Same-saying
Somewhat related to the Most Certain Principle is the following constraint on semantic content:
Same-Saying Constraint: if A utters a sentence S1, and B utters a sentence S2, then they say the same thing iff S1 and S2 have the same content.
"Saying the same thing" is here obviously not meant as "saying something with the same content". That would make the constraint empty. Rather, it's supposed to be an intuitive, pre-theoretic notion.
I think it is a bad idea to impose this constraint on semantic content. I'll give a few reasons for that first, and then make a tentative proposal for what else -- instead of sameness of content -- makes for A and B saying the same thing.
One problem with the constraint is that at least my pre-theoretic concept of same-saying is too flexible. In many contexts, it is perfectly okay to say
1) when asked about the war, all candidates said the same thing: that they rejected it from the beginning and would withdraw troops as soon as possible.
In other contexts, however, A and B can count as saying different things when both utter "I rejected the war": A talks about A, B talks about B. The semantic content of a sentence in a content can hardly be that flexible
Second, same-saying constructions can be more hyper-intensional than is plausible for semantic content. Suppose there are fines for uttering certain words on television. (Makes no sense, yes, but just suppose.) And suppose A was fined for something he said. Then one could say
2) it's not about A. If B had said the same thing, he would also have been fined,
even though B might not have been fined if he had said something synonymous to what A said but using different words.
Third, in judgements about same-saying we often ignore parts of the content of an utterance:
3) A: It's cold in here. B: Yes, it's cold in your office.
Surely the content of B's sentence has something to do with being in an office, and the content of A's sentence does not.
Fourth, our judgements about same-saying depend not only on literal semantic content but also on things that should better count as pragmatic:
4) A: Fred tried to open the jar. B: Fred had difficulties opening the jar.
5) A: That was a really great performance. [ironic] B: That was an aweful performance.
One could argue that in all those cases, we are speaking sloppily when we say that A and B do or do not say the same thing. There is, one might argue, an intuitive notion of strict same-saying that's tied to things like agreeing and giving the same information. The presidential candidates in (1) don't all give us the same information; each utterance of "I rejected the war" tells us something new, something we didn't know already from the corresponding utterance of the previous candidate. Moreover, the second candidate might well disagree with the first candidate's utterance of "I rejected the war". As for (2), that is simply false when same-saying is understood in the sense tied to information. In (3), only B's utterance gives us information about A being in an office. And in (4) and (5), the difference in the informational content can be tested by the cancelation test, which most people quickly understand.
-- OK. I agree that there is an intuitive notion of same-saying somehow linked to giving the same information. But I still think that there are equally established uses of "saying the same thing" that go with something like sameness of character or diagonal intension, as in (1), or with sameness of uttered words -- as in (2). So the current proposal is to replace the original Same-Saying Constraint by a Strictly Same-Saying Constraint, where "strictly same-saying" is explicated by appeals to new information, agreement, etc.
Kaplanian character or diagonal or primary intension doesn't pass that Same-Saying Constraint. While this is widely accepted (and often employed as an argument against taking those things as semantic values), it is less known that Kaplanian content aka horizontal aka secondary intension also doesn't pass the test. Here are 10 counterexamples. In all the following cases, the secondary intensions of the uttered sentences are the same, but it is not obvious (to put it mildly) that what is said is the same in the relevant sense: that the second utterance simply expresses agreement with the first (try prefixing it with "I agree" or "as you said"), and that the two give exactly same information.
6) A: I am Fred. B: Fred is Fred.
7) A: That is Frank. [pointing at Frank] B: Frank is Frank.
8) A and B think it is 11:00 even though it is actually noon. A: The meeting starts now. B: The meeting starts at noon.
9) Let the two objectively omniscient Gods in Lewis's "Attitides de dicto and de se" be called A and B. A is the one that throws manna, B is the other one. A: A lives on the tallest Mountain, B on the coldest. B: You live on the tallest Mountain, I live on the coldest.
10) A: That is visible in the evening sky. [pointing at Hesperus in the morning] B: That is visible in the evening sky. [pointing at Phosphorus in the evening]
11) A: Hesperus is Phosphorus. B: Hesperus is Hesperus.
12) A: hummingbirds can fly backwards. B: if hummingbirds can actually fly backwards, then hummingbirds can fly backwards.
13) A [in world alpha]: hummingbirds can fly backwards. B [in world beta]: hummingbirds can fly backwards.
There are also converse cases where people say the same thing in the relevant sense (they agree, they give the same information), even though the secondary intensions of the uttered sentences are very different:
14) A: it's raining. B: it's actually raining.
15) A: let us use 'Julius' to refer to whoever invented the zip. Julius was American. B: the inventor of the zip was American.
