In his pioneering work Arthur Burks [1949], distinguishes the following aspects of an utterance containing indexicals:
Suppose, for example, Burks tells me, pointing to a house on
Monroe Street in Ann Arbor: ``I live in that house.'' (i)
The sign itself is the token or burst of sound that Burks
utters; it is a token of an English sentence of a certain
type, namely, ``I live in that house", and it occurs at a
certain spatio-temporal location. (ii) This token has
``existential relations'' to other objects. That is,
there is a person who uttered it (Burks), there
is a house at which that person was pointing at the time
of utterance, and so forth. (iii) English associates a
meaning with the type, the same for every token of it. Any
token of ``I live in that house,'' will be true if the
speaker of that token lives in the house he or she points to
at the time they produce the token. This is what all tokens
of the type have in common. (iv) Each token also has an indexical
meaning, which results from the combination of the type
meaning and the the particular token. Call the token Burks
produced 
. Imagine David Kaplan pointing to a house
in Pacific Palasaides at some other time and producing a
token 
. and have the same type meaning, but
different indexical meanings. will be true if
the house Burks points to is the one he lives in, will be true if the house Kaplan points at
is the one he lives in.
Aspect (v) is the information conveyed
by the sign. Let's add a third token to our example. Let 
be my token of ``You live in that
house,'' said to Burks, pointing to the house on Monroe
Street. My token doesn't have the same symbolic meaning
or the same indexical meaning as , Burks' token of
``I live in that house''. But there is something
important that my token and Burks' have in common. Each
of them will be true if a certain person, Burks, lives in
a certain house, the one on Monroe Street. Once we factor
in the contextual or ``existential'' facts that are
relevant to each token, they have the same
truth-conditions. Their truth places the same conditions
on the same objects. Burks calls this ``conveying the
same information''.
The reflexive-referential theory that I advocate builds on Burks' basic framework. In sections 3.2-3.6, I go through the five aspects, usually starting with a discussion of Burks' basic idea. I discuss various issues, elaborating and qualifying the basic idea; the reflexive-referential theory is the account that emerges from this process. Here is an overview, highlighting the differences in terminology:
Aspect (i): Burks takes the signs itself to be the token. I think there is an ambiguity in ``token''; it is sometimes used for the act, and sometimes for something produced by or at least used in the act. I'll use ``utterance'' for the first and reserve ``token'' for the second. In some kinds of discourse tokens are epistemically basic, but utterances are always semantically basic. (As I use the term ``utterance'' it does not have the implication of speech as opposed to writing.)
Aspect (ii): What Burks calls the ``existential relations'' is now usually referred to as the ``context''; indexicals are expressions whose designation shifts from context to context. I will distinguish several different uses we make of context, and distinguish various contextual factors that are relevant to different types of indexicals.
Aspect (iii): For Burks the ``type meaning'' is associated by language with expressions. I simply call this ``meaning''--I try to always use ``meaning'' for what the conventions of language associate with types. The key idea here in our account of the meaning of indexicals comes from Reichenbach, who emphasized the reflexivity of indexicals.
Aspects (iv) and (v): I take content to be a property of
specific utterances. Burks recognizes two kinds of content,
while I recognize (at least) three. What Burks calls
``indexical meaning'' I call ``content given facts about
meaning'' or ``content 
''. What Burks calls
``information conveyed'' I call ``content given facts about
designation'' or ``content 
''. I claim that neither of
these is our official, intuitive notion of content; that is,
neither corresponds to ``what is said'' by an utterance.
That role is played by ``content given facts about
context,'' or ``content 
''. All three kinds of
content, however, play important roles in the epistemology
of language.
For Burks, the sign itself is simply the token. But the term ``token'' is used in two ways in the literature. Sometimes it is used for the act of speaking, writing, or otherwise using language. At other times, it is used for an object that is produced by, or at least used in, such an act. Reichenbach, for example, says that tokens are acts of speaking, but then talks about the position of a token on a page.
I use ``utterance'' for the first sense. Utterances are intentional acts. The term ``utterance'' often connotes spoken language, but as I use it an utterance may involve speech, writing, typing, gestures or any other sort of linguistic activity.
