Indexicals and utterance production

Co-authored with Paula Sweeney. Forthcoming in Philosophical Studies.

We make various distinctions within a Kaplanian semantical framework for indexicals. Then we propose how to handle various non-standard uses of indexicals such as post-it notes, answering machines, and other cases as well. We argue against Stefano Predelli's analysis of these cases.

I've Read This

Please login or signup to add this paper to your Papers I've Read list

Added to your Read List

Readers (1)

• 30 Views
• Like Liked (1) Undo

Open Preview Close Preview Fullscreen Download (.pdf)

Indexicals and Utterance Production
Dylan Dodd  & Paula Sweeney  Arché – Philosophical Research Centre The University of St Andrews Forthcoming in Philosophical Studies

Steven Hawking is having a bad day. He has just lost another nurse; the fourth this month. The problem is that the voice synthesizer he uses to communicate has been malfunctioning, ‘saying’ things that Hawking has not instructed it to say. The first three times it was inconvenient and embarrassing when he ‘said’ “I hate you!” to his worthy nurses, but what he ‘said’ to this latest nurse was liable to land him with a harassment suit. “The machine is putting words in my mouth,” he tells the engineer. “I am not saying these things!” Intuitively, Hawking is not saying anything; and the machine (having no communicative intentions) is not saying anything. In fact, nothing is said; no proposition is expressed. We use this case and cases like it to both motivate our response to what is known as the ‘answering machine paradoxes’ and to shed light on recent variants of these cases. We claim that the most intuitive solution to these paradoxes requires one to distinguish between the agent (of a context), the tokening of a sentence and the agent’s chosen mechanism of communicating this tokening.

1

Introduction: the Kaplanian framework

Indexical expressions like ‘I’, ‘here’ and ‘now’ refer to different things in different contexts. When John says ‘I am sad’ he expresses the proposition that John is sad, but when Fred says it he expresses the proposition that Fred is sad; when John says ‘I am sad now’ at time t, he says that he is sad at t, but when he says it at t′ he says that he is sad at t′. 1

In his landmark paper “Demonstratives”, David Kaplan put forward an influential and elegant theory of the semantics of indexical expressions such as ‘I’, ‘here’, and ‘now’. According to Kaplan, in a context there is someone who makes an utterance. This is the agent of the context. This utterance occurs at a time, location and a possible world. In order to determine the referent or content of an indexical, we need an index of the context 〈w, t, l, a, . . . 〉 (where w is the possible world of the context, t the time, l the location, a the agent, . . . ). ‘I’, ‘here’ and ‘now’ refer to a, l and t respectively.1 Kaplan thought that ‘I am here now’ is a logical truth. As he puts it, this sentence “is deeply, and in some sense . . . universally true. One need only understand the meaning of [it] to know that it cannot be uttered falsely . . . A Logic of Indexicals which does not reflect this . . . has bypassed something essential to the logic of indexicals” (1989: 509). We shall argue that Kaplan is incorrect. ‘I am here now’ is not after all a logical truth. But he definitely says some true things here, and we agree that our semantics for indexicals should capture at least of some of the respects in which this sentence is special. One thing that is true is that in virtue of understanding its meaning one can know that ‘I am here now’ cannot be falsely uttered. But that is also true of ‘I am speaking now’, which, as we shall shortly discuss, Kaplan rightly does not consider a logical truth. Is there a sense in which ‘I am here now’ is more deeply and universally true than ‘I am speaking now’? We think there is. Think to yourself right now ‘I am speaking now’. Unless you happened to be talking, you just thought something false. On the other hand, you were thinking, so how about ‘I am thinking now’? Just understanding the meaning of that sentence is sufficient to know that it cannot be falsely thought, and if we just add the minimal knowledge that one cannot speak without thinking, we also are able to know that it cannot be falsely uttered. Kaplan also wouldn’t consider ‘I
1

We will be ignoring the “demonstrative uses” of ‘here’ and ‘now’ in this paper. Corazza 2004 cogently argues that the cases used by Smith 1989 to motivate the thesis that indexicals like ‘here’ and ‘now’ are ambiguous are in fact cases where the the words are being used anaphorically. We shall be ignoring anaphoric uses of ‘here’ and ‘now’ in this paper too. 2

am thinking now’ to be a logical truth. Is there a sense in which ‘I am here now’ is more deeply and universally true than ‘I am thinking now’? Yes. At every point in time in your life, you are at whatever location you’re at, but you’re not always thinking. If we pair ‘I am here now’ with you at every point in your life, and assign you as the referent of ‘I’, the time in question as the referent of ‘now’, and your location at that time as the referent of ‘here’, the result will be a true proposition. But since you’re not always thinking, we can’t do the same thing with ‘I am thinking now’. Note that we can only capture what separates ‘I am now here’ from ‘I am thinking now’ and from ‘I am speaking now’ by pairing these sentences with an individual; we can’t capture it by pairing an utterance of these sentences with an individual, or a thinking of these sentences with an individual. We will argue that semantics should not have ‘I am here now’ be a logical truth, but we agree that semantics should capture the way in which it’s special. Let any person a be the agent of the context, and let the time and location a occupies at any point in her life be the time and location of the context, pairing the sentence ‘I am here now’ with a context with this agent, location and time in the index of the context will yield a content that is true.2 Since we agree that the specialness just pointed out of ‘I am now here’ should be capture in our semantics for indexicals, we adopt Kaplan’s framework where the semantics takes one from sentence-context pairs to contents. In this respect we differ, for instance, from Perry 2003 who does the semantics in terms of utterance-context pairs. There are other reasons for adopting Kaplan’s sentence-based semantics over an utterance-based semantics. For instance, we want it to be a consequence of our semantics that ‘There’s no beer’ to entail

