February 04, 2004

All Conversational Implicatures are Cancelable

and I am the Queen of Romania. I mean that literally.

[Update: Edited for clarity. Go to Liberman's post to see the Dorothy Parker poem that is related to this entry.]

Mark Liberman has said all I wanted to say in response to Allan's post on conditionals. Brief recap: If you say "If you're a good citizen, I'm Donald Duck" it means "You're not a good citizen" because of contraposition; if you say "You're a good citizen, and I'm Donald Duck" it means "You're not a good citizen" because the second clause flouts the maxim "say what you believe to be true"--so that it is clear that the first clause is also meant to violate that maxim.

But, since I'm going to use Mark's post as a peg to hang a crotchet on.

Namely: People widely take it that conversational implicatures have to be cancelable. "Some of you passed the test" implicates "Some of you failed," but you can cancel the implicature by saying "Some of you passed the test--in fact, all of you did."

Timothy Williamson uses this in arguing for the Knowledge Account of Assertion ("Knowledge and Assertion" in Phil. Rev., revised as the "Assertion" chapter in Knowledge and Its Limits; Geoff Nunberg gives it as an explicit test for conversational implicatures; and Brian Weatherson, in his Truer paper, talks about a category of implicature that is "a much stronger implicature than conversational implicature, since it is not cancellable."

I don't see this, and the first paragraph of this post is meant to show it. Does saying "I mean it literally" really cancel the implicature expressed by "I am the Queen of Romania"? I've written a short paper (pdf, word) arguing that it doesn't.

You should read this paper because
(1) It's short (6 pages)
(2) It's probably the first philosophy of language paper ever to discuss the Sex Pistols.

The basic idea is quite simple: The act of explicit cancelation is itself bound by conversational maxims. So even if A implicates B, saying "A. And not B" may not amount to an assertion of A with the implicature canceled--because the maxims might determine that the utterance of "And not B" shouldn't be taken literally.

One could define conversational implicatures as those that are cancelable--but then why would we care about them? I'd say that conversational implicatures should be those that are generated in roughly Gricean fashion, and that not all of these are cancelable.

There's some deeper thoughts lurking in the background.

This is: What we do when we perform a speech act, generally, is to do something that has certain normative consequences. In my dissertation I argue that testimony makes the teller liable to lose some credibility if the testimony turns out false; it also affects the hearer's justification for believing what she's been told. An order makes the orderee responsible for carrying it out, on pain of court-martial or some other sanction. (Reminders and requests are hard to work in, but I think it can be done.)

And you're not always in position to bring about whatever normative effects you please. Obviously you can't just order anyone about; it would be lovely for me if I could make people responsible for doing things, on pain of court-martial, but I can't. But I don't think you can even assert whatever you like. I can't make it the case that a reasonable person would take me to be less credible if it turns out that I am not the Queen of Romania--no matter how much I insist that I am the Queen of Romania, the rational thing to do is to take me as telling some bizarre joke. So--on my theory--I just can't tell you that I am the Queen of Romania.

And when I say "I am the Queen of Romania," I can't cancel that utterance's implicatures and leave the bare assertion that I am the Queen of Romania. The rational hearer won't take my canceling utterance any more seriously than my original utterance.

Anyway, I'd appreciate any comments on the paper, or on the other thoughts herein expressed.

Oh, and the scorecard: In the cases Nunberg and Weatherson discuss, they're right that the failure of cancelability indicates an absence of conversational implicature. But Williamson is wrong--he's laid hold of a Gricean implicature that can't be canceled. I argue for that in my paper "Must We Know What We Say?" (word, pdf).

Posted by Matt Weiner at February 4, 2004 01:38 PM

Relates strongly to the game-theoretic concept of "cheap talk". If implicatures were cancelable, then one prisoner could in a one-shot Prisoner' Dilemma could make the following offer to the other:

"If you say to me that 'I promise I won't talk' and then say 'I mean it, man', then I promise I won't talk either. I mean it, man".

It's crystal clear in a number of economic contexts that there is a "time-consistency problem"; that you can't just by saying something, cause someone else to believe that you mean it, simply because there's enough common knowledge that the other person knows that you will in future have the incentive to break that promise. Hmmmm.

Posted by: dsquared at April 8, 2004 09:34 AM

I didn't know that term. It seems relevant to my main field of testimony--what reason to you have to believe what other people tell you? If you know that the other person is a one-shot maximizer, then you know that you have no reason to believe what they tell you about the PD, and so there's no point in their telling you. (I've ranted about how it often is rational to cooperate in PDs, the idea being that economists should pay more attention to philosophers. Though if economists want to pay more money to philosophers, I'll settle.)

One of the commonly made points in the epistemology of testimony is that you don't usually gain any credibiliity by saying "You can trust me on this." I actually think this can invoke a kind of hyper-credibility--ordinarily I may joke around or shoot my mouth off, which will damage my credibility for ordinary utterances. When I say "You can trust me on this," I may be able to free this testimony from my past violations, if they weren't too serious. But if I turn out to be wrong on one of these high-stakes utterances, my credibility is really shot.

Posted by: Matt Weiner at April 9, 2004 12:53 PM