I use ``token'' in the second sense, in the way Reichenbach used it when he said that a certain token was to be found on a certain page of a certain copy of a book. Tokens, in this sense, are traces left by utterances. They can be perceived when the utterances cannot, and be can used as evidence for them. Modern technology allows for their reproduction. The paradigm tokens are the ink marks produced in writing or typing. When we read, tokens are epistemically basic, and the utterances that produced them hardly thought of. But the utterances are semantically basic; it is from the intentional acts of speakers and writers that the content derives.
An utterance may involve a token, but not be the act of producing it. My wife Frenchie and I were once Resident Fellows in a dormitory at Stanford, eating with the students each evening in the cafeteria. If she went to dinner before I returned, she would write on a small blackboard on the counter, ``I have gone to the cafeteria," and set it on the table near the front door of our apartment. I would put it back on the counter. There was no need for her to write out the message anew each time I was late; if the blackboard had not been used for something else in the interim, she could simply move it from the counter back to the table. Frenchie used the same token to say different things on different days. Each use of the token was a separate utterance.
One can imagine the same token being reused as a token of a different type of sentence. Suppose there is a sign in a flying school, intended to warn would-be pilots: ``Flying planes can be dangerous". The flying school goes bankrupt; the manager of a park near the airport buys the sign and puts it next to a sign that prohibits walking on high tightropes. In its new use the sign is a token of a type with a different syntax and a different meaning than in its original use. In principle, tokens could even be re-used for utterances in different languages; I leave finding such examples as an exercise for the reader.
In the case of spoken utterances in face to face
communication, the utterance/token distinction becomes
pretty subtle. One who
hears the token will see the utterance which produces it.
Writing brings with it the possibility of larger gaps
between use and perception; letters are sent, books are put
on shelves, to be read months or even years later, and so
forth. The utterance/token distinction is most at home
in the case of written text. It grows in importance
as culture and technology
develop. Modern technology allows for the storage and
reproduction of both spoken and written tokens, and with
such devices as email an utterance involves the production
of numbers of tokens around the world.
So, to review some distinctions and terminological decisions made thus far:
What Burks calls the ``existential relations'' of a token or utterance is now usually referred to as its ``context''. The ``context-dependence" of indexicals is often taken as their defining feature: what an indexical designates shifts from context to context. But there are many kinds of shiftiness, with corresponding conceptions of context. Until we clarify what we mean by ``context", this defining feature remains unclear.
The key distinction is between presemantic and semantic uses of context. Sometimes we use context to figure out with which meaning a word is being used, or which of several words that look or sound alike is being used, or even which language is being spoken. These are presemantic uses of context. In the case of indexicals, however, context is used semantically. It remains relevant after the language, words and meaning are all known; the meaning directs us to certain aspects of context.
Consider these utterances:
With (7) nowing that the utterance occurred in Frankfort rather than San Francisco might help us determine that it was German teenagers expressing enthusiasm and not American teenagers expressing disgust.
With (8), knowing whether our speaker has just arrived from Germany or just arrived from Saudi Arabia might help us to decide what the syntactic structure of the sentence is and whether ``good'' was being used as an adjective or an adverb.
With (9), knowing a little about the situation that this utterance is describing will help us to decide whether the person in question had lost her pet or was seeking security in an earthquake, and whether ``duck'' is a noun or a verb.
In each of these cases, the context, the environment of the utterance, the larger situation in which it occurs, helps us to determine what is said. But these cases differ from indexicals. In these cases it is a sort of accident, external to the utterance, that context is needed. We need the context to identify which name, syntactic structure or meaning is used because the very same shapes and sounds happen to be shared by other words, structures, or meanings. In the case of indexicals we still need context after we determine which words, syntactic structures and meanings are being used. The meanings exploit the context to perform their function. The meaning of the indexical ``directs us'' to certain features of the context, in order to fix the designation.
It seems then that a defining feature of indexicals is that the meanings of these words fix the designation of specific utterances of them in terms of facts about those specific utterances. The facts that meaning of a particular indexical deems relevant are the contextual facts for particular uses of it. However, indexicality is not the only phenomenon in which context plays a semantic role. anaphora provides another case.
In anaphora what one word designates depends on what another word in the same bit of discourse, to which the word in question is anaphorically related, designates. Compare
The designation of ``she'' in (10) simply depends on a contextual fact, whom the speaker was indicating. But in (11) the designation of ``she'' depends on which previous word in the discourse is taken as its antecedent. In both anaphora and indexicality we have semantic use of context; the difference is in the sorts of contextual facts that are relevant.
