Frege's Mistake

In a recent post I told the story of Ralph, 'Hesperus' and 'Phosphorus'.  We saw there why Ralph might assert

('h≠p')    'Hesperus ≠ Phosphorus'

and how astronomical observation might lead him to change his mind. Remarkably, at no point in telling that story did we find any reason to deny that

(Bh=p)    Ralph believes Hesperus = Phosphorus.

It was not that we refused to take Ralph at his word.  Ex hypothesi, Ralph is a competent speaker of English and we took him to be sincere when he asserted ('h≠p').  We had to conclude that:

(Bh≠p)    Ralph believes Hesperus ≠ Phosphorus.

The principle at work here is an important one.  I call it (following Kripke) "The Disquotation Principle":

(DP)    If someone understands a sentence and believes that it is true then they will believe what the sentence says.

If we restrict ourselves to unambiguous, non-indexical English sentences we can express this more formally as.

(DP)    If x believes^┌p^┐ is true & x understands ^┌p^┐ .⊃ x believes p.

I think (DP) is true, indeed I'm happy to call it a necessary or, if you prefer, an analytic truth about belief and understanding.  (DP) is what makes language useful: learning what sentences others think are true or false tells us what they believe and disbelieve. So when Ralph asserts ('h≠p') I conclude that Bh≠p.

But Bh≠p doen't entail that Bh=p is false. Of course Ralph doesn't think the sentence,

(h=p)    'Hesperus = Phosphorus'

is true.  But Ralph does think that the sentence

('h=h')    'Hesperus = Hesperus'

is true.  Which means, given (DP) that we must say that:

(Bh=h) Ralph believes that Hesperus = Hesperus.

And if we say (Bh=h) is true then Leibniz’s Law requires that we say that (Bh=p) is true. After all Hesperus is Phosphorus, and by the principle of the Indiscernibility of Identicals , anything true of Hesperus must be true of Phosphorus.  It follows that if Hesperus has the property of being-believed-by-Ralph-to-be-Hesperus then so must Phosphorus.

Thus we found ourselves saying that Ralph believes inconsistent and necessarily false propositions. But that did not trouble us because we saw that it was a mistake to judge the reasonableness of Ralph's inferences or states of mind by looking at the propositions he believes.  We understood what sort of situation Ralph thinks he is in and why and how this might lead him to say the things that he does.

All this should seem remarkable because it means that we managed to make sense of Ralph without ever having to invoke what philosophers call the "narrow" or "referentially opaque" or "notional style" of belief talk.  We never had any reason to say that Ralph believes anything about Hesperus that he doesn't believe about Phosphorus. And a good thing too since that would have put us in a jam: it would have put our psychology on a collision course with Leibniz’s Law.

That is precisely the jam that Frege finds himself in at the beginning of On Sense and Reference.

People forget that OS&R begins with a psychological question: how to explain the difference in "cognitive value" between identity sentences like 'h=h' and 'h=p'.  How, Frege asks, can Ralph simultaneously think (h=p) is false and (h=h) true? Frege takes the obvious answer -- the obvious psychological explanation for these different attitudes -- to be that (Bh=p) is false even while (Bh=h) is true.  The balance of the essay (and, one might add, a good deal of twentieth century philosophy) is an attempt to explain how that can be so.

But is it so?  Is Frege's psychology sound?  Frege takes it for granted that if Ralph denies 'h=p' it must be because he doesn't believe that Hesperus = Phosphorus.  This assumes what we will call (again, following Kripke), "The Quotation Principle".

(QP)    If x understands a sentence and believes what it says, then x will believe the sentence is true.

or

(QP)    If x understands ^┌p^┐ & x believes p .⊃ x believes ^┌p^┐ is true.

It is only this assumption that forces us to count Ralph's denial of 'h=p' as proof that Bh=p is false.

I think (QP) is wrong.  I don't mean that I think there are puzzle cases or tricky counter-examples.  I mean wrong.  I don't think people ever believe that sentences are true because they believe the propositions those sentences express.  I offer this as an empirical claim, because I take (QP) to be contingently false and I take cases like Ralph's as evidence that this is so.  But the argument won't be that simple because, as we shall see, Frege assumes (QP) to be analytically true, and this assumption regiments everything he and his successors have understood about meaning and belief.

