Here is an old problem for semantic eternalism which still fascinates me. This is Kusumoto's (1999) version of it. According to the eternalist, 'believe' operates on eternal propositions, that is, propositions that are timelessly true or false. So, an occurrence of 'John believes that Mary is pregnant' is true iff John believes at t* that Mary is pregnant at t*, where t* is the time of speech. But consider now the following case.
Mary is pregnant on December 24, 2006 and is expected to give birth on January 15, 2007. But on the morning of December 24, 2006, John and Mary are in a car accident. Mary and the baby are fine. But John is in a coma. Exactly four months later John wakes up and remembers the accident but he believes it is still December 24, 2006, so he says: 'Where is Mary? She is pregnant'. One can truthfully report what John believes using the sentence: 'John believes that Mary is pregnant'. But according to the eternalist, an occurrence of 'John believes that Mary is pregnant' is true iff John believes on April 24, 2007 that Mary is pregnant on April 24, 2007. The problem with this analysis, however, is that John couldn't possibly believe that Mary is pregnant on April 24, 2007. For he believes that it is still December 24, 2006. Eternalism would thus seem to get the truth-conditions wrong.
Here is another way of making the same point. Suppose we treat 'John believes that' as a modal operator. 'John believes that p' is then true iff for all worlds w compatible with what John believes at (@, t), p is true at (w, t). But John believes (among other things) that no human pregnancy lasts 13 months and that Mary was almost 9 months pregnant on December 24, 2006. So, if eternalism is right, then it is true at any relevant world of evaluation that no human pregnancy takes 13 months, that Mary was almost 9 months pregnant on December 24, 2006, and that Mary is pregnant on April 24, 2007. This is inconsistent. So, contrary to appearances, John's belief set is inconsistent!
Reference:
Kusumoto, K. 1999. Tense in Embedded Contexts, Doctoral Dissertation, University of Massachusetts, Amherst.
Friday, September 08, 2006
An Old Problem for Semantic Eternalism
Subscribe to:
Post Comments (Atom)
17 comments:
Is eternalism really committed to the view that for any expression 'E' that denotes t*, 'John believes at E that Mary is pregnant at E' is true? That seems to be what the arguments against eternalism rely on (else eternalism is not committed to the claim that on April 24, 2007, John believes that Mary is pregnant on April 24, 2007). Are these arguments as suspect as any arguments that rely on the free substitution of coreferential terms within propositional attitude contexts?
Hi Leo,
There may well be versions of eternalism that can avoid this sort of objection. But Fregean eternalists take sentences without explicit time determination to make implicit reference to the time of speech (see e.g. Richard 1981). And they take 'S believes that S is F' to express the proposition that S believes at t* that S is F at t* (where t* is the time of speech).
Reference:
Richard, M., 1981, 'Temporalism and Eternalism', Philosophical Studies 39, 1-13.
Hi Brit,
I'm sure it's my fault for being too terse, but I don't think you quite got what I was worrying about. I wasn't quibbling with your characterization of eternalism -- what was worrying me was the move from (1) to (2):
(1) At t* John believes that Mary is pregnant at t* (where t* is the time of utterance).
(2) On April 24th, 2007, John believes that Mary is pregnant on April 24th, 2007.
I understand that eternalism is committed to (1), but my worry was whether it's kosher to move from (1) to (2). (2) seems to present John as conceiving of the time of utterance in a certain way, namely as April 24th, 2007. And the case as described makes it clear that John doesn't conceive of the time of utterance in that way.
Let me put the worry another way. If these sorts of arguments work against eternalism, similar arguments ought to work against the standard view that the semantic content of 'now' in a context is the time of utterance. So suppose John says, 'Mary is pregant now'. Then John believes that Mary is pregnant at t* (the time of utterance). So John believes that Mary is pregnant on April 24th, 2007. Therefore John's belief set is inconsistent. A defender of the view that 'now' refers to the time of utterance could respond that it's not okay to replace t* with 'April 24th, 2007', perhaps because the semantic value of 'April 24th, 2007' is not itself just a time. This reply seems legitimate -- why isn't a similar reply available to eternalists?
Hi Leo,
I see what you're saying.
But "At t* John believes that Mary is pregnant at t* (where t* is the time of utterance)" is just a schema. So, if eternalists think that this is good paraphrase of the content of "John believes that Mary is pregnant", then they are committed to the move from (1) to (2). (2) is just a substitution instance.