What is the relation between the ``she'' used in (10) and the ``she'' used in (11)? No one supposes they are mere homonyms. Many philosophers are at least tempted to suppose they are occurrences of a single ambiguous word, which sometimes functions as a variable and sometimes as an indexical[Kaplan, 1989a]. Many linguists find this implausible, and would prefer an account that gives a uniform treatment of pronouns, bringing the relativity to linguistic and other contextual factors into a single framework for a subject matter called ``deixis'' [Partee, 1989]. I have some sympathy with this point of view, but for the purposes of this essay I will set the issue of the precise connection of anaphoric and demonstrative uses of pronouns to one side.
With respect to contexts for indexicals, I need to emphasize two distinctions, which together create the four categories exhibited in Table 1:
I'll show which expressions fit into these categories, and then explain them:
The narrow context consists of the constitutive facts about the utterance, which I will take to be the agent, time and position. These roles are filled with every utterance. The clearest case of an indexical that relies only on the narrow context is ``I", whose designation depends on the agent and nothing else.
The wider context consists of those facts, plus anything else that might be relevant, according to the workings of a particular indexical.
The sorts of factors on which an indexical can be made to depend seem, in principle, limitless. For example,
When Rip Van Winkle says, ``I fell asleep yesterday," he intended to designate (let us suppose), July 3, 1766. He in fact designated July 2, 1786, for he awoke twenty years to the day after he fell asleep. An utterance of ``yesterday" designates the day before the utterance occurs, no matter what the speaker intends. Given the meaning and context, the designation is automatic. No further intention, than that of using the words with their ordinary meaning, is relevant.
The designation of an utterance of ``that man'', however, is not automatic. The speaker's intention is relevant. There may be several men standing across the street when I say, ``That man stole my jacket". Which of them I refer to depends on my intention.
However, we need to be careful here. Suppose there are two men across the street, Harold dressed in brown and Fred in blue. I think that Harold stole my wallet and I also think wrongly that the man dressed in blue is Harold. I intend to designate Harold by designating the man in blue. So I point towards the man in blue as I say ``that man''. In this case I designate the man in blue--even if my pointing is a bit off target. My intention to point to the man in blue is relevant to the issue of whom I designate, and what I say, but my intention to refer to Harold is not. In this case, I say something I don't intend to say, that Fred, the man in blue, stole my wallet, and fail say what I intended to, that Harold did. So it is not just any referential intention that is relevant to demonstratives, but only the more basic ones, which I will call directing intentions, following Kaplan [1989b].
In a case like this I will typically perceive the man I refer to, and may often point to or otherwise demonstrate that person. But neither perceiving nor pointing seems necessary to referring with a demonstrative.
The indexicals ``I", ``now", and ``here" are often given an honored place as ``pure" or ``essential" indexicals. Some writers emphasize the possibility of translating away other indexicals in favor of them [Castañeda, 1967], [Corazza, forthcoming]. In Table 1, this honored place is represented by the cell labelled ``narrow'' and ``automatic''. However, it is not clear that ``now" and ``here" deserve this status, hence the asterisks. With ``here" there is the question of how large an area is to count, and with ``now" the question of how large a stretch of time. If I say, ``I left my pen here," I would be taken to designate a relatively small area, say the office in which I was looking. If I say, ``The evenings are cooler than you expect here" I might mean to include the whole San Francisco Bay area. In ``Now that we walk upright, we have lots of back problems," ``now" would seem to designate a large if indefinite period of time that includes the very instant of utterance, while in ``Why did you wait until now to tell me?" it seems to designate a considerably smaller stretch. It seems then that these indexicals really have an intentional element.
``Here" also has a demonstrative use. One can point to a place on a map and refer to it as ``here'' [Kaplan, 1989a]. ``Now" and the present tense can be used to draw attention to and confer immediacy on the time of a past or future event, as when a history teacher says, ``Now Napoleon had a dilemma..."[Smith, 1989].
To repeat, as I use the terms, meaning is what the rules of language associate with simple and complex expressions; content is an attribute of individual utterances. The simple theory into which I am trying to incorporate indexicals focusses on the contents of utterances of four kinds. The content of a statement is a proposition, incorporating the conditions of truth of the statement. The content of an utterance of a predicate (for our purposes, a declarative sentence with some of its terms replaced by variables) is a condition on objects. The content of an utterance of a definite description will be a mode of presentation. The content of the utterance of a name will be an individual. The contents of utterances of terms combine with the contents of utterances of predicates to yield propositions.