So my purpose just now is not to convince you that (QP) is wrong but to show you how much hangs upon it.

We've just observed that (QP) is essential to Frege's way of arguing for the referential opacity of belief.  Now notice that there is no other way to argue for the opacity of belief except Frege's way.  The referential opacity of the attitudes is usually cited as the hallmark of the "intentionality" of the propositional attitudes.  It is supposed to separate the intentional from the merely intensional.  But if we reject (QP) we would have no reason treat the propositional attitudes as referentially different from any intensional construction.

I'm sure that there are philosophers who will want to tell me that failures of substitutivity in the attitudes are somehow “intuitively” obvious.  That they are part of "folk psychology".  They might point out that it would sound odd or inappropriate to most folks to say that Ralph believes Hesperus is Phosphorus when Ralph is all the while insisting 'Hesperus isn't Phosphorus'. But intuitions about what it might seem odd or appropriate to say are beside the point.  For substitutivity to fail we need evidence that bh=p is literally false even when bh=h is true.  And this needs evidence. The propositional attitudes are theoretical entities.  We postulate them to explain behavior.  And however folksy our psychology may be, Ockham's razor applies: If you are going to claim that there is a difference between one belief and another you must be prepared to say what difference having one rather than the other might make in someone’s behavior.

But what difference could having a Hesperus belief rather than a Phosphorus belief make to Ralph's behavior? He doesn't behave differently towards Hesperus than he does towards Phosphorus. He can't; Hesperus is Phosphorus.  He can't look at, point to, throw rocks at, travel to, flee from or fart in the general direction of Hesperus without doing likewise to Phosphorus and vice versa.

No one can behave differently toward towards Hesperus than Phosphorus they can only behave differently to the words 'Hesperus' and 'Phosphorus'.  The only evidence for the referential opacity of the attitudes -- the only evidence there can be -- is linguistic behavior; and it is only evidence under the assumption of (QP).

Comments

[Justin]

This seems to be a pretty standard Millian line. Nathan Salmon says something similar, I think.

[Tristan Haze]

I think there is a way of saying, in a certain sense, that Ralph believes that Hesperus is F, without implying that he believes that Phosphorus is F, without flouting Liebniz's Law.

Let's stick to belief-reports involving names, for simplicity. In short, my view is that the name 'Hesperus' in a belief report like:

(A) 'Ralph believes that Hesperus is F'

can be read as doing two things at once. (1) specifying the object of Ralph's belief, and (2) specifying the concept (or mode of presentation) via which he has it. On such a reading, (1) could be expanded to:

(B) Ralph believes, of Hesperus, via his Hesperus-concept, that it is F.

(A similar thing could be done for the 'F', but I'll keep it simple.) Some belief reports, on the other hand - purely de re belief-reports - may be read as only specifying the object. (A) read this way could be expanded to:

(C) Ralph believes, of Hesperus, via *some* concept(s), that it is F.

This analysis, I think, can shed light on Kent Bach's puzzle about belief (http://userwww.sfsu.edu/~kbach/puzzle.html), by making it clear how 'S believes that o is F' can fail to entail 'S believes the proposition that o is F'. The first report can be read in the (C) way, whereas the second report induces a (B) type reading.

Likewise Kripke's puzzle: in a (C)-sense, Pierre can believe that London is beautiful and also that London is not beautiful - once via one concept, once via another.

This treatment admittedly makes belief-reports quite ambiguous and flexible devices, but I think that's to be expected. Furthermore, it seems to make more sense out of the way people actually talk, than your resolutely Millian approach. What do you think?

[tomkow]

Tristan,

Thanks for your comment!

I don't believe in "concepts", "senses", "characters" or any other sort of mediating mental representation . I think that the only argument for them is Frege's and I think that turns on a mistake.

The best way to understand what philosophers mean when they speak of the opaque/de dicto/narrow/intentional sense of:

(1) Ralph believes p.

is

(2) Ralph is in the state of mind that would lead a competent English speaker to believe the sentence ┌p┐ is true.

Which is why they get puzzled about how to ascribe beliefs to foreigners, animals, or less than competent speakers and indexicals… &c.

