Two Dogmas of Empiricism

The two dogmas are

I take this to be an important analogy:

Quine::Carnap = Austin::Ayer.

We have devoted the semester to Austin's criticisms of Ayer's presentation of and his version of foundationalism. Quine's "Two dogmas" is a criticism of Carnap's presentation of and his version of foundationalism.

Austin's criticism of Ayer was regarded as devastating, and it was tremendously influential in Oxford and Cambridge. (I'm indulging in a bit of oversimplification for literary purposes.) Quine's criticism was regarded as a devastating criticism, not just of Carnap, but of Logical Positivism, which was, at least according to hindsight, the dominant philosophy among all English-speaking philosophers for most of the 20th Century. The story of dominance is so extreme, that it is common today for Ayer to be described as a Logical Positivist (even a "leading" Logical Positivist), though, by his own characterization, he was no such thing. His views, however, are similar to those of the Logical Positivists in many important respects.

I have described, in this course, Austin's criticisms of Ayer as being the end of foundationalism. That is not completely, or even primarily, true, as a matter of the historical record. Quine's criticisms of Carnap, and, in particular, "Two Dogmas," was really what did the job. So, today, I'm going to focus more on the ways in which Quine's criticisms are like those of Austin than I am on the details of Quine's criticisms.

Why did I focus on Ayer::Austin instead of Carnap::Quine in the class? If you thought Ayer was hard to read, you should try Carnap. Also, though it becomes clear from correspondence between Quine and Carnap (Dear Carnap; Dear Van) that "Two Dogmas," and most of everything else Quine wrote was directed at Carnap, Quine almost never makes that explicit. Austin, on the other hand, cites Ayer chapter and verse, which makes the argument much easier to follow.

Aside: The last chapter of Sense and Sensibilia criticizes Warnock. Warnock is the author of Sense and Sensibilia.

Quine's second dogma, about "reductionism," is largely a corollary of his first dogma, that there is no analytic-synthetic distinction. I think that that is true and as important as Quine did, and maybe even more important. I may be the only one in our department today who rejects the analytic-synthetic distinction in any form, though, of course, present-day versions of the distinction incorporate some awareness of "Two dogmas," and hence are not exactly the same as the one Quine criticizes. I think they fall to variants of Quine's arguments.

What is an analytic-synthetic distinction? A truth is analytic if it is a truth of reason (Leibniz). A truth is synthetic if it is not analytic, and synthetic truths are taken to be truths of fact. An epistemological consequence is usually taken to follow from those notions: If a truth is analytic, it can be known to be true simply on the basis of an understanding of what it asserts. If a truth is synthetic, one also needs to know something about the world to know whether it is true.

Since Quine doesn't think the distinction is coherent, he doesn't think that there are any examples of analytic or synthetic truths, but here are the kinds of truths that have traditionally been taken to be analytic or synthetic, examples of the sort that Quine discusses:

Analytic
A bachelor is an unmarried male. (Involves knowledge of the meanings of the words used.)
Analytic
Jack and Jill went up the hill if and only if Jack went up the hill and Jill went up the hill. (Involves logical principles.)
Synthetic
Brutus killed Caesar.
Synthetic
There are brick houses on Elm Street.

What does any of this have to do with Ayer? Ayer's perceptual truths and empirical truths are all synthetic, and, if there is no distinction between analytic and synthetic, there is no distinction between empirical or perceptual properties and any other properties. Ayer talks about adopting different languages that are "empirically equivalent." That amounts to claiming that the principles that are just part of the language are analytic while the empirical facts on which such languages agree are synthetic. In addition, Ayer admits that his incorrigible truths might be incorrect when they involve a misuse of language (that is, when they violate some analytic truths) but that they are incorrigible with respect to the empirical data (that is, they cannot violate synthetic truths).

Ayer's doctrine, though he doesn't put it that way, is imbued through and through with an analytic-synthetic distinction. Like Ayer, Carnap thinks that we can and should adopt special languages for special purposes. They use very parallel terminology, and Quine's criticism applies nearly unchanged to Ayer. Austin does not, so far as I can see, ever attack precisely the analytic-synthetic distinction, but, when he attacks "false dichotomies," many of the dichotomies he actually attacks can be taken to be cases of an analytic-synthetic distinction. His argument that particular sentences can be used in different ways in different circumstances, accompanied by unexpected examples, is much like Quine's argument that particular sentences can be taken to be more on the analytic or the synthetic side in particular cases.

Note that it is critical for Ayer that the distinction between truths of language (analytic) and empirical truths (synthetic) be a fundamental distinction, a genuine dichotomy, since, for example, if there were intermediate possibilities, his argument for incorrigibility would fail. Austin, in effect, notes other examples. It would not do for Ayer if it were merely the case that some statements were more on the analytic side, others more on the synthetic side. The same is true of Carnap, though I won't stop to explain. The point is important since many modern defenders of a distinction of analytic-synthetic type only defend a dimension on which some statements are closer to one side than the other. Quine agrees that there is a such a dimension and explicitly endorses it. Somehow it has become common to take Quine to be denying even that.