The contents of utterances derive from the meaning language associates with expressions. The simplest way for this to happen is equisignificance: the meaning of an expression assigns the same content to each and every utterance of the expression.
But, as I explained in section 3.3, indexicals don't work this way. The meaning directs us to certain aspects of the context of the utterance, which are needed to determine the content. The object designated by an indexical will be the object that bears some more or less complicated relation to the utterance. Instead of the usual twofold distinction--sinn and bedeutung, meaning and denotation, intension and extension--we have a threefold one:
The reflexivity apparent in the second level has long been one of the major themes in the study of indexicals. Reichenbach put forward a token-reflexive theory in his Introduction to Symbolic Logic [1947].
Reichenbach claimed that token-reflexive words could be defined in terms of the token-reflexive phrase ``this token", and in particular, as he put it, ``The word `I"... means the same as `the person who utters this token'...(284)".
If we take Reichenbach's claim as a literal claim of synonymy between ``I'' and ``the person who utters this token'', it is wrong. The two terms may be assigned the same condition, but ``I'' refers whereas ``the person who utters this token'' describes. But Reichenbach was clearly on to something. There is an intimate connection between the meanings of ``I'' and ``the person who utters this token'', even if it falls short of synonymy. The second phrase does not have the meaning of ``I'', but it gives part of the meaning of ``I''. It supplies the condition of designation that English associates with ``I''. We can put this in a rule that brings out the reflexivity:
Here we see that the condition of designation assigned to an utterance u has that very utterance as a constituent, hence it is reflexive. (I discussed the reasons for using ``utterance'' rather than ``token'' above in section 3.2.)
This rule does two things. First, it assigns a binary condition to the type, ``I''. The condition is that x is the speaker of u. This condition has a parameter for the object designated and one for utterances. Second, the rule assigns unary condition, on objects, to each utterance of ``I'', by specifying that the utterance parameter is to be filled with that very utterance. To state this sort of rule in English, we would naturally make use of a reflexive pronoun:
Here are the conditions of designation for some familiar indexicals, in line with the discussion in section 3.3.
x(x is the speaker of u & x addresses y with u)
x
(x is the speaker of u & x directs u at t during
part of t)
: u designates y iff x (x is the speaker of u & x
directs u towards y)
n
In considering the meanings of sentences, it is helpful to think of propositions as 0-ary conditions. English assigns 0-ary conditions, propositions, to indexical-free sentences but assigns unary conditions on utterances to sentences with indexicals in them.
So our conditions of designation give rise to conditions of truth that also are reflexive. Meaning does not associate a proposition or 0-ary condition with a sentence containing an indexical, but a unary condition on utterances:
An utterance u of the form, where
is the subutterance of an indexical
, is true iff
(
)
So, for example,
is the subutterance
of ``you", is true iff (
designates y & y was born in Los Angeles);
(x is the speaker of
& x addresses y with u &
y was born in Los Angeles)
On David Kaplan's approach, the meaning of expressions in languages with indexicals are regarded as characters. Characters are functions from contexts to contents. So the meaning of ``I'' is a function, whose value is a for contexts in which a is the speaker and the meaning of ``I am sitting'' is a function whose value is the singular proposition that a is sitting for such contexts. This theory neatly captures what is special about context in the case of indexicality; that it plays a semantic role, rather than merely a presemantic one. I don't think Kaplan's view does as well with what is special about content in the case of indexicals, however. Kaplan provides only one level of content--official content--where I agree with Burks that more than one level of content is needed in the case of indexical utterances. In the next two sections I will defend Burks' perspective.
Reichenbach analyzed Luther's utterance, ``Here I stand,''
in terms of the relation 
, where x
is a person, 
is Luther's token and z a place:
is Reichenbach's term for Luther's
utterance; his analysis amounts to:
stands in
the place where is made
In our scheme, we have here a general proposition
about two modes of presentation, being the speaker of
and being the place of . Each of these
modes is a singular condition, with as a
constituent.