And when they find that folk's use of "belief" does not fit with their (philosophical theory driven) "intuitions" about how to ascribe beliefs they start sneering at "folk psychology".

I don't think (2) names a natural kind of mental state because I think it is contingently false that speakers believe sentences true because they believe the propositions those sentences express. So I think the sense of 'belief' you are hoping to account for with an appeal to concepts is a chimera born of this mistaken assumption

To see how you can handle Hesperus and Phosphorus without appeal to concepts, senses or any other sort of mental representation I invite you to have a look at Now, Me .

I would be very interested to hear what you think.

[Tristan Haze]

I've taken one look at your proposal, a few days ago, and thought a bit about it. I will have to have another look, but for now here's a preliminary response.

You say you don't believe in concepts, and that you think the only argument for them is Frege's. Arguments aside, I hardly think it's right to regard Frege's considerations about identity statements as being the only thing which might make us want to talk about concepts. I think concept-talk - or the concept of a concept, to put it provocatively - is much more basic than that. They're not just some posit which theorists invoke to solve puzzles. Pretty much all thoughtful, educated people talk and think about concepts.

Relatedly, in your 'Now, Me' paper, you indicate that one of the chief things in favour of your account is that this account doesn't require one to talk about concepts, senses, modes of presentation, or whatever. But I don't see that as particularly desirable. (To butcher a phrase of Lewis's: if we're going to have to posit concepts anyway, we might as well enjoy them.)

If you're going to try to do without concepts and the like entirely, I think you're going to have a very hard time in many areas. One particular challenge, close to my concerns, is: how are you going to make sense of modality? (cf. http://sprachlogik.blogspot.com/2011/06/sketch-of-way-of-thinking-about.html)

There are pretty basic criticisms, so I'm not sure how conducive they are to fruitful debate, but let's see. Again, I will have another look at your specific proposal and may have some further comments after that.

[tomkow]

Tristan,

As you say, lots of philosophers talk about concepts these days. I know. It has taken a whole generation of philosophers for them to come back in style.

They went out of style for a while thanks to Quine, they have crept back as Quine's lessons have been forgotten. It has been a deliberate amnesia: it was just too hard to do philosophy in a post-Quinean regime, so people started to pretend he never existed.

So never mind my account. If you want to talk about "concepts " you need to answer Quine's objections.

Start with the problem set by Two Dogmas of Empiricism: If there are concepts then there must be a difference between people who share the same concepts but disagree about the empirical facts and people who agree about the empirical facts but have different concepts. How do we tell them apart?

[Tristan Haze]

Regarding this problem: first, and in general, I don't think it's generally true that a distinction is only defensible when one can give a method for applying it in all cases. So the lack of a general and informative answer to your final question ('How do we tell them apart?') needn't be perceived as a worry.

There will be difficult cases, and perhaps cases where the distinction between conceptual and non-conceptual "disagreement" is not helpful. But there will also be cases where it helps a good deal, for example where people who thought they had a factual disagreement come to agree that they are simply using different concepts.

But there's another issue here: the distinctions between apriori/empirical and necessary/contingent are being run together, and the analytic/synthetic distinction's in the mix somewhere too.

This is intolerable if, like me, you agree with Kripke that there are aposteriori (empirical) necessities, and furthermore think that these are conceptual truths in a certain sense. On such a view, one can disagree about 'empirical facts' by having different (or differently arranged) concepts. (This hangs together with semantic externalism: which conceptual arrangement is best depends on the identity and nature of the objects of the concepts, and this is not determined by the concepts themselves.)

I mention this in case it looked like I was committed to a view which sorts claims (propositions) and disagreements into non-overlapping categories 'conceptual'/'grammatical' and 'empirical'. What has made the philosophy of modality so difficult, in my view, is that we've got to get over this old-fashioned and over-simple way of looking at things, but without losing our grips entirely. We still need to talk about concepts, and make distinctions like conceptual/non-conceptual, apriori/empirical.

I guess I see Quine as responding to difficulties by rubbing out all of these distinctions as though they are essentially just one bad idea, in contrast to Kripke, who tackles the difficulties by separating out different distinctions in the area. And I'm much more sympathetic with this latter approach.