Jason Bridges

Phil340,W02-lecture 17 notes

Indeterminacy of reference II


Going to have to move quickly today.


Recall Quine’s argument:

1.      Systematic reference permutations are always possible.

2.      The only evidence an interpreter has for the meanings of the words of a speaker’s language is the speaker’s utterances of whole sentences.

3.      There are no facts about what a speaker’s expressions mean that are not in principle available to an interpreter.

4.      Therefore, reference is indeterminate (from 1, 2 and 3).


I. Recapitulating the idea of a proxy function

Based on some questions after class from last time, I realized it would be good just to say a bit in clarification of the notion of a proxy function.


A proxy function, to simplify, is a one-to-one mapping of every object in the universe onto another object in the universe.

Such functions exist, whether we specify them or not, in the same sense that gigantic natural numbers exist before anyone gets around to specifying them in mathematical notation.


So suppose there were only three objects in the universe: Austin, the desk, my mug.

Then we can define a proxy function such that:

The proxy of Austin is my desk

The proxy of my desk is my mug

The proxy of my mug is Austin


Now here’s the point needed to establish the availability of systematic reference permutation.

Let p be the proxy function just defined.  Let T be a theory of meaning for a language L.  Then let T* be the theory of meaning for L such that: whenever according to T a name n refers to an object x, then according to T* n refers to p(x); whenever according to T a predicate is true of an object x, then according to T* the predicate is true of p(x).

Then T and T* will assign identical truth conditions to all sentences


So suppose T involves the axioms:

1. “Austin” is a name that refers to Austin.

2. If x is a name, then x + “is a student” is true iff the object referred to by x is a student. (I.e., “is a student” is true of students.)


Together these axioms entail the T-sentence:

Austin is a student” is true iff Austin is a student.


So let’s define T* in accordance with the stipulations above.  Instead of 1 and 2, we’ll have:

1*. “Austin” is a name that refers to the proxy of Austin.

2* if x is a name, then x + “is a student” is true iff the object referred to by x is the proxy of a student.  (I.e., “is a student” is true of proxies of students.)


Together, these axioms entail:

Austin is a student” is true iff the proxy of Austin is the proxy of a student.

Same truth condition!  (And this is true no matter what the proxy of Austin is.)  But all the names refer to different things, and all the predicates take different extensions


II. Second premise

The second principle we have in effect already discussed back when we looked at Davidson on interpretation.  The point is that the only evidence we have for what words mean, for the semantic properties of subsentential expressions, is indirect, via evidence about the semantic properties of sentences.


Of course, an actual interpreter will not first assign truth conditions to all of a speaker’s sentences and then attempt to construct a theory of truth for them; he must switch back and forth between sentences and terms, using hypotheses about one to test hypotheses about the other, if he is to make any headway.  But the level of sentences and their truth conditions, if not the first to be settled in practice, is nonetheless evidentially fundamental.  As Frege emphasized, it is only through the utterance of a sentence that one can perform a full-fledged linguistic act—in Wittgensteinian terms, only sentence-utterances constitute moves in (most) language games. This is particularly clear in the language-game of greatest importance to the interpreter:  the making of assertions.  You could not begin to make sense of the speaker’s talk unless you construed her as advancing, at least some of the time, claims about the nature of the shared environment. To advance a claim, of course, is to assert something, and an assertion is an utterance of a sentence.  The general point is just that however creative the interpreter is at prompting the speaker to various sorts of linguistic behavior,  the resultant evidence for the construction of the theory of meaning will always be in the form of sentences.  No smaller meaningful unit can be experimentally isolated.


Of course, it’s true that sometimes one performs a speech act with just a word: “What’s that?”  “A chicken.”  But here the context provided by the initial question is essential.  If unprovoked you had just burst forth with, “a chicken”, you would not have performed any speech act….



This conclusion is reinforced by recalling the context principle, which, as we said, traces all the way back to Frege.  The principle is that words have meaning only in the context of sentences.  This is not to say that if I just utter a word by itself, what I utter does not have any meaning in the language.  Of course not.  The point is rather that there’s nothing for the meaning of a word to be but the role it plays in determining the meaning of a sentence.  Words have no semantic point, hence no meaning, considered in isolation; they aren’t good for anything.  What words do is join together with other words to produce expressions, namely sentences, in the utterance of which we can say things.