This proposition fits pretty well Burks' description of his fourth aspect, as what results from combining the meaning with the token or utterance. On the reflexive-referential account, the meaning of a sentence like Luther's is a condition on utterances, and Reichenbach's analysis fills the parameter of that condition with the utterance itself. It seems that Reichenbach's proposition or something like it deserves a central place in our account.
However, (12) is clearly not what Luther said. He didn't say anything about his own utterance, and he referred to himself with ``I'', rather than describing himself. (12) is not a good candidate for the official content of Luther's remark. Where then does it fit in?
On Kaplan's approach, the level of analysis represented by
(12) and by Burks' fourth aspect is bypassed [1989, 1989a].
The meaning, or character, of an indexical is, on
Kaplan's theory, a function from context to official
content, to what is said. The approach Barwise and I took
in Situations and Attitudes [1983] was similar,
although we did compensate somewhat with what we called
``inverse interpretation''. Stalnaker complained that
something was missing from such approaches [1981], and I have
come to think that he and Burks were correct. In
fact, we need a variety of contents.
A problem that underlies the simple picture of meaning and content is now going to come to the surface. The problem is that the concept of ``truth-conditions of an utterance'' is a relative concept, although it is often treated as if it were absolute. Instead of thinking in terms of the truth-conditions of an utterance, we should think of the truth conditions of an utterance given various facts about it. And when we do this we are led to see that talking about the content of an utterance is an oversimplification.
Suppose that you are at an international philosophy meeting. During what seems a stupid lecture, the person next to you writes a note which he passes to you. It say, ``Cet homme est brillant''. He then whispers, in English, ``Don't you agree?'' You are a confirmed monolingual, and don't even recognize in which language the message is written. To avoid compounding ignorance with impertinence, you nod. All you can infer about the message is that it is a statement, with which one could agree or disagree. Do you know the truth-conditions of his message?
Given the ordinary philosophical concept of the truth-conditions of an utterance, you certainly do not. You have no idea what proposition is expressed. If you did, you wouldn't have nodded as if you agreed.
But you could list some conditions, such that, were they met, the message would be true. Call the message m. It is true if there is a proposition P, such that in the language in which m is written, its words have a certain meaning, and in the context in which m was written, words with that meaning express P, and P is true.
It is fair to call these truth-conditions of m, because they are conditions such that, were they satisfied, m would be true. But they are not what philosophers usually have in mind when they talk about the truth-conditions or content of the message. They would have in mind the proposition that a certain person, the lecturer, was brilliant.
But this philosophical concept of truth-conditons is a special case of a more general one: the truth-conditions of an utterance given certain facts about it. What you know about m is its truth-conditions given only the barest facts about it, that it is a statement. You can specify conditions under which m would be true, but because you know so little about m itself, those conditions have a lot to do with m's relation to the rest of the world and say little about the world independently of m. The philosophical concept of truth-conditions corresponds to the case in which one knows a lot about m; in this case the conditions will pertain to the world outside m, not m itself.
If your high-school French started to return to you, you might reason as follows:
As you figure out more about m, fixing more of its linguistic properties, the conditions that had to be fulfilled for its truth become more focussed on the world. The additional or incremental conditions required for the truth of m, given all that you knew about m, were conditions on a certain person, that he be brilliant. Our philosophical concept of truth-conditions of an utterance is the incremental conditions required for truth, given that all of these linguistic factors are fixed.
This picture of truth-conditions as relative is just a matter of treating them like other conditions we ask about. Whenever we ask about the conditions under which something has a certain property, we take certain facts as given. What we want to know is what else, what additional facts, have to obtain, for the thing to have that property, given the facts we assume. I ask you ``Under what conditions will Clinton get re-elected?'' and you say, ``He has to carry California''. You are taking for granted a number of things--that he will lose the South, do well in the in the Northeast, get at least two midwestern states. Given all of this, what else does he need? To carry California.
It's the same with truth-conditions. What does the world have to be like for m to be true? That guy must be a brilliant lecturer. Right--given the facts about the language, the words, the meaning and the context of m, that's what else is needed.
As I mentioned above, I use three different kinds of content in the account of indexicals. These correspond to three kinds of facts one might take as fixed in assessing truth-conditions:
of an utterance corresponds to the
truth-conditions of the utterance given the facts that fix the
language of the utterance, the words involved, their syntax and their
meaning.
of an utterance corresponds to the
truth-conditions given all of these factors, plus the facts about
the context of the utterance that are needed to fix
the designation of indexicals.
of an utterance corresponds to the
truth-conditions given all of these factors, plus the additional facts
that
are needed to fix the designation of the terms that
remain (definite descriptions in particular, but also possessives, etc.).
