Thursday 20 March 2014

Empty Names and Negative Existentials

Singular negative existential propositions such as 'Santa Claus does not exist' can be made to look puzzling without any explicit theoretical view on board, with Socrates-style questions such as: how can anything not exist? How can one ever truly say that something doesn't exist? For if one is right, there is no such thing as what one is talking about, and therefore one is talking about nothing.

Or: 'Santa Claus doesn't exist' – what is this thing which doesn't exist? How could there be such a thing?

For those who have a Millian conception of naming (and associated views of propositions), the problem of negative existentials assumes a particularly acute form. If all there is to the meaning of a name is its bearer, and if substituting co-referring names does not affect the 'proposition' (not my usage) expressed by the sentence, then how can a statement like 'Santa Claus does not exist' mean anything at all? Furthermore, how can it be true? And how can different true negative existentials have different meanings, as they seem to do?

All these questions push the Millian toward analysing existence statements – giving an account of what they really mean. (Witness for example Kripke's tortured discussion in Reference and Existence. Of course, he never officially and unequivocally endorses Millianism, but he's flirted outrageously with it in public and finds it intuitive.)
Having a non-Millian conception of naming such as the one I propose, on which names are recognized as being tied to individual concepts (or having uses - roles in the systems of language they occupy - which are semantically relevant), makes it a lot easier to answer these questions. But this does not mean that we are using individual concepts (or name-uses) as elements in an analysis of existence statements.

'N does not exist' does not, for instance, mean exactly the same thing as'“N” has no bearer' or 'The “N”-concept has empty extension' (that is, on a natural and sufficiently fine-grained conception of proposition-meanings): the existential proposition is not about a name or a concept.

We can say it is about N, if we understand 'about' as not having existential import, or we can say that it is not about any real thing, but it would muddy the waters intolerably to say that it is about a name or a concept.

And yet we can see what would make a person want to say that it is about a name or a concept. What we can truly and properly say is that, in 'N does not exist', the function of the name 'N' is not to pick out an object – rather, this name (rather than some other name) is used in order to bring a particular individual concept (or name-use) into the act (though I would not want to say 'under consideration', for it need not become an object of thought).

But then what do we say about the function of a name in a proposition like 'John is tall'? It is no less true to say that 'John' functions to bring a particular individual concept or name-use into the act, but here we can also say that it functions to pick out an object. But some have found it intuitive to say that a name functions purely to pick out an object. Given a certain very narrow concept of 'function', this is fine too, although it could be misleading – it could lead to the troubles of Millianism.

Let us now take a representative selection of the questions raised at the beginning of this section, showing how they can be answered with the conceptions of propositions and naming I favour. This is done without giving an analysis of existence statements (in the classical sense of giving something else which they are then said to mean). As with identity statements, the trick is to treat existence statements on their own terms, and to recognize that they occupy a special role for us, and work in a quite particular way.

How can anything not exist?

Just as we sometimes use names in a way which carries existential import, as in 'John is tall', and sometimes use them in a way which does not, as in 'Santa Claus does not exist', 'Some children believe in Santa Claus', we use terms like 'there is', 'something' and 'anything' in two different ways: with or without existential import. A clear example of the latter sort of use would be a kid saying 'I don't believe in Santa Claus, the Easter Bunny, or anything like that'.

The difficulty and puzzlingness of the above question derives from this ambiguity. The thorough answer is: in the sense of 'anything' etc. in which they carry existential import, nothing can fail to exist – i.e. it is not the case that it is possible for anything to not exist, but in the sense of 'anything' etc. in which they do not carry existential import, there are indeed things which do not exist.

At this point, individual concepts (notions of things), or name-uses can be brought into consideration, to help us make sense of the fact that we talk like this. What gives rise to it is that we sometimes have individual concepts without objects, name uses where the name lacks a bearer. We then formulate propositions which, if we treat them by analogy with propositions like 'John is tall' and 'Someone is in this room', look as though they would have to involve (existing) things in order to be true, in the way that these propositions would have to involve John and a person in the room – but in fact they face no such requirement. We use them in connection with objectless concepts, bearerless names etc. (and this connection is quite different from that which holds between 'John is tall' and John himself).

