building, I just have two quick questions.

First, does anyone have good references on this ambiguity?

Second, does anyone know how widespread this ambiguity is in other languages? Since it appears twice over in English (for should as well as ought) I wouldn’t be surprised if it is widespread. And that would be a neat result I think, since it would undermine a view that is rather widespread in philosophy. The view in question, which is made most explicit in Kripke’s response to Donnellan, is that ambiguity is always a matter of coincidence, so we should assume that English ambiguities will not appear in other languages, especially those not closely related to English. (Actually, Kripke isn’t very clear on the last qualification, so I’m possibly been generous in stating his view here.)

6 Replies to “Ought”

  1. I’ve only just read this interesting post. I must get Judy to send me her Carus lectures. (I am planning to work on the meaning of ‘ought’ over the summer.)

    I think that the view of most formal semanticists like Angelika Kratzer is right: ‘ought’ is basically a modal term. Most modal terms have paradigmatically normative uses as well as epistemic uses (‘may’, ‘can’, ‘must’, etc.), and ‘ought’ and ‘should’ are no exception. (This is certainly not just a quirk of English. A great many languages’ modal terms have both paradigmatically normative and epistemic uses.)

    Surely these terms aren’t ambiguous, but systematically context-sensitive. With ‘ought’, e.g., there must be some contextually salient ranking of possibilities (e.g., this may be a ranking with respect to how probable those possibilities are, or with respect to how good they are in some way or other). It “ought” to be the case that p if and only if it is the case that p on all the sufficiently highly-ranked possibilities.

    (I’ve been saying “paradigmatically normative” because I am tempted by the view that the epistemic reading is also normative in a way.)

    BTW, there are some nice examples that suggest to me that there is no ambiguity between the so-called moral ‘ought’ and the ‘ought’ of rational belief. E.g.: ‘You ought to keep your promises and proportion your belief to the evidence’ sounds perfectly OK to me.

  2. Ralph, I don’t think this affects your main point, but I do think it’s misleading to describe words like “ought” as context-sensitive. It’s not the context but the speaker that determines (in the context) the more specific way in which the term is used. For example, in what might clearly be an epistemic context (people are wondering when George will arrive), a speaker might use “ought” normatively (“George ought to be here by noon”). The speaker will be misunderstood all right, but this shows only that what he is taken to mean can be different from what he does mean. I think the phrase “semantically underspecified” (or “underdetermined”) better describes such terms as “ought”.

    In an interesting 1989 book, On Monosemy, Charles Ruhl discusses a variety of common terms that fall into this category, including verbs like “go”, “get”, and “take” and pronouns like “from”, “in”, and “on”, and suggests that dictionaries “overspecify” meanings of such words. Instead, he proposes, each has but one “highly general meaning” and must be used in some more specific way that fleshes out this skeletal meaning.

    No doubt I am oversensitive to phrases like “context-sensitive” and “contextually determined.” For an explanation of this pet peeve, see a piece called “Context ex Machina,” downloadable from my website (click on my name below).

  3. There’s another underdetermination that takes place with “ought.” I’ve seen it described as the difference between ought to be and ought to do, but that’s not quite right—it’s not determined by the verb following “ought.”

    Consider the following two situations.

    (1) We are at a family reunion hosted by my notoriously feckless nephew. There is anxiety about whether he has laid in enough supplies. I say, “He ought to have enough food for his family.”

    (2) We are watching a documentary about poverty in America. A Wal-Mart cashier is describing how, given his and his wife’s low salaries, they sometimes do not know where their children’s next meal is coming from. I say, “He ought to have enough food for his family.”

    Both of these are cases of the moral “ought,” but the onus falls differently. In (1), the onus falls on the subject of “ought”; in (2), the onus is impersonal. My nephew has done wrong if he does not have enough food to feed his family; it is wrong tout court if the Wal-Mart cashier does not have enough food to feed his family.

