To: Chris M.
From: Geoffrey Klempner
Subject: Grice on conversational implicature
Date: 25th October 2011 14:16

Dear Christian,

Thank you for your email of 16 October, with your essay written under exam conditions for the University of London BA Philosophy of Language module, in response to the question, 'How does Grice distinguish between the semantic and the pragmatic contributions to what speakers convey by their utterances? How plausible is this distinction?'

This essay gives a lot of information about Grice's theory of conversational implicature, and raises difficulties with the idea (for example, the extent to which we can write the 'rules' for conversational implicature). You also mention views (Searle, or Borg, Cappelan, Lepore) which are hostile to the idea. For example, according to Searle, you say, 'there is no literal 'semantic' meaning', even in such a seemingly straightforward statement as, 'The cat is on the mat.'

What I didn't get from this is any sense of why the distinction between semantic properties and pragmatic considerations matters. What turns on the debate? Why are we interested in semantics at all? What is the philosophical significance of the claim that there is such a thing as 'semantic meaning' of a word, phrase or sentence? Why, for that matter, do we study the philosophy of language?

Grice's own program of defining meaning in terms of speaker's intentions is one thing we could talk about. But the issue is somewhat clearer if one looks at Davidsonian semantics. The idea behind Davidson's program is to account for a speaker's ability to produce and understand a potentially infinite number of meaningful statements in terms of an understanding of the semantic properties of a finite number of semantic particles: proper names, predicates, logical constants, quantifiers, etc. A Davidson-style 'theory of meaning' is an axiomatic system from which one can derive the meanings of indefinitely many sentences.

One comes across the idea of 'theory' in other areas of philosophy. A theory of ethics would tell you what the appropriate course of action would be in any given set of circumstances. It is getting less controversial that there cannot be a theory of ethics, because the situations requiring ethical choice are too complex. There will always be irresolvable ethical dilemmas. Williams argues something along these lines in his 'Ethics and the Limits of Philosophy' (as indeed I do in the Moral Philosophy program).

The situation with regard to theories of meaning does not seem so bad. Arguably, Grice's program is not inconsistent with Davidson (see Mark Platts 'Ways of Meaning'). Frege, Tarski and the idea of truth-conditional semantics seems a good, solid foundation for accounting for the endless variety of linguistic usage -- provided one is able to make the appropriate 'cut', separating off core semantic meaning from pragmatic considerations. (One annoying issue here is context dependence: how does this differ exactly from pragmatics, where does one draw the line? That's something you could have mentioned in your essay.)

Consider a seemingly simple case. There was a time when philosophers of language were happy to defend a truth functional analysis of indicative conditionals. I would still favour such an approach, despite recent work (e.g. by my old tutor Dorothy Edgington) which seems to show a far greater complexity or 'meatiness' in the semantics for indicative conditionals, involving probability theory.

Grice's theory seems to offer a powerful defence of a truth-conditional account of indicative conditionals, which happily copes with the so-called paradoxes of material implication etc. But how good is it, really?

Let's say, I am near a lift and you ask me, 'How do I get the door open?' I reply, nonchalantly, 'If you press the green button, the lift door will open.' You press the green button, and the lift door opens. 'Thank you!' As you disappear into the lift I allow myself a little chuckle. I could see from the panel that the lift had arrived. And I knew that the door opens automatically after 10 seconds. The green button has no effect whatsoever because it's purpose is to call the lift.

Did I lie? I spoke the literal truth. It is not the case that you press the green button and the door does not open. That's what the indicative conditional, 'If... then...' means. 'If P then Q' is equivalent to 'not-(P and not-Q)'. Frege held this. (Interestingly, Frege offers an example where there is no misleading implication of a direct connection between antecedent and consequent, 'If the sun has not gone down then the sky is very cloudy.') What I said is literally 'true' because the door does open. But what you inferred from my statement is that pressing the green button causes the door to open.

How would one adjudicate a debate between someone who holds the truth-functional view of indicative conditionals and someone who holds that the semantic content of conditionals includes an implication about a causal connection between antecedent and consequent? (If there is no direct causal connection, as in Frege's example, then the claim would be that there is an implication that assuming the truth of the antecedent has an effect one's judgement of the probability of the truth of the consequent.)

If you could answer this question -- or pick an issue that you find gripping -- then you would have given the reader a much clearer idea of why anyone would want to distinguish between semantics and pragmatics using Grice's notion of conversational implicature, as well as providing a useful test case for Grice's theory.

All the best,

Geoffrey