The past few days, I was in DC for the annual Georgetown University Roundtable, where I gave a plenary talk on conditionals. The slides for the talk (supplemented with two slides that clarify a point that I failed to make coherently in the talk) can be downloaded, but I don’t know how much sense they make without me talking at the same time (and even then …). The talk is an evolution, with a new twist at the end, of my talk at the conditionals conference at UConn last year.
I hope to be able to write up this material in a proper format some time soon. I’d appreciate comments or questions.
PS. I would bet a lot of money that this is the first talk that cites Bertrand Russell and Dolly Parton.
How much?
The Economics of Defined Benefit Pensions and the Rationality of Pension Funding Strategies (see p. 2 and p. 17).
March 12th, 2007, at 2:11 am #David: Oh well, I forgot the cardinal rule that one shouldn’t forget to google before making such rash claims. At least, he is not also citing Ockham and Cicero.
March 12th, 2007, at 2:26 am #Hi Kai - I’m sorry I wasn’t at Georgetown this year to hear your talk (as you know, I’m spending my sabbatical away). I read through the slides, though, and found them pretty clear even without the presenter — and very interesting. Here would have been my questions if I had been back home in DC:
The arguments you discuss that indic conditionals lack truth conditions really both seem to me to be arguments that the meanings of conditionals are not completely analyzable in terms of truth conditions. That is, they may have truth conditions, but if they do, they have more meaning beside that. Do you think this is accurate?
Your proposal doesn’t seem to relate the debate about conditionals to a very similar issue in the semantics of epistemic modals, but I suspect that there is a connection.
Let me explain: In the talk, you work with a traditional, truth-conditional theory of the semantics of epistemic modals, adding that “there are some other components of meaning [for epistemic modals] (evidentiality in particular), but this will do for now.” To this I would add subjectivity, in Lyons sense - which may be the same thing as evidentiality, I’m not sure. Lyons thinks subjectivity involves a difference at the level of speech acts, and I think you agree (or agreed a few years ago); there are other theories as well (e.g., Papafragou, several functional linguists, me myself; Kratzer 1981 has a brief discussion, though her analysis puts subjectivity into the truth conditions). Whatever the right theory is, it seems that this subjective/evidential aspect of meaning dominates intuitions about what epistemically modalized sentences “mean” in simple root sentence, hence definitions of epistemic modality involving level of commitment and the like. But this does not show that they lack truth conditions; subjectivity/evidentiality may be part of truth conditions or in addition to truth conditions. So the situation may be more or less similar to indic conditionals.
Now, Bennett makes his claim based on the intuition that the meanings of indic conditionals are “subjective”. Given this, one possibility would be to analyze the intuition of NTV in terms of ‘if’ restricting an epistemic modal, and this having an effect on the “other components of meaning” (evidential or subjective) of the modal. In other words, people who support the NTV hypothesis are attending to the effects of the conditional on the evidential/subjective meaning of the covert epistemic modal.
Now I can really ask my questions: Do you think that Bennett is on to something when he says that conditionals are subjective? Do you think that this subjectivity is the same as that discussed in connection with epistemic modals? Do you wish to make any bold claims concerning its nature (speech act-y, evidential, truth conditional, etc.)?
March 12th, 2007, at 5:21 am #Hi Paul — thanks for the comments and questions. Here are some unordered reactions:
- I have maintained for a while that some of the funny properties of indicative conditionals that NTVers highlight should follow from them being epistemic conditionals (in Lewis/Kratzer/Heim terms: epistemic modals restricted by an if-clause). So, yes, I think that is the right approach. I guess I didn’t stress it in this talk because I’ve said so a number of times before (quite clearly in my UConn talk, I believe, but also in my UMass colloquium in 2003).
- Another proponent of such a view is Brian Weatherson, who has a paper where he connects an epistemic analysis of indicative conditionals to the kind of assessment-relative semantics for epistemic modals that he has argued for.
- And Tamina Stephenson is working on a chapter of her dissertation where she will extend her analysis of the subjectivity of epistemic modals to the data on indicative conditionals.
- There are some comments on the peculiarities of epistemic modals in the papers I have written with Thony Gillies: “CIA Leaks” and “An Opinionated Guide to Epistemic Modality”. Our “magnum opus” on epistemic modals is under development and we hope to write a paper on epistemic conditionals at some point as well.
- The subjectivity of indicative conditionals should indeed entirely trace back to the subjectivity of epistemic modals, however that is analyzed.
- But there are other problems as well. In particular, there are the issues with embedding under probability operators and the cross-speaker anaphora cases I discuss in the GURT talk. Here, I believe, the solution is in the Lewis/Kratzer/Heim analysis or, rather, in the Belnap/Lewis variant, if that is indeed the best way to deal with the cross-speaker anaphora problem.
March 13th, 2007, at 9:57 pm #