[Update: A much extended version of this dialogue will appear in the journal Philosophy.]
Alice: I don’t know what you mean by ‘glory’.
Humpty Dumpty: Of course you don’t---till I tell you. I meant ‘there’s a nice knock-down argument for you!’
A: But ‘glory’ doesn’t mean ‘a nice knock-down argument’.
HD: When I use a word, it means just what I choose it to mean—neither more nor less.
A: But what you mean by ‘glory’ is different from what I mean by ‘glory’; which one is what ‘glory’ means?
HD: What does ‘means’ mean? ‘means’ can mean something persons mean and also something words mean. What a person means by a word is what the word means in the language they speak. What I mean by ‘glory’ is what ‘glory’ means in the language I speak; what you mean by ‘glory’ is what ‘glory’ means in the language you speak.
A: Does something "connect" the string of letters ‘glory’ and the meaning?
HD: Are there not two kinds of connection, not one? The meaning-relation for a language connects its words to their meanings. The cognizing relation connects minds to the language they speak.
A: Surely I connect the word and the meaning?
HD: I say that you---your mind---cognizes a structure, a bunch of syntax and meaning functions. You speak a language by cognizing these functions. I speak Humptese by cognizing its syntax and meaning functions; you speak Alicese by cognizing its syntax and meaning functions.
A: I see. The meaning relation has to be distinguished from the cognizing relation. And then, as Professor Saussure insists, the meaning relation is conventional. A word---a string of phonemes or letters---doesn’t mean anything ‘intrinsically’. Professor Tarski’s work seems to make this very clear.
HD: Some authors have attacked Professor Tarski for his clarity about the conventionality of the meaning-relation. The shock! For them, the meaning-relation has to be explained in terms of ‘natural facts’, the use of the string; the mental states involved in speech-acts producing the string; its role in reasoning; its causal connections to its referent; and so on. But these theories cannot even get the simplest fact about language---its conventionality---right! For the conventionality of meaning is explained by the fact that there is not just one meaning relation! For example, with two basic words, and a thousand possible meanings, there are one million meaning functions. When Professor Saussure says the meaning-relation is conventional, all he means is that each distinct function yields a distinct language. Change the function and you have a new langage. It seems to me that these authors confuse the meaning-relation (what a string means in a language) with the cognizing relation (what language a speaker speaks).
A: So, ... the meaning-relation is a mathematical relation. Its conventionality consists in that. I do like the view you advocate. But here is an interesting consequence: I could have spoken a language whose meaning function maps ‘cat’ to dogs. And if I had spoken that language, then I would not have been speaking the language I do in fact speak. For ‘cat’ means cats in Alicese, but means dogs in that language. And if ‘cat’ means dogs in a language, then that language cannot be Alicese. Not only is it a semantic fact that ‘cat’ mean cats in Alicese, it couldn’t have been otherwise: So, Alicese has an essential property: the property that ‘cat’ means cats in Alicese.
HD: Interesting, isn’t it, that conventionality and necessity are so closely linked?
A: But conventionality seems initially to be a matter of choice. So, ‘cat’ can mean one thing and mean another thing! But now I examine your theory, this is merely a multiplicity of mathematical possibilites. When we “choose”, our mind “chooses” one of these, by cognizing it---in a sense, speaking a language is a form of mathematical cognition; but, if this view is correct, then irrespective of which language one cognizes, what a string of phonenes means---relative to that meaning function---is necessary.
HD: For example, the string ‘gloobydooby’ doesn’t mean anything in Alicese! But, mathematically speaking, every word has a meaning relative to some language.
A: Yes, for example, any language in which ‘gloobydooby’ means Liverpool FC. Then it does mean something! But no one cognizes that language. So, another consequence is that the fact that there is no person who speaks a language in which ‘gloobydooby’ means Liverpool FC is no reason to deny there being such a language. Existir es no ser hablado!
HD: I thought Berkeley was Irish? But Latin is old-hat, I suppose. What shall we call these unspoken langages? Disembodied languages?
A: A second consequence of this view is this: if linguistic facts concerning disembodied languages are necessities and not contingencies, then linguistic facts cannot be discovered, or refuted, or explained by empirical science. A contingency cannot explain a necessity. How I happen to use a string cannot explain what it means! The usage somehow explains what I mean, but not what it means.
HD: How you use a word can help explain what language you happen to cognize.
A: Everyone speaks their own idiolect, which may be chopping and changing all the time. Maybe it can change in a nanosecond. Maybe you shift into different languages as the context changes. Why not? But what then is this thing called English? I suppose, it is a language mathematically defined by some grammar textbook, but is a language that no one speaks exactly. After all, not everyone has the same lexicon, the same meanings for words, the same pronunciations, the same pragmatic rules, and so on. If they all vary, even a tiny bit, then they are different. I suppose that, to use Professor Quine’s terminology, languages are individuated very finely: by exact sameness of phonetics, syntax, semantics and pragmatics. We might say that I speak Alice-English and you speak Humpty-English. But these aren’t instances of something more abstract, like Über-English. They are imperfect copies of each other.
HD: You mentioned Professor Quine. Do you recall his famous puzzle? A human rushes past the Mad Hatter who shouts out, ‘gavagai’. The Mad Hatter always tends to behave like this. Let us define Madhatterese to be the language that the Mad Hatter cognizes. Then, given his speech behaviour, what does the Madhatterese word ‘gavagai’ mean?