We shall see below that we need all three kinds of content to adequately describe the epistemology of indexicals and other terms.
as cognitively relevant content
As we saw in section 3.4, the meaning of an indexical or sentence
containing indexicals provides a condition on utterances. We move from this condition to the content of an
utterance of that type by filling the parameter of that condition with the
utterance itself. In the case of indexical terms, we go from binary
conditions on objects and utterances to 1-ary conditions on objects. In
the case of sentences containing indexicals, we go from 1-ary conditions on
utterances to 0-ary conditions, propositions. These are propositions
about utterances.
Consider,
The content of (13) is a proposition about (13)
itself:
is the addressee of (13) &
y is the capital of Nebraska &
x was born in y)
As we noted, this proposition certainly does not seem to be the official content of (13), what the speaker said when he uttered (13)--a point I will emphasize in the next section.
Nevertheless, content is very important in understanding the
connection between meaning and cognition, how we use language to express
our beliefs, and influence the beliefs of others. It is
cognitively relevant content.
Imagine that I am standing next to W.V.O. Quine at a party. Consider the difference between my saying ``I would like to shake your hand'' and ``John Perry would like to shake your hand.'' In response to the first, we would expect Quine to extend his hand; in response to the second, he might well ask ``Well, where is he?'' (See [Castañeda, 1966], [Perry, 1979], [Stalnaker, 1981], [Perry, 1993].)
If we ask what I hoped to accomplish by saying ``I'd like to shake your hand'', we might just say that I wanted to make him aware that I wanted to shake his hand, so I said that I wanted to. This would be accurate, but incomplete. It leaves out many of my subgoals and my plans for acheiving them. I spoke the sentence, rather than including it in a letter or email, because I realized that he was standing where he could hear me. I said it in English because I thought that he understood English. I wanted him to be aware of that, in order to get him to turn and offer his hand for me to shake. In order to get that effect, I wanted to produce a certain kind of thought in him. I wanted him to think that the person in front of him wanted to shake his hand. My plan might be summarized as follows:
, thinking ``that utterance
is spoken by someone who wants to shake the hand of the
person he or she is addressing.''
Now the content of my utterance is the key to this plan. The
content of my utterance is simply the singular propositions that
John Perry want to shake W.V.O. Quine's hand. This is the same as the
content of ``John Perry wants to shake your
hand''. But there
would be no reason to expect this utterance to have the desired effect,
given my assumptions. The difference between them comes out at the level
of content .
as official content
Content is a useful tool for understanding the
motivation and impact of utterances. But it is not our
ordinary concept of content. It is not what I have called
official content, the content that corresponds to what the
speaker says. There are two main arguments for this; the
reader may be convinced by and familiar with the arguments,
but I want to highlight them to help us reflect on just what
they show.
The first and simplest I'll call this the ``samesaying argument''. Consider my utterance, directed at my son Jim:
The content of (14) is a proposition about
(14). But we would ordinarily count me as having said the
same thing to him as he said to me with his utterance
and the same thing I say to a third party with my utterance
But these two utterances have quite different
contents than (14). The content of (15) is a
proposition about (15) itself, and the content of (1)
is just a singular proposition about Jim (since names name,
their designation is fixed by their meaning). It seems then
that it is the individual designated by the subutterance of
``you'', and not the condition of being the addressee of
that subutterance, that makes it into the official content
of (14).
The second argument I call the ``counterfactual circumstances argument.'' To understand it, one needs to keep clearly in mind the difference between the conditions under which an utterance is true, and conditions under which what is said by the utterance (or perhaps better, what the speaker says, in virtue of making the utterance) is true. We can separate these, by considering counterfactual circumstances in which the utterance is false, but what is said by the utterance is true [Kaplan, 1989a].
Now suppose, contrary to fact, that when I uttered (14) I was mistaken, and was talking to my son Joe rather than Jim. In those circumstances, my utterance would have been false, since Joe was born in California. And what I would have said in those circumstances, that Joe was born in Lincoln, is false. But what I actually said, since I actually was talking to Jim, was that he was born in Lincoln. And that proposition, that Jim was born in Lincoln, would have been true, even if, when I uttered (14), I was talking to Joe.
