Lewis on Warrant and Truth

Michael Hand just called me to ask a question, and thought I’d post about it. He seems to recall seeing a passing remark in some David Lewis paper against antirealist attempts to substitute some notion of warrant in place of truth in the theory of meaning, but can’t remember where. The remark was a very cursory argument against such attempts, to the effect that any clarification of the meaning of warrant will have to appeal to the concept of truth anyway (perhaps through some connection to likelihood of truth), so the antirealist attempts to replace truth with warrant are bound to fail.

I don’t recall seeing any such argument in Lewis, but others know his work better. Michael also wondered about the argument, whether epistemologists talk much about it, and whether it is thought to be a good argument. I told him what I thought, but I’d rather see what others think before saying here. Any thoughts on either question?


Lewis on Warrant and Truth — 10 Comments

  1. For what (little) it’s worth, there’s no occurrence of the words “antirealism” and “antirealist” anywhere in Counterfactuals, On the Plurality of Worlds, Parts of Classes, or any of the papers in Philosophical Papers Volumes I and II or in Papers in Metaphysics and Epistemology or in Papers in Philosophical Logic. “Warrant” appears in “Scorekeeping in a Language Game” and in “Humean Supervenience Debugged”, but in neither case in a relevant way.

  2. The closest relevant word to “warrant” that I could find while cruising through JStore was “assertability”.

    Anyway: can’t much comment on an argument that hasn’t been provided.

  3. A quick thought, if I may.

    Looks like there are two notions of assertability. A strong one, according to which we can only correctly assert what is true, and a weaker one, according to which, for every p, p is assertable iff p is rationally believable, or something like that. If ‘warrant’ means whatever makes an assertion correct in the strong sense, then it is difficult to see how such a notion could be defined without appealing to the notion of truth.

  4. On the one hand, to take those comments a step further, I have trouble understanding how even the utterances governed by a standard of “rational believability” may be separate from the logical sense of truth; since the logical sense of truth is quite a bit more wide than the narrower everyday sense, and seems to extend across most intelligible utterances (including expressives).

    On the other hand, I presume that one of the attractions of an account of semantics that reaches beyond ‘truth’ is that we must be able, with a straight face, to provide a straightforward way of treating seemingly problematic, non-indicative statements (especially commands). And commands are only arguably potential bearers of truth. For instance, it seems strange to say that the utterance, “Do your laundry right now!” bears any truth-conditions. If we were to hold the view that it was, then we’d need to take a circuitous route by postulating a number of implicit statements.

    Another appeal of alternative accounts is the intuition that semantics is not just the study of sentences or languages, but also the poor forgotten study of words themselves as bearers of meaning. Truth-values can only be assigned to propositions (or so I’m told); and a single lexical item is not a proposition. Presumably an alternative account would be broad enough to treat isolated lexical items. (For Tarski, these were satisfaction conditions; for others, propositional functions; etc.)

  5. The argument suggested strongly reminded me of a paper by Paolo Casalegno (2002), “The Problem of Non-Conclusiveness”, (in Topoi 21). It argues that no condition of assertibility is adequate for expressing the meaning of a senstence unless it implies the truth of the sentence itself.
    The structure of the argument is similar to that of Huemer’s (2001) “The problem of defeasible justification”(Erkenntnis 54; this is also available in Huemer’s webpage, I think), which argues that, independently of its relation to meaning, the concept of defeasible warrant (warrant not entailing truth) is incoherent. Although I am generally a fallibilist, I think I have conclusive grounds to say these are both excellent papers.
    On the more general question, I’d say epistemologists tend to think that warrant is conceptually related to truth, via either subjective or objective probability. The only sort of theory I can think of that could escape this tendency is a certain kind of coherentist theories of justification. What epistemologists certainly do not tend to do is worrying about the relation within these views and the different theories of meaning.

  6. I recall seeing an argument of this sort in a paper my students read for John MacFarlane’s class on philosophical logic this semester. Looking back at the syllabus, I’m pretty sure it was the excerpt from Alston’s A Realist Conception of Truth, which I found fairly unconvincing. I think he was talking about the notion of “justification” rather than “warrant” though.

  7. Kenny, that sounds like an appropriate location for such an argument, and one that Alston, I’m sure, would endorse. I think you’re right that it’s not very convincing. It’s not clear why the truth connection point assumed (which is itself controversial) has to be built into the meaning of ‘warrant’ or ‘justification’.

  8. It seems to me that there at least an implicit connection between warrant and truth. One can opt out of the traditional answer of ‘true’ to the question ‘What is P warranted as?’ by pointing to other epistemic values, such as predictive power, explanatory success, and so forth. But what makes these epistemic values? One answer is that the statements that possess them tend to be true, although there are exceptions here and there. If the manner in which one replaces truth with warrant in a theory of meaning is to offer some (global) epistemic theory of truth, it seems this answer cannot be offered non-trivially. If, on the other hand, one intends to retain a realist theory of truth but thinks something like warrant or assertibility conditions gives a better account of meaning than truth conditions, then it seems one can explain various epistemic values in terms of conduciveness to truth. However, this latter route seems just as bad, given the difficulties of accounting for compositionality of sentences in terms of assertibility conditions.

  9. David, I don’t see why any antirealist, or other philosopher for that matter, should think about the nature of warrant in terms of the question you pose. Why not simply the question of which beliefs are warranted or, perhaps, superwarranted? To think that a belief is warranted iff believing it to be true is warranted is to make a mistake. People can have warranted beliefs who have no concept of truth at all. Beliefs, and propositions can be warranted, and so can the belief or claim that some proposition is true. It’s pretty obvious, however, that the former is fundamental.

  10. Jon, what I had in mind is a view that statements are warranted or unwarranted only according to the standards of evaluation appropriate to the discourses in which the statements occur. Where the discourses differ, so may the standards of evaluation (consider warrant for “I am in pain” vs. “The Mavericks played a good game tonight”). In lots of cases it is sensible to ask whether P is warranted or not, but this doesn’t show warrant is a property simplicitor. One might say that cases in which it is sensible are those where there is enough shared background beliefs and knowledge of context that the question of what P is warranted as is superfluous, but that otherwise there would be nothing wrong with my question. Perhaps neither question tells us much about the nature of warrant (I don’t put much stock in my own linguistic intuitions), in which case the above view would need further argument for someone who would like to establish the connection between warrant and truth.

Leave a Reply

Your email address will not be published. Required fields are marked *