Now, to Quine's actual argument. Austin breaks down a host of "false dichotomies" by providing examples that fall squarely in the middle of the supposed gulf between the two sides. Quine takes a group of false dichotomies, analyzes them all as having their root in the analytic-synthetic dichotomy, and then (though he does produce examples in the middle—"everything green is extended") he primarily argues against the distinction by showing that it has not and cannot be characterized, defined, described in any noncircular way. That (unlike Austin's argument) leaves open just taking the distinction to be fundamental, well-understood, not in need of further definition, and the like. Quine might have argued against that as follows (though he never did, the ingredients of my proposed response are in what he did say): The distinction is not one taken from ordinary language. The terms "analytic" and "synthetic" are technical philosophical terms par excellence, and so they don't have any prephilosophical meaning. Thus, they, unlike terms like, for example, those in "The earth has existed for many years past," do need to be introduced somehow, and there just isn't any way to do it.

Of course, that isn't exactly true: we do introduce the terms, much as Ayer introduced the term "material object," by means of examples. That method was criticized very effectively by Austin, and, at any rate, the method is not suitable for introducing a sharp dichotomy, only a dimension, and Quine doesn't object to the idea that there is such a dimension (though he thinks it is a pretty useless one).

Since "synthetic" just means "not analytic," the term we need to introduce is "analytic." I gave two examples of "analytic truths," one of which is true in virtue of meaning and the other of which is a truth of logic. Quine does not object to the idea that there is a sharp distinction between truths of logic and other truths, it is the part about meaning he criticizes. Since "A bachelor is an unmarried male" is analytic, so is "A bachelor is unmarried." Analytic truths can combine issues of meaning with some logic. What is it to say that two words mean the same thing? We cannot put the meanings on display and compare them. Meanings aren't like that. Suppose we just try taking "meaning the same," that is, synonymy, as basic, and defining analyticity from synonymy. Here is the proposal:

A sentence expresses an analytic truth if it can be converted into a logical truth by substituting synonyms for some of the words in it.

Examples:

A bachelor is an unmarried male. A bachelor is a bachelor.
A bachelor is unmarried. An unmarried male is unmarried.

That works, so why aren't we done? What is it for two words to be synonymous? It is exactly that they "mean the same thing," that is, that a sentence asserting their interchangeability is analytic. For example,

The first sentence is true if and only if the second is analytic. All we have shown is that defining "analytic" and defining "synonymous" are the same problem, we haven't made any progress toward solving either one.

So, what are some ways one might define or characterize synonymy?

  1. Two words are synonymous if they mean the same thing by definition.
  2. Two words are synonymous if they are everywhere interchangeable in a truth preserving way.

Definition
Normal words have meanings before they appear in dictionaries. ("Mcjob" recently made it into the Oxford English Dictionary, and McDonald? 's has been trying to claim that they defined it incorrectly: jobs at McDonald? 's, according the McDonald? 's publicity people, are wonderful jobs.) It isn't up to someone who writes a dictionary to decide what words mean, which words are synonymous, that is what it is before the dictionary is written, before the definitions are given, and so definitions report synonymy, if there is such a thing, they don't establish synonymy. Thus, definitions are beside the point in trying to see what synonymy is.

There is an exceptional case: on occasion, words are introduced into a language by explicit stipulation of a definition (radar is radio detecting and ranging). In such cases, the defined term is guaranteed to be synonymous with its definition in virtue of the definition. That isn't very helpful for two reasons: it is an exceptional case and it doesn't give us any information about what we mean by "synonymy," which was supposed to be the goal.

Interchangeability Salva Veritate (preserving truth)
The case of stipulative definition guarantees truth-preserving interchangeability. That suggests using such interchangeability to define synonymy.

Two definition I learned from Quine:

renate
Having a kidney.
cordate
Having a heart.

Quine takes it to be the case that everything that is renate is cordate, and vice versa, and so, much of time, "renate" and "cordate" are interchangeable:

All renates are cordates. All renates are renates.
That won't do. To get the right answer from the test we need sentences like
Necessarily, all renates are cordates.
Unfortunately, a sentence is "necessarily true" if and only if it is analytic. We are going in circles.

If there is no analytic-synthetic distinction, there is no distinction between "the facts" and our theories about them, nor is there any particular set of facts that any single sentence asserts to hold. The meaning of a sentence is determined by its relation to other sentences and to experiences not one by one, but as a whole. There is no such thing as an experimental or observational test of the truth of a single sentence.

What is left? Our picture of the world as a whole. We, in response to experiences, adjust our beliefs and language. Sometimes the changes are local: we just change a single sentence. Sometimes we change the very way in which we talk about the world. There is not basic distinction that makes it necessary to do one or the other in any particular case. We pragmatically adopt theories beliefs and facts that are most useful for our form of life.

-- ShaughanLavine - 24 Apr 2007