The upshot of these arguments is that the official content of (14) is a singular proposition about Jim. This is the same proposition that Jim expressed with (15), and that I expressed with (1). And it is a proposition that would still be true even if I were talking to Joe rather than Jim, although of course then I would not have expressed it, but a quite different and false proposition about Joe.
Our other two kinds of content, content and content , both
assign this proposition to (14), (15) and (1). But these
differ with respect to
Content , recall, corresponds to truth-conditions with the contextual
facts fixed. The content of (6) is not a singular proposition about
Jim. The first argument role of x was born in y gets filled with a
mode of presentation of Jim, not Jim himself.
Content corresponds to truth-conditions with all
the facts that determine designation of terms fixed,
including in this case the fact that Jim is the manager of
Kinko's. So the content of (6) is our
singular proposition about Jim.
Content corresponds to Burks' concept of ``information
conveyed''. On this concept (14), (1) and (6) all
convey the same information, ``for they both refer to the
same object and predicate the same property of it.''
Which corresponds to official content, content or
content ? It depends on whether we think of
definite descriptions as referring or describing. If they refer,
then they contribute the objects they designate to official
content, and the right answer is that content is
official content. If they describe, then content
is the right answer. For the purposes of this essay, I
have accepted the traditional account of definite descriptions as
describing.
With this understanding of definite descriptions, it seems
that content corresponds to official content. When we
compare what people say, and consider the counterfactual circumstances in
which what they say is true, we fix the meaning and context, but let other
facts vary, even the ones that fix the designation of definite descriptions.
Consider,
said to Jim. When we think of the possible worlds in which this is true, what do we require of them? Worlds in which Jim was born in Iowa, but ``You were born in the capital of Nebraska'' means that 2 + 2 = 4 don't get in. We fix the meaning, before we consider the world. Worlds in which Jim was born in Iowa, but I am talking to Sue, who was born in the capital of Nebraska, don't get in. We fix the contextual facts, and so the designation of indexicals, before we consider the worlds. But worlds in which Jim was born in Omaha, and Omaha is the capital of Nebraska do get in. We consider the worlds, before we fix the facts that determine the designation of definite descriptions.
In maintaining that content is official
content, I agree with a movement in the philosophy of
language I call ``referentialism''. The referentialist
thinks that names and indexicals refer, and statements
containing them express singular propositions. This set of
views constitutes a movement because it had to overthrow an
opposing orthodoxy, which dated back to Frege and Russell.
Frege was troubled by singular propositions. How can
a proposition have an object in it? Won't there always be different ways of
thinking about the object? So won't a belief or desire or hope about an
object always involve some specific way of thinking about it? Shouldn't the
propositions we are worried about be ones that incorporate those ways of
thinking--shouldn't propositions always have modes of presentation, not
objects, as constituents?
This line of thinking led Frege and Russell away from
singular propositions; Frege didn't have them at all, and
Russell made less and less use of them as time went on.
Both concluded that names were something like hidden
definite descriptions; in our terminology, ordinary names
denote and describe
rather than name and refer. And
this became the standard view for the first two-thirds of
the century--with some dissenters like Burks and Ruth
Marcus ([Marcus, 1961], pp. 309-310). When Donnellan and Kripke attacked
description theories of names and argued that
names referred and statements containing them expressed
singular propositions, the feeling was that something like a
conceptual revolution was occurring. And Kaplan's ``direct
reference'' theory of indexicals seemed to turn the
revolutionary doctrine into unassailable common sense.
It seems to me that the referentialist movement was basically correct. Names and indexicals refer; they do not describe. Singular propositions may be sort of fishy, but they play a central role in the way that we classify content for the purpose of describing minds and utterances. Our concept of what is said is, as such things go, fairly robust.
Still, it is not entirely clear how far reaching the philosophical consequences of this revolution are. There are three attitudes towards the referentialist treatment of ``what is said'' or official content:
or content .
So the true contents must be something else.