I take it that you're questioning whether the "time of speech" should always be the time of the speaker's context. Well, maybe it shouldn't, but Frege and Richard (1981) argue that it should. As far as they are concerned, John's conception of time is irrelevant. What matters is the time at which he is speaking.
I agree that there are related arguments involving "now". But these cases go against eternalism as well. A standard solution to the problems of temporal indexicals is Lewis'. On Lewis' account, when John says "Mary is pregnant now", he self-ascribes the property of being located at a world and time at which the temporal proposition that Mary is pregnant is true. Since Lewis' solution makes appeal to temporal propositions, it is inconsistent with eternalism (as defended by Frege and Richard)
The reply you are suggesting may well be legitimate and may well be available to eternalists. But it is not a reply that any eternalist has defended in print (as far as I know).
Thanks, Brit; that helped.
I think I've found a way to make my problem a bit more concrete ... I hope you'll bear with me a bit longer.
You say eternalists are committed to (1), and are therefore committed to the truth of every substitution instance of (1) where 't*' is replaced by an expression referring to the time of utterance. 'July 24th, 2007' is an expression that refers to the time of utterance, so eternalists are committed to (2). It seems to me that if that's what eternalists are committed to, then Frege was not an eternalist. After all, he thought that expressions refer to their ordinary senses in belief contexts. So, on Frege's view, the first occurence of 'July 24th, 2007' in (2) and the second occurence of 'July 24th, 2007' in (2) have different referents -- the first refers to a time, the second to a sense. For Frege, a belief ascription is true only if the sense that the 'that' clause refers to adequately characterizes the way the subject of the ascription conceives of things. (2) fails that test, so Frege would deny (2). So Frege is not an eternalist. What do you think?
Hi Leo,
Well, all it takes to be a temporalist is one damn temporal proposition.
Frege thought that there was none. So he is an eternalist.
You're right, one has to be careful when speaking of Frege and belief contexts, for the reasons you point out. But he is an eternalist.
Richard (1981) is the only eternalist that I know of, who has officially defended the view that schema (1) is a good paraphrase of what is expressed by 'John believes Mary is pregnant'. But Salmon (1989) and King (2003) come close.
The problem with the line you want to take is that if you do not defend some version of Richard's (1981) view, then that might get you into the sort of trouble outlined in Richard (1981).
Brit,
Well, I think the time has come for me to do less talking and more reading. (Maybe it was a few comments ago?) Thanks for all the helpful references. I am now officially interested in this topic.
I think the problem here has to do with the notion of context of utterance. Kusumoto's example involves a very special context --- one in which the time of the context is not transparent to the speaker. In such cases, perhaps, one might stipulate that it is, so to say, the time of the speaker that counts.
Think about the film 'Goodbye Lenin' (2003). The protagonist, a convinced communist woman, wakes up from a coma in Berlin, in the post 1989 era. Her heart being very weak, her sons let her believe that communism still rules in Eastern Europe. The film is funny and very enjoyable. Anyway. In that case, I think, there's not a problem with time, but there's a problem with indexicals such as 'the (actual) government', 'the (actual) president' etc. As uttered by the woman, the semantic values of those expression should be determined by the post 1989 context of utterance. But of course the woman is referring to the pre 1989 government and to the pre 1989 president. People understand the woman because they know hers context is, so to say, pre 1989. Thus, I think Kusumoto's example shouldn't be used as an argument against temporalism. Otherwise, it would prove too much, namely that, for instance, 'I an hungry', as uttered by me and as uttered by Berit express the same proposition. Are temporalists committed to this latter view?
Hi Julien
Well, we cannot just stipulate it. We need a theory that predicts it.
I saw the film you mention. It's great. I think sentences containing 'the (actual) president' will be false under these circumstances.
Actually Kusumoto's argument is not used against temporalism. It is used against one form of eternalism.
No, temporalists are NOT committed to 'I am hungry' expressing the same proposition in all contextss
Yes, I was meaning eternalism --- not temporalism.
Ok, so sentences such as 'The actual president is a member of the Communist Party' would false, as uttered by the protagonist of Goodbye Lenin in the post 1989 scenario. The analogy was probably misleading.
What I wanted to say is that, for a speaker unaware of John's accident, the eternalist's truth-conditions for 'John believes that Mary is pregnant' may seem correct after all. In another sense, you're right: they are incorrect.