How can one ever truly say that something doesn't exist? For if one is right, there is no such thing as what one is talking about, and therefore one is talking about nothing.

In light of the above, this question can be disposed of quickly. We can truly say that something doesn't exist by using 'something' in the sense in which it doesn't carry existential import, and in virtue of the fact that we have objectless individual concepts and involve them in our talk. In the sentence after the question ('For if one ...'), 'there is no such thing as' and 'nothing' are used in their existential-import-having senses, and so there is no real conflict in what is being said here. It is just being said in a potentially misleading way.

How can a statement like 'Santa Claus does not exist' mean anything at all?

The proposition works by means of the fact that the name 'Santa Claus' brings an individual concept (or a way of using a name) into the act – not by referring to it, but because that is the concept tied to that name (or that is the way that name is used). The proposition is true iff the concept (or name-use) of 'Santa Claus' has an object.

This is not to say that the proposition means the same as any proposition about concepts or name-uses, or that the proposition holds of just the same possible situations as those of which what is said on the right hand side of the 'iff' holds. We are using the biconditional here not to give an analysis but to give a necessary and sufficient condition for the proposition in question, which we have before us, actually being true.

How can different true negative existentials have different meanings, as they seem to do? 

By bringing different individual concepts (or name-uses) into the act. 


Kripke, Saul A. (2013). Reference and Existence. The John Locke Lectures. Oxford University Press. 

Monday 3 March 2014

Kripke's Puzzle and Semantic Granularity

(Added October 2016: my most up-to-date treatment of granularity can be found in Chapter 6 of my PhD thesis. This is an early, undeveloped attempt.)

This is the post where I first introduced my doctrine of semantic granularity. Follow-ups so far:

Facts and Granularity
Granularity and Quine
Metaphysical Realism and Conceptual Relativity: An Application of Granularity
Granularity and Relativism about Truth
Granularity and the Paradox of Analysis
The Principle of Compositionality and Semantic Granularity
Two Opposite Types of Granularity Difference

Meanings of expressions and belief-contents can be carved up at different granularities. That is, it can sometimes be the case that, when operating at one granularity it is correct to bundle two expressions or beliefs together as having the same meaning or content, while at another granularity it is correct to put them in separate bundles. I think this must be acknowledged in order to fully solve Kripke's puzzle about belief.

There are aspects of the puzzle which do not require this - i.e. the puzzle has some morals which do not involve this. But until semantic granularity is recognized there will be a remainder.

One aspect of Kripke's puzzle is like Frege's puzzle: we need difference-makers for 'Londres' and 'London' and the propositions they appear in, so that we can avoid the conclusion that Pierre here believes some proposition as well as the negation of that very proposition (i.e. we need to differentiate his 'Londres'-mediated beliefs from his 'London' ones). I do this with my accounts of names and propositions.

Another, closely related, aspect of the puzzle is that, once we have the required difference-makers, we need to put them to work somehow in distinguishing the sense in which Pierre has inconsistent beliefs from the sense in which he does not have inconsistent beliefs. Accordingly I distinguish internal and external inconsistency. Two beliefs are internally inconsistent iff no two beliefs with the same internal meaning could both be true. Two beliefs are externally inconsistent iff those very two beliefs, with their actual external projective relations to reality, could not both be true. Internal inconsistency implies external, but not the other way around. And one of the morals of Kripke's puzzle is that merely internal inconsistency does not constitute irrationality.

But there is a further aspect that remains puzzling even with the required difference-makers, and the distinction between internal and external consistency with its associated moral about rationality. And this comes out in Kripke's summing up of the puzzle: does Pierre, or does he not, believe that London is pretty?

And here, to see that this is still puzzling, it is very important that we take to heart Kripke's stipulation that he is using the language of belief reports not in a de re sense, but in a de dicto sense; that he is using forms like 'S believes that a is F' not in the sense of 'S believes, of a, that it is F' or 'S has a belief concerning a to the effect that it is F', but to actually specify belief-contents. Kripke gives a supplementary explanation of his meaning by saying that we could emphasize it by putting a colon in place of the that-clause: 'S believes: a is F'.