The third, reasonable sounding view is of course my own. I
call it ``critical referentialism''--a term so ugly only
moderates could like it. The critical
referentialist believes that one commits ``the fallacy of misplaced
information'' [Barwise and Perry, 1983] when one expects that all of the content a
meaningful utterance carries can all be found at the level official
content. Critical referentialism is simply referentialism without the
fallacy. Free of the fallacy, the referentialist can employ other aspects
of content, such as content , to explain the motivations and impact of
language on semantically competent speakers and listeners,
without having to elevate it to official content.
According to critical referentialism, the counterfactual test and samesaying tests identify the proposition that best fits our intuitive conception of what is said by an utterance or what the speaker says in making an utterance. There are many other propositions systematically associated with an utterance in virtue of the meaning of the words used in it, which can and must enter into the explanation of the significance the utterance has for competent speakers and listeners.
The ``reflexive-referential'' account of
indexicals developed in this essay is an example of critical
referentialism. We need to consider the
content of statements containing indexicals to deal with
the sorts of cases that bothered Frege, such as our example
of meeting Quine. But for other
purposes, including those enshrined in our everyday concepts
for describing utterances, the referentialist concept of
what is said is useful and legitimate. Burks' original
account was also critically referentialist; he recognized
the importance of content for certain purposes, and of
content for others.
The importance of the contextual or official level of content stems from the basic facts of communication and the purposes for which our ordinary tools for classifying and reporting content are adapted.
In the paradigm communicative situation, the speaker suits the message to the listener's knowledge of the context of utterance and the impact on belief he hopes to achieve. That is, he assumes the listener to know the relevant contextual facts, and tries to convey the incremental content. I assume that Quine will recognize the speaker of ``I'd like to shake your hand,'(``that person in front of me'') and the addressee (``me''). Given this knowledge, the additional information he receives is: that person would like to shake my hand. The incremental content of my utterance, given the facts about context--the singular proposition that John Perry would like to shake Quine's hand--does a good job of characterizing what additional fact I am trying to convey to Quine, given what he knows and what will be obvious to him.
In a non-philosophical moment someone might explain Quine's action, of turning and extending his hand to me, by simply saying:
The embedded sentence here, ``he wanted to shake his hand'', does not seem to identify any of the modes of presentation that were crucial to my plan and Quine's understanding, as explored in section 3.5. And yet (16) is a perfectly adequate explanation.
We have to see this as a situated explanation. In the background
is the assumption that Quine and I were engaged in a normal case of face to face
communication. The explainer tells what I was trying to add to
what Quine knew and could easily perceive, and to do this it suffices to
identify the singular proposition that is the content of my
utterance. This is what the ordinary report does.
Frege's insight was that there are multiple ways to cognize
any object. Any utterance that adds to a listener's
knowledge in a significant way will connect to the modes of
presentation by which the listener already cognizes the
object, or can easily do so, and the modes of presentation
that connect with the ways the listener has for acting on
the object or dealing with information about the object. To
trace these interactions in a completely unsituated way,
making no assumptions, dealing with the listener's thought
processes in a way that doesn't rely on the external world
to suggest internal connections, would require what we might
call completely ``Fregean'' content, totally without lumps.
For practical purposes, what we need is ``Fregean-enough''
content. That is, we must specify the modes of presentation
that are actually involved in cognition and the ways they
are linked in the mind insofar as there is
something in the context of explanation that suggests that
the ordinary links might be broken.
Thus when I raised, in 3.5, the question of the difference between ``I'd
like to shake your hand'' and ``John Perry would like to shake your
hand'', I undermined the assumptions that make (16) an adequate
explanation of Quine's action. I asked for an account of exactly what is
taken for granted by (16), the planned connections between the modes of
presentation involved in the utterance (being the speaker of it, being
the addressee), and those involved in the cognitions the led to Quine's
action (being the man he sees, being himself).
When we retreat from the content of my utterance to
its content , to provide an explanation for the links now brought
into question, we
retreat to more Fregean, less lumpy content, in the sense that I and Quine are
replaced by modes of presentation. But note that the content is not
without lumps. For the content of an utterance is also a
singular proposition, about the utterance itself. The explanation I
gave in 3.5 is also situated; the assumption is that Quine hears my
utterance in the usual way, as it comes out of my mouth. If we asked why
I could get him to shake my hand by talking to him, but not by
saying the same thing in such a manner that his first perception of my
utterance was of an echo from a far room (details left to reader), we
would have to revert to even more Fregean content, with
modes of presentations of the utterance, rather than the
utterance itself, appearing in the contents.