My point was that, perhaps, the beliefs of the speaker may be contextually relevant in the determination of the content/proposition expressed. Eros Corazza in his talk at the Sopha triennal meeting in Aix-en-Provence (a very nice conference, by the way) was arguing that it might be useful to add some point-of-view parameter to the traditional ones, in order to determine the content/proposition expressed in a given context. The time believed by the speaker my be one such perspectival parameter.
would *be* false, sorry
Julien: Right, I am willing to grant this. The example is only meant to undermine eternalism, as this view is normally construed.
The problem with your and Leo's suggestion is this: if the value of the temporal variable can be fixed relative to a perspective, then Richard's (1981) argument against temporalism becomes an argument against eternalism as well, which means that eternalists cannot appeal to Richard (1981) when they argue against temporalism.
Wait a moment.
But isn't the temporalist committed to the view that propositions do not have fixed temporal parameters *at all* - no matters whether they can be fixed relative to a perspective or not?
If I remember correclty, Richard's argument, or some version of it, was as follows:
1. Yesterday Julia believed it was raining
2. Today Julia still believes what she believed yesterday
3. Today Julia believes it is raining.
The argument seems invalid, but the temporalist should accept it as valid.
I presently can't see how the temporalist could block the argument without abandoning her metaphysical tenet. Perhaps you could help me out here.
More generally, however, the temporalist should provide a non ad hoc argument to the effect that temporal variables should be fixed relative to a perspective not only in the clearly exceptional coma cases (if correct at all, one of my points was that a speaker unaware of the accident would attribute *eternalist* truth-conditions to 'John believes that Mary is pregnant'), but also in ordinary cases such as the Julia example above.
Or am I missing something?
Shit, I guess I've lost my post... Very quickly, I was reconsidering Richard's 1981 argument. How temporal variables fixed relative to a perspective would invalidate
1. Yesterday Julia believed it was raining
2. Today Julia still believes what she believed yesterday
3. Today Julia believes it is raining
?
More importantly, even if they could, the eternalist should explain why, in *ordinary* cases such as 1-3 temporal variables should receive a perspectival treatment. A perspectival treatment might be justified in *extra*ordinary cases, such as the coma one. After all, as I said before, the speaker unaware of John's accident would attribute *eternalist* non-perspectival truth-conditions to 'John believes that his wife is pregnant'.
Finally, isn't the temporalist committed to the view that propositions lack any temporal, B-type, determination? But then non-perspectival times should not figure in temporalist propositions. If so, how could the temporalist use them in order to block Richard's argument?
Remember, temporalism just says that there is a temporal proposition. Eternalists deny this.
Traditionally, of course, temporalists have made a claim about sentences as well (but this is not essential to temporalism). The claim about sentences is that sentences without explicit time determination do not make reference to a time.
The point I was making about fixing values relative to perspectives was this: Richard has an argument against temporalism. If the eternalist fixes values relative to perspectives, the argument can be turned against her position as well.
So, let us look at Richard's argument.
Mary believed Nixon was president
Mary believes everything she once believed.
So, Mary believes Nixon is president.
This poses a challenge to the temporalist (but as I argue in my book, not an insuperable one). The eternalist makes the right predictions.
But now, look what happens if you allow the values of time variables to vary with a perspective.
On Dec 24 John believed that Mary was pregnant
John believes everything he believed on Dec 24
So, John believes that Mary is pregnant.
The argument shouldn't be valid (we just agreed on that). But if you allow the time variables to be relative to speaker perspectives or whatever, then it has a valid reading.
That was my point.
1. On Dec 24 John believed that Mary was pregnant on December 24
2. John believes everything he believed on Dec 24
3. So, John believes that Mary was pregnant on December 24.
If this is the valid reading, then I would say that the eternalist should welcome it.
Anyway, it is true that Richard's argument is often presented as *the* definitive objection against temporalism (See Kuenne 2003, for instance). But maybe it is just an objection against *one form* of temporalism.
Concerning your definition of temporalism, I guess MacFarlane's claim that future contingents are assessment sensitive would count as a form of temporalism.
The view according to which future contingents acquire a determinate truth-value at the time to which their truth-conditions refer, also, would count as a temporalist view.
Is that correct?
If so, then I think we might need some more fine-grained labels (relative to a context of assessment, truth-values are eternal, and once future contingents acquire a truth-value, they get it eternally, on the view sketched above).
Hi Julien,
The argument you are presenting is unproblematic (except the conclusion is ambiguous). Temporalists and eternalists can agree that it is valid. So I am not sure what your point is. My earlier objection still stands.
Temporal propositions are propositions that can have different truth-values at different times. But relativists need not include times in their indices. So they need not commmit themselves to temporal propositions.
Post a Comment