So neither primary nor secondary intensions satisfy the Same-Saying Constraint. Nor do Russellian propositions or any other familiar type of semantic value. Now we could try to postulate a completely new type of semantic value. But I think a better strategy is to use what we've already got and try to derive facts about same-saying from them. Here is an idea of how that could look like.
Under what conditions do two utterances give us the same information? Well, when the second one tells us nothing we didn't already know from the first one. When A says "hummingbirds can fly backwards" and B says (in the same world, at roughly the same time) "hummingbirds can fly backwards", then all the possibilities B's utterance rules out are already ruled out by A's utterance.
The relevant possibilities and the relevant knowledge are not always just possible worlds. More often than not, what we do or do not learn is centered information. When we hear "it's noon" at noon, we don't learn the same as when we hear "it's noon at noon". What we learn from the former is something about us, not about the world. Lewis's two Gods know all facts about the world, but they would still learn something new from "it is noon" or "you are on the tallest mountain".
What we learn when we hear "it is noon" is not strictly speaking information about ourselves. In the first place, we learn something about the utterance situation (that it was noon then), from which we can usually infer something about our own situation. When we're on the phone with somebody who tells us "it's raining", we first of all only learn that it's raining wherever the other person is. If we happen to know that they are, say, in Kuala Lumpur, we can infer that it's raining in Kuala Lumpur. When we know that they are 2 km South of us, we can infer that it's raining 2 km South of us (even if we don't know where we are, except relative to one another).
So when A utters something, we first of all learn something about A's situation. And when B utters something, we learn something about B's situation. How then could we learn nothing new from B's utterance?
Well, A and B might have uttered the same eternal sentences, say, "it's raining in Kuala Lumpur on May 10 at 22:25". We learn from A's utterance that his situation is in a world where it's raining in Kuala Lumpur on May 10 at 22:25, and we learn from B's utterance that B's situation is in such a world. Given our knowledge that the two utterances take place in the same world, we learn nothing new.
Notice the restriction: given our knowledge that the two utterances take place in the same world. Without that, we wouldn't learn the same thing. If A is a trustworthy speaker in world alpha and B in world beta, what we can infer from B's utterance is not the same as what we can infer from A's utterance. A's utterance is evidence for rain in alpha, B's is evidence for rain in beta.
Now suppose A and B both utter some non-eternal sentence: "it is noon". And suppose we know that their utterances take place at the same time. Given that knowledge, we again learn nothing new from B.
Or suppose we know that A is presently 2 kilometers North of B, and now they say
16) A: it's raining 1 km South of here. B: it's raining 1 km North of here.
Given our background knowledge, we learn nothing new from B.
The general notion here is that of same-saying relative to a body of information:
A and B say the same thing relative to a body of information X iff given X, A's utterance is true iff B's is.
The 'iff' here is meant to be fairly strong, as something like a priori or analytic equivalence: in any conceivable X-world, A's utterance and B's utterance have the same truth value.
In fact, as mentioned above, X can be centered information. The information we acquire from an utterance is in the first place centered information about the utterance situation, and derivatively information about our own situation. (Centered information about other things than oneself is unfamiliar, but actually quite intuitive: just as we can locate ourselves in centered logical space, within the class of time-slices of possible individuals, so we can locate other things in this space.)
Suppose, to simplify matters, that we're in A's position. When now B says something and we trust him, we can first of all infer something about his situation. Ignoring implicatures and the like, the information we acquire is that his situation verifies the sentence he uttered. To draw conclusions about ourselves, we need information that links our situation to B's situation: that he is located in the same world, at the same time, or something like this.
Example. B says "it's cold in here". This tells us various things, e.g. that somebody is uttering words. Understanding and trusting those words, we can also infer that the speaker's situation is one of those where it is cold. (This is the primary or diagonal intension of the uttered sentence.) If we moreover know that the speaker is in the same room as us, we can infer that what he says is true iff it is actually cold in this room. Holding fixed the fact that we're in the same room, we could have expressed the same by saying "it's cold in here": whenever A and B are in the same room, then B's "it's cold in here" is true iff A's "it's cold in here" is true. Relative to the information that we're in the same room, we say the same thing. If, on the other hand, we know that the speaker is in the room next door, we'd have to say something like "it's cold in the room next door" to say the same thing: holding this relation between our contexts fixed, our utterances will always have the same truth value.
What if we're not in A's position, or in B's, but somewhere else? Then the information we get from (the semantic content of) A's utterance depends on what we know about the relation between our situation and A's situation, and the information we get from B's utterance depends on what we know about the relation between our situation and B's situation. However, at least in ordinary cases, what matters for whether or not we get different information from the two is determined purely by what we know about the relation between A's situation and B's. For instance, if in (16) we know that we're 1 km South of A and 1 km North of B, then their utterances tell us the same thing. But if we know that we're 1 km South of A and 1 km North of B, then we also know that A is 2 km North of B, and that is enough to figure out that A's utterance and B's utterance give us the same information, no matter where we are located relative to them. What we learn depends on our own location: that it's raining here, that it's raining somewhere in our world, etc. But that we learn the same from both is fixed once we know about their location relative to one another.
