blog.kennypearce.net

The Problem of Analyticity

The new quarter has begun, and I have just finished reading Quine's "Two Dogmas of Empiricism". One of Quine's chief purposes here is to argue that the difference between 'analytic' and 'synthetic' truths is one of degree, and not of kind, so that there is no neat division between the two. I want discuss this difficulty here, although I shall treat it slightly differently than Quine does.

Anyone who has been exposed to post-Kantian philosophy is familiar with the distinction between the analytic and the synthetic. For instance, 'no bachelor is married' is an analytic truth, whereas 'no bachelor lives in my apartment' is a synthetic truth. To a first approximation, we might say that the truth (or falsehood) of an analytic statement depends on the meanings of the words and the structure of the statement, whereas the truth (or falsehood) of a synthetic statement depends on the facts about the world. This is rather fuzzy, but I think it can be turned into a statement which does a good job of capturing most of our (philosophical) uses of the terms 'analytic' and 'synthetic':

A mistake about the truth-value of an analytic statement is rightly attributed to lack of linguistic competence; a mistake about the truth value of a synthetic statement is rightly attributed to lack of acquaintance with the facts of the world.

This, as I said, captures most of our uses of the terms involved. For instance, if you say 'there is a married bachelor,' I will suppose that you don't know the meaning of the word 'bachelor'. On the other hand, if you say (to me) 'there is a bachelor living in your apartment,' I will suppose that there are certain empirically determinable facts with which you are not acquainted, and I might provide evidence of these facts. Consider, however, the following dialog, in the spirit of Donald Davidson:

A: I have several rhinoceroses in my refrigerator.
B: Rhinoceroses fit in your refrigerator?
A: Yes. I could fit lots of them in there if I wanted to.
B: What do you do with them?
A: Each morning, I squeeze two of them into a glass and drink it with breakfast.
B: How?
A: By hand. It takes a little while, but it helps me wake up in the morning.
B: Where do you get these rhinoceroses?
A: From the grocery store.
B: Where does the grocery store get them?
A: They grow on trees.

Davidson rightly points out that at some point in a dialog like this one B ought to begin to doubt A's linguistic competence: that is, B should suppose that A does not mean the same thing by 'rhinoceros' which other English speakers do. Which of A's statements is analytic, then? None of them, it seems to me. Perhaps, though, someone will suppose that it is an analytic truth that rhinoceroses do not grow on trees. Surely, though, at some point before the end of the dialog B ought to conclude that A's problem lies not in lack of acquaintance with the facts about rhinoceroses, but, rather, that A uses the word 'rhinoceros' to refer to something other than 'rhinoceroses'. Clearly, however, it is a synthetic truth that rhinoceroses do not fit in refrigerators. In fact, it is in principle possible to build a refrigerator capable of accommodating a herd of rhinoceroses.

This is more or less Quine's point: the 'factual' and 'linguistic' components of meaning/truth are not neatly separable.

Why is this? Well, it seems to me that there is an obvious reason why this is so, which Quine does not mention, perhaps because he was writing in the year 1951: language takes place in the world and, therefore, facts about language just are facts about the world. Of course, then, it is not possible to neatly separate the two.

We might try another direction to distinguish the two and do a sort of thought experiment. Suppose we consider some sentence S. If S's truth-value can be changed by changing only non-linguistic facts, then S is synthetic, but if S's truth-value can be changed only by changing the linguistic facts, then S is analytic.

A problem suggests itself. Let S be the English sentence 'There have been, in history, at least two people such that they spoke to each other in English.' S ought to be synthetic. However, by our standard, it might very well come out analytic because, arguably, if S was false then there would be no linguistic facts about English, so rendering S false would require changing the linguistic facts about the language in which S is expressed.

This problem could be solved by a sort of platonism about language: if the language exists in 'Platonic heaven' regardless of whether or not anyone in fact speaks it, then I think we might be able to get this second definition to do some work.

Posted by Kenny at January 5, 2009 4:50 PM