It is important to take this to heart because, if we stick to a de re sense, we can give an answer with what we already have; we can say, in answer to Kripke's puzzle question reproduced at the end of the second last paragraph, 'He does; Pierre believes, of London, via his "Londres" concept (or via his symbol "Londres" with its attendant use or internal meaning) that it is pretty. But he also believes, of London, that it is not pretty, but in that case his belief goes via his "London" concept (or via his symbol "London")'.

And this is just using the stuff we needed anyway to solve Frege's puzzle. And furthermore we can add that there is no irrationality on Pierre's part here, since his two conflicting beliefs, while externally inconsistent, are internally consistent, and he doesn't know that they concern the same object.

But this doesn't enable us to answer Kripke's puzzle-question as he intended it, namely in a belief-content specifying sense. Indeed, it can seem to be part of the problem. We wanted to allow that Pierre is not being irrational, and distinguish his 'London' concept from his 'Londres' concept. But then what was going on when, in the first part of the story, we felt the pull of saying that Pierre believes that London is pretty – i.e. that Pierre believes the same thing that we mean when we say 'London is pretty'?

The solution is to see that a shift in granularity has taken place, and that the answer to Kripke's question - indeed, the meaning of that question - depends on what granularity one is operating at. In the first part of the story, we naturally go for a granularity coarser than the one we will end up at, in order to capture in an efficient way what Pierre's and our contents have in common. Then, when the special “splitting” (mistaking one for two) situation arises, it becomes much more convenient to describe the situation using the same device of belief reports, but at a finer granularity. 

Kripke's puzzle is puzzling because one part of the story induces one granularity, and another part induces another. With granularity kept in the background as an unarticulated and untheorized contextually variable aspect of the sense of belief reports, the results seem to contradict each other. Once we realize what is going on, the results can be seen to be no more contradictory than 'All the beer is in the fridge over there', under a certain natural contextual restriction of quantifiers, is of 'There is beer at the pub'.

Philosophers already talk about different granularities, but generally the distinction is made between two quite different notions: for example, the set-of-worlds conception of propositions is said to be more coarse grained than the Russellian. Here, I am keeping to one conception (which, in comparison with those just mentioned, is left more intuitive), but saying that we can operate at different granularities in individuating meanings, roles in language systems, and the contents of beliefs. The idea is that semantic notions such as that of synonymy and belief content are flexible devices, in that they can be used to bundle expressions and representations together in multiple ways.

The underlying idea here, analogues of which appear in connection with other things besides linguistic meaning and the content of belief, is quite commonsensical. For example, consider someone who takes a board game and alters some rules, inaugurating a social institution of playing to the altered rules which goes on along side the practice of playing the original game. Are we to say there are two different games here, or two different versions of the one game? It seems like common sense to say that one can say either. It's not as though there's some answer here which we haven't yet managed to find out. So, we individuate games at different granularities. And this is part and parcel of the usefulness and flexibility of our concept of a game. I think the same holds for the concept of meaning.

While this idea is quite commonsensical, the idea that it should be taken seriously in analytic philosophy of language appears quite radical. (It is as though, without really thinking it over, people have rejected any such move as inherently inimical to analytic conceptions and methods. A bit like vagueness before analytic philosophers began taking that seriously.)

Interestingly, after I had independently started applying the terminology of granularity and bundling to the matters of Kripke's puzzle and internal meaning, I found that AI researchers working on word sense disambiguation have been talking the same way (without apparently reflecting much on it philosophically, let alone from the point of view of the problems of analytic philosophy of language).

In this post I have tried to introduce the doctrine of semantic granularity, a doctrine which has come to assume an important role in my thinking. I have motivated it in the first instance using Kripke's famous puzzle about belief, which is also how I arrived at it. In subsequent posts I will develop the idea further and outline further applications of it.

[Here is one further application. - TH 25/8/14.]


Kripke, Saul A. (1979). A puzzle about belief. In A. Margalit (ed.), Meaning and Use. Reidel. 239--83. [Online here and here as of 4/3/14.]