    I think you can’t conjoin these oughts:
    (3)??My nephew and the cashier both ought to have enough food to feed their famililes.
    But of course (3) is an odd thing to say anyway; I’d have to come up with something more natural to prove this case.

    Interestingly, in the epistemic “ought” the onus doesn’t fall on the subject of the “ought.” In the epistemic sense, “P ought to phi” seems to mean, given what the speaker knows, it is likely that P phis; or maybe that it is likely tout court that P phis.

    (I’m going to cross-post this on my own blog here. The two senses of “supposed to” came up there, because someone posted “Aren’t you supposed to be dead?”)

  4. Matt, you’re quite right about this sort of “underspecificity” (to use the term that Kent approves of — I must read Kent’s paper to understand whether I should give up my enthusiasm for appealing to “context” here). The David Lewis / Angelika Kratzer picture can handle this point quite easily, though.

    In the cases where you say that the “onus” falls on the subject of ‘ought’, the relevant ranking of possibilities is what moral philosophers would call “agent-relative”: it’s a ranking of the possibilities with respect to how well the agent acts on those possibilities. In the cases where you say the “onus” does not fall on the subject of ‘ought’, the relevant ranking is “agent-neutral”: it’s a ranking of the possibilities with respect to how good those possibilities are in terms of more impersonal values like fairness, etc. With the epistemic ‘ought’, the relevant ranking may be speaker-relative, but it is not agent-relative.

    So I also completely agree with you that it’s not a distinction between ‘ought to be’ and ‘ought to do’. There is no reason to insist (as some philosophers like Gil Harman and Peter Geach used to do) that ‘ought’, in the so-called ‘ought to do’ cases, cannot be understood as a propositional operator.

  5. Does the Lewis/Kratzer picture run into trouble with the gentle murder paradox? From your description of it as a ranking of the possibilities, it seems to me that it shouldn’t—“If you murder him, you ought to murder him gently” comes out fine if you interpret it as “Of the possibilities in which you murder him, the best one is the one in which you murder him gently.”

    In any case it looks absolutely right that the epistemic “ought” is agent-neutral.

    I’ve thought of a couple of examples that may or may not shed light on ambiguities in “ought”:

    (1) Jordan, a 13-year-old, has been told not to play in the street. One day Jordan and a one-year-old baby are playing at opposite ends of the street. Someone says: ?Neither Jordan nor the baby ought to be in the street.

    This strikes me as a bit funny—Jordan’s “ought” is agent-centered, the baby’s “ought” is agent-neutral. But I wouldn’t trust my linguistic intuition on this one.

    (2) Can you say: You ought to be kind to others and put nutmeg on a Georgia Flip? It seems to me that there are two different “oughts” here—the second is purely instrumental about how to make a Georgia Flip. The sentence sounds a bit odd to me, but perhaps that’s because it’s odd to conjoin those two things anyway.

  6. The Lewis/ Kratzer picture handles “paradoxes” of this sort in the way that is fairly standard in deontic logic.

    In every context when we use ‘ought’, there are certain facts that are “held fixed”. (With the agent-relative ‘ought’, these are often just: all the facts that cannot be changed by the agent at the time in question.) The only relevant possibilities are those that are compatible with these facts.

    E.g. suppose I say to a drug addict: ‘You ought to stop injecting heroin’; then later I say, ‘You ought to inject yourself with clean needles’. Both could be true, if in the first context it is not being held fixed that the addict will continue injecting himself with heroin, while in the second context it is.

    I’ll try to think about your interesting examples (1) and (2). I’m inclined to agree with your reactions. ‘You ought to put nutmeg on a Georgia Flip’ is a bit like an example that I heard Bernard Williams use: ‘He ought to use a Phillips screwdriver to open that safe’. In these cases, the ranking of possibilities is a “purpose-relative” ranking — i.e. it ranks the relevant possibilities purely with respect to how well or efficiently the agent achieves a certain purpose (viz. making a Georgia Flip, or opening the safe).

Comments are closed.