A: As I can now see, we should formulate the problem by first defining two languages L and L*, such that L maps ‘gavagai’ to humans, and L* maps ‘gavagai’ to undetached human parts (or to the image of the humans under some suitable proxy function). Then there is no question as to what the strings mean. It is a matter of definition that ‘gavagai’ means humans in L and means undetached human parts in L*. So, it really isn’t a puzzle about semantics. However, that said, we still do not know, on the basis of the Mad Hatter’s behaviour, which language Madhatterese is. We do not know if the Mad Hatter cognizes L or cognizes L*. But this is a problem concerning cognition, not semantics.
HD: All of these puzzles about semantics, for which Wittgenstein, Quine, Putnam and Kripke are rightly famous, can be reformulated in the same manner. Suppose that L maps 'number' to the numbers 0, 1, 2, ... while L* maps ‘number’ to elements of a non-standard model M of true arithmetic, then do number theorists cognize L or L*? Or suppose L assigns electrons to ‘electron’ while L* assigns protons to ‘electron’. Do scientists cognize L or L*? Suppose that the symbol ‘+’ means addition in L, and means quaddition in L*. Do we cognize L or L*? These are not problems in semantics at all! They are problems in cognitive science.
A: The problem doesn’t go away though, does it? For we have just moved the goalposts, from semantics to cognizing. For example, now we wish to know if the language that Professor Gowers cognizes is a language in which the meaning of ‘+’ is addition or a language in which its meaning is quaddition. And if the former, then how do we account for this?
HD: Like I said, earlier, when I use a word, it means just what I choose it to mean—neither more nor less!
A: And, in a nutshell, that is the following theory: an agent cognizes L just when the meaning of a word in L is exactly the meaning the agent assigns to that word. So have we solved the fundamental problem of twentieth century philosophy?
HD: Well, not really: for what does ‘the agent a assigns meaning m to word w’ mean?
A: Do I really have to analyse that relationship?
HD: Professor Putnam insists that it would involve “noetic rays”.
A: Well, I do assign meanings to words! I don’t know how, but surely I do.
HD: These meanings being what? Extensions, referents, intensions, what?
A: I think meanings have to be concepts---Fregean concepts.
HD: You---your mind---assigns Fregean concepts to words. Isn’t that Platonism?
HD: And you speak L if and only if L assigns to each word exactly what your mind assigns to it?
HD: One thing that occurs to me is this. If you speak L and I speak L*, then we speak different languages; this makes communication occasionally non-optimal; should you change or should I change? Neither?
A: This is the question of collective linguistic normativity. There are only two reasons I can think of for changing my idiolect somehow. The first is to improve social co-ordination: then it’s easier for each of us to process incoming messages from others. The second is that my meanings---concepts---might be gerrymandered in the manner that has been discussed by Professors Goodman and Miller.
HD: I am not going to change my idiolect just to suit the demands of others! They can interpret me if they make a bit of an effort.
A: Despite your protestations, I bet you do modify your idiolect on some occasions. Several years ago, I noticed that experts in areas related to law, politics and epistemology always assign the concept UNBIASED to the string ‘disinterested’, whereas I, and quite a few other speakers, sometimes assign a different concept, NOT-BEING-INTERESTED to the string. When I noticed this, I deliberately changed my idiolect. I deferred to expertise in order to improve social co-ordination. In some cases, often with technical words like ‘isomorphism’, I’m sure that the concept I assign to the word is muddled and unclear. So, I defer to the expert, and modify my idiolect slightly.
HD: Well, I suppose in that sort of case, we decide to do that, yes. Isn’t this the topic of linguistic prescriptivism that many linguists get all huffed up about?
A: Yes: but they are being silly. Social co-ordination, expertise, and so on, all involve normativity. It is rational for humans to improve co-ordination and to share expertise. Still, I’m not sure about the other reason. It would require a kind of hierarchy or ordering of concepts: some concepts are more natural than others, while others, like GRUE, are gerrymandered. Perhaps that is right.
HD: If a language assigns some meanings which are gerrymandered, then the language itself would be somehow deficient. So, some languages would be "better" than others!
A: Yes, but it’s damn difficult to be convinced that this is right. Maybe it’s connected to simplicity somehow. But it’s very hard to see how to make sense of ideas like concepts being more natural or the world being simpler.
HD: Anyway, good luck with all that, Alice! Disembodied languages, meaning-functions, noetic rays and natural concepts! Let’s call your cognizing function the noetic-function. Speaking a language L involves the equivalence of a noetic-function and L’s meaning-function. Do you think you can get a funding agency to support that? Your research career is ruined.
A: Why should I care?
HD: You’re right.
A: So, the question is how our noetic rays can make words mean so many different things.
HD: The question is which is to be master---that’s all.
A: But with that, Mr Dumpty, I disagree!
[This dialogue is based on ideas from a couple of talks I've given over the last few years. One, "Meaning, Use and Modality" in Madrid in 2008; another, "Cognizing a Language", at the Edinburgh Linguistics Society in October 2010; and a related talk, "Deflationism and Representationalism", I gave in Vienna and Munich in March 2012.]