So we can think of the relevant information X as always being centered information about A's situation and B's situation and their relation to one another.
So far, I've only talked about same-saying relative to a body of information. What about same-saying simpliciter? There are a number of options: it could be a matter of same-saying relative to no information (equivalently, relative to every information), or it could be same-saying relative to some privileged kind of information. If the latter, the privileged kind could be invariant or context-dependent. Let's look at some more examples.
17) A: I am tired. B: you are tired. [to A]
A and B are saying the same thing here. So the information that B refers to A with "you" seems privileged. In general, one might suspect that information about the reference of indexicals is always privileged. And one might extend this to proper names:
18) A: Hesperus is a planet. B: Phosphorus is a planet.
Again, at least in many contexts it would be okay to say that A and B said the same thing. One could even argue that this is true in all (actual) contexts, even though people can reasonably believe that it is false, if they don't know that Hesperus is Phosphorus. Privileged information need not be known. After all, we can sometimes discover that two people said different things, or that they did not. When we realize that we are in different towns talking to each other on the phone, we realize that we didn't actually disagree when I said "it's raining" and you said "it's not raining". -- Incidentally, this last example appears to show that information about the time and place (and world) of the utterances is also privileged, even without explicit "here" and "now" indexicals.
But now we run into many of the problems I mentioned for secondary intensions, especially cases (6)--(11). What's more, same-saying judgements about (18) seem just as acceptable to me (in a suitable context) as same-saying judgements about (19), assuming last week Fred was talking about Frank Jackson:
19) A: Fred met the philosopher he talked about last week. B: Fred met Frank Jackson.
So should we extend the privileged information to include the reference of definite descriptions? That will give us lots of even more counter-intutive results. Consider this slight variation:
20) A: Fred met his favourite philosopher. B: Fred met Frank Jackson.
Assuming Frank Jackson is Fred's favourite philosopher, it is still far from clear that A and B say the same thing.
I think it is better to let the relevant information X vary with context. In a context where we take it for granted that Frank Jackson is Fred's favourite philosopher, it is okay to describe (20) as a case of same-saying. Not so in a context where this is unobvious or surprising. Similarly, (19) is not correctly described as a case of same-saying when doubts are raised as to which philosopher Fred talked about last week (some say it was Frank Jackson, others say it was Judith Butler). With respect to (18), suppose surprising evidence emerges suggesting that Hesperus is Venus but Phosphorus is actually an alien spaceship. In this context, the exchange (18) takes place between A, a defender of the spaceship hypothesis, and B, a defender of the orthodox view. Shortly after (18), somebody says "A and B said the same thing". This seems clearly false to me, even if it later turns out that the evidence was misleading and Hesperus is really Phosphorus. (In all those cases, it seems to matter at least to a certain extent that the doubts are known to A and B themselves.)
Cases where there is doubt or confusion about there relative reference of "I" and "you" are harder to find. But they do arise from time to time.
21) A is talking on the phone to an anonymous caller. The caller is actually A's boyfriend B, who is simultaneously chatting to A via IM. A doesn't suspect B to be the caller, but wants to break up with him anyway, for different reasons. A: I don't want you to call me any more. [on the phone] A: I don't want you to call me any more. [on IM]
Does A here say the same thing twice over?
I don't think it's worth arguing over which information exactly is relevant for a given case. As same-saying judgements are vague and context-dependent, it's good to leave some vagueness and context-dependence in the analysis.
So here's the proposal:
By uttering S1 and S2 (respectively), A and B say the same thing iff holding fixed salient information X about A's and B's contexts and their relation to one another, A's utterance is true iff B's utterance is true,
where the salient information X usually includes information about co-reference of indexicals, relative spatiotemporal and modal location and things that are taken for granted by all parties.
A nice thing about this is that it can be used to explain the sloppy uses of "same-saying" with which I began. Remember
3) A: It's cold in here. B: Yes, it's cold in your office.
If it is taken for granted that the addressee of B's utterance is the speaker of A's, and that this person is in an office, then A and B really say the same thing. If it is not taken for granted that A is in an office, they don't say the same thing. That's exactly what we want. And it is not surprising that implicatures enter into same-saying judgements: we generally take for granted that speakers follow the cooperative principle; holding fixed that A and B do so, A's utterance will be true iff B's is (also again ignoring some outlandish possibilities):
4) A: Fred tried to open the jar. B: Fred had difficulties opening the jar.