III.  Third premise

We have already implicitly encountered this premise as well, in Davidson’s views on interpretation.  It is implicit in Davidson’s methodology, in his imposition of the interpretive constraint on a theory of meaning.  The thought at work here is that the appropriate context to think about meaning is in thinking about how one might come to understand someone else.

This goes against a traditional view that meaning is private, in the sense that each of us can never be sure we know what other people mean by what they say, that some aspect of that meaning might always escape us.  Only the individual really knows what she means.


The problem with this view, for Quine, is not that people can’t possibly have private associations with words, that is, associate words with feelings and images that other people are not privy to.  Of course we can; nothing’s stopping us.  The claim is rather that any such material ought not, properly speaking, be thought of as part of the meaning of what is said.


Why not?  Because, says Quine, language is a social phenomenon.  He means this in a very limited sense, a sense that Strawson, Davidson and Wiggins, whatever their disagreements, would all endorse.  Language is social in that it is essentially a tool for communication.  It is essentially fit for people’s conveying things to each other, for telling each other things, asking each other things, and so forth.

It’s simply misguided to think of linguistic meaning as something that isn’t in principle available to both the speaker and her audience.  For the point of the notion of linguistic meaning is to capture what is shared in successful communication. 


It’s tempting to suppose that there might be some residue, some bit of meaning, that is not available to the audience.  But what those who advocate the move to reflecting on meaning in the communicative context would say is that whatever is not even in principle available to the audience ought not, on pain of serious confusion, be counted a part of the meaning.  Meaning just is that which is present for both parties in successful communication.


Of course, sometimes people don’t understand each other.  Sometimes one can feel, and rightly so, that no one understands one.  And of course one can make up a code that no one else can crack.

The point is not that people always in fact understand each other’s meanings; it’s that meanings is always in principle available to the audience.  If I were just to tell you my code, you’d be in a position to understand my meanings.  And if one’s feeling misunderstood persists despite all attempts to be understood (say, when one says something in a philosophy seminar), this should lead one to doubt one is saying something coherent.

Much more could be said about this.  The general acceptance of this move away from the private first-personal context to the communicative context marks one of the central developments of 20th century philosophy.  But it’s a large issue we can’t pursue at length here.  I shall assume the truth of premise 3 for the current discussion.


Note that Quine associates the denial of 3 with what he calls the museum myth of meaning.  (p. 27.)  I think the proper way to understand Quine’s problem with this, so far as premise 3 goes, is not that we can’t think of meanings as entities.  The point is that we can’t think of meanings as entities that somehow determinately have properties that could not even in principle make a difference to what happens in communication, in someone speaking and someone else understanding.  For Quine, once we appreciate that reflection on meaning is fundamentally reflection on a social phenomenon, that idea is revealed as analogous to the claim that the notion of a triangle involves some further property other than being three-side polygons—it’s just that this property has no bearing on any geometrical goings on of triangles.  That claim is ridiculous—a triangle is simply a geometrical figure.  Likewise, meaning is a social phenomenon, and any alleged meaning property that plays no possible role in what happens among people when they speak with each other is an untenable posit.

(Quote p. 27.)


So it looks like we have reason to buy all three premises.  And on the face of it, they do seem to imply the conclusion.  For if there are no facts about meaning that an interpreter could not have evidence for, and if the only evidence an interpreter could have for the meanings and references and extensions of words is the utterance of whole sentences, and any number of different assignments of references and extension are consistent with according utterances of whole sentences the same truth conditions, then….


IV.  A brief objection

I want very briefly to suggest a problem with this argument, without questioning the truth of any of these premises.

The problem, I think is this: you don’t get to the conclusion unless you add a further premise, and it this premise that is dubious.


The premise is this:

The only facts available to an interpreter about a speaker’s sentences are the conditions under which they are held true (or feared true or desired true….)


Let us look again at the beginning of radical interpretation.  A speaker says, “gavagai,” whenever a rabbit runs by, and not otherwise.  Being charitable, that is, assuming the speaker has correct beliefs about her environment, we ascribed the following truth condition to the sentence:

“Gavagai!” is true at time t at place p iff there’s a rabbit at t at p .

As we saw, T-sentences like this are compatible with all manner of reference assignments to the term, “gavagai”, and the situation does not markedly change, given the proxy functions, as we accumulate more and more T-sentences.  For if the above is true than so is:

“Gavagai!” is true at time t at place p iff there’s an undetached rabbit at t at p.


But is this really all that we’re going to think we can say about the sentence?  Aren’t we also inclined to suppose the following:

In uttering “Gavagai!” the speaker said that a rabbit is present.

We’re not inclined to suppose:

In uttering “Gavagai!” the speaker said that an undetached rabbit is present.

These are incompatible, unlike the pair above.  We have to decide between these two.  And obviously we’re quite inclined to go with the former.


So if we want a theory that satisfies convention T also to satisfy the interpretive constraint, and we think such facts as above are available in interpretation, we do have to choose between the above T-sentences.  For only in the former can the right side of the conditional be understood as expressing what is said by an utterance of the object-language sentence.

The idea is that a theory of truth-conditions can serve as a genuine theory of meaning, if such facts are indeed available, only if the T-sentences are understood in a way that goes beyond what they actually state.  All they actually state are truth-conditions.  But only some statements of truth conditions can be construed as expressing what is said by an utterance of that sentence.

So this would give us a way of deciding among the competing theories.  We wouldn’t have people saying that the proxy of….


But do we have available such facts as above in interpretation?  Well, why not?  For present purposes, the point is only that none of the premises of the argument suggest that we don’t.

I won’t pretend this is not a large issue, and I can’t get into it at length at all, but why can’t we say that such decisions are forced on us by the principle of charity, broadly understood?

Davidson enjoins us to find beliefs and desires that make sense by our lights.  Well, it makes no sense by our lights, unless we imagine some very special circumstances the presence of which we could test for, to spend all one’s time saying that undetached rabbit parts are present, rather than rabbits.  That seems nuts.  And we should go with that feeling.  Certainly, we’re inclined to doubt most people even have such artificial concepts.


V. Stroud’s diagnosis

Not too much time to talk about Stroud.  Two points.


1. One of the main themes of the article is that what’s really motivating Quine is an just an unwillingness to accept that there could be such facts like this about what a person said.  Quine portrays his argument as flowing out of the considerations outlined in the argument as I summarized it, but what’s really at stake is just a brute reluctance to accept that there could be such facts as these.  These facts involve the idea that the speaker has said something with an explicit content, a content that he also presumably believes, or at least intends us to believe.  These are mental facts, facts entailing a speaker’s capacity to understand language, to intend to say things, to believe that such-and-such.

And Quine, for Stroud, like many philosophers before and after him, has an uneasiness about the very idea of mental, thought-involving facts.  He thinks there’s no place for them in a properly scientific picture of the world.  So he is drawn to the other sort of fact, which makes no mention of a specific content, but only of an attitude toward a sentence.  We still have the attitude to make sense of, and it seems mentalistic, but the sentence is a physical object.

Much, much to be said about the frame of mind Stroud imputes to Quine.  But can’t here.  (Davidson has it to.)


Stroud sees as connected to this Quine’s talk of translation manuals rather than theories of meanings.  For translation manuals, as we saw, only relate various words to each other: they don’t say anything about what these words actually mean, or what they refer to.  And Stroud sees in Quine a desire to understand meaning, talking, wholly in terms of relations among words, to submerge talk of meaning and reference and so forth.


2. This leads to the second point, which concerns what Stroud says about Quine’s idea of how to reconcile ourselves to his result.  On the face of it, the indeterminacy of reference seems a devastating result.  It turns out that it’s never true that we’re talking about this rather than that.  Our thought never, as it were, makes contact with any objects at all.  How can we make sense of this?  Don’t we lose the idea the very idea of speech?

Quine’s response is that we can avoid this insane result by doing what he calls ‘taking the home language at face value’.  According to Quine, this involves refusing to think of one’s own language in terms of these crazy proxified theories.  What we should do is accept what he calls homophonic theories of meaning (or homophonic translation manuals, as he insists on speaking in terms of).  These are ones where the metalanguage and object language are the same, and the reference or extension of a term n is stated in terms of itself.

Thus only the first could be part of a homophonic theory:

The extension of “rabbit” is rabbits.

The extension of “rabbit” is undetached rabbit parts.

Thus we’ve resolved things.


But Stroud’s point is that, if reference is really indeterminate, this is useless.  For if the reference of the metalanguage expressions is not determinate, and according to Quine it isn’t, these sentences tell us nothing whatsoever about the extension of the term “rabbit”.  It’s still all up for grabs.