2

Predelli 2008 says that what is special about ‘I am here now’ is just that it’s “selfverifying”, which means this. Take the set of all contexts C such that agent of the context is at the location and time of the context, this sentence can only be uttered truly at C. He then points out rightly that ‘I am uttering something now’ is also self-verifying, and concludes that ‘I am here now’ is not special in a way that ‘I am uttering something now’ is not (62-4). He fails to see how ‘I am here now’ is different because he pairs utterances of sentences with contexts, not simply sentences with contexts. 3

‘There’s no beer in Sally’s liquor store’. But we can only capture this entailment of we have a sentence-based semantics. Say an utterance of ‘There’s no beer’ is true because there is a quantifier domain restriction. It’s true because all the salient beer – all the beer at the party – has been consumed. But if one were to utter ‘There’s no beer in Sally’s liquor store’, that would expand the domain of the quantifier to include the beer in Sally’s liquor store. Given that there’s always plenty of beer there, this sentence can never be uttered truly. So how do we get our semantics to have the utterance of ‘There’s no beer’ be true, and entail the truth of ‘There’s no beer in Sally’s liquor store’? By following Kaplan in doing our semantics in terms of sentences rather than utterances. We associate the sentence ‘There’s no beer’ with a context C. The semantics takes us from <‘There’s no beer, C> to a content that is true. The semantics also takes us from <‘There’s no beer in Sally’s liquor store’, C> to a true content. In spite of the fact that this sentence can never be truly uttered, it is true when paired with certain contexts, just like ‘I’m not speaking’. We shall now spell out our Kaplanian framework in more detail. As we see things, there are at least two phases of getting from an utterance to the content of the utterance – the proposition the utterance expressed. First, there is the pre-semantic phase were we associate a sentence S – the sentence that was uttered – and a context C with the utterance. Determining which context to associate with the utterance gives us the index of the context , for that is part of the context. Then there’s the semantic phase. When we feed a sentencecontext pair  into the semantics, it outputs a content, the proposition expressed. Also when we feed word-context pairs into the semantics, it outputs a referent for the term. For instance, on our view <‘I’, C> always outputs the agent of C as the referent of ‘I’. Kaplan called the function from an indexical-context pair to a referent the indexical’s character. In this paper we shall have much to say about the pre-semantic phase; in particular, we shall focus on difficulties in figuring out what context gets associated with a particular

4

utterance. It should be pointed out, however, that there can also be difficulties in determining what sentence was uttered, and thus what sentence to associate with a particular utterance. For instance, according to Kaplan 1990, names like ‘Mike’ are not ambiguous between the different people named ‘Mike’; rather, “different Mikes” have different names that share a spelling and pronounciation. If this view is right, then when one hears ‘Mike is happy’, it won’t necessarily be obvious what sentence was uttered because we may not know what name was just used. It could’ve been Mike Smith’s name, or it could’ve been Mike Jones’ name. Thus there is a pre-semantic difficulty of determining which name, and therefore which sentence, to associate with an utterance. To repeat, this is a view according to which names are unambiguous, and the epistemological issue of determining what is being said is pushed into pre-semantics. Similarly, although ‘bank’ is often presented as a classic case of ambiguity, we think this is a mistake. The fact that there are river banks and banks that are financial institutions does not show that there is a single word ‘bank’ that is ambiguous. Rather, there are distinct words that share a spelling and pronounciation. (This fact is noted in dictionaries. They will list ‘bank1’ and ‘bank2’ as distinct entries, whereas they also have single entries with distinct definitions (meanings).) If one hears an utterance of ‘Sally is at the bank today’, it may not be obvious which sentence was uttered due to an uncertainty about what word is the fifth word of the uttered sentence.3 Finally, Kaplan devised a semantics for indexicals according to which ‘I am here now’ is a logical truth. In order to do this, he said that semantics can only use what he called “proper contexts”, contexts with an index  such that the agent a is at location l, at time t, and world w. We agree that if only let proper contexts play a role in our
3

As Bach 1998 puts it, “Ambiguity is, strictly speaking, a property of linguistic expressions. A word phrase, or sentence is ambiguous if it has more than one meaning . . . However, it is not always clear when we have only one word. The verb ‘desert’ and the nount ‘dessert’, which sound the same but are spelled differently, count as distinct words (they are homonyms). So do the noun ‘bear’ and the verb ‘bear’, even though they not only sound the same but are spelled the same.” 5

semantics, ‘I am here now’ will come out as a logical truth – it will be true at all the contexts in the semantics. But we shall deny Kaplan’s claim that this sentence is a logical truth. Our disagreement with Kaplan is about whether there are true utterances of the sentence. We want our semantics, which is done completely in terms of sentences, to model the truth of utterances, among other things. In explaining why ‘I am here now’ is special in a way that ‘I am speaking now’ is not, and in claiming that our semantics should be able to account for this, we pointed to the fact that we want our semantics to model things besides the truth of utterances. Nevertheless, we do want our semantics to able to provide the truth conditions of any possible utterance. If ‘I am not here now’ can be uttered truly, we need to allow contexts C into semantics such that the semantics will take <‘I am not here now’, C> to a true proposition. Such contexts would be in the class of improper contexts that Kaplan purged from his semantics, for they would be contexts where a was not at l at t at w. If we let such contexts into our semantics, when they’re paired with ‘I am here now’ the semantics will take the pair to a falsehood, and thus ‘I am here now’ won’t be a logical truth. So, if there are true utterances of ‘I am not here now’, then we need to let some contexts into our semantics which will prevent ‘I am here now’ from being a logical truth. Are there true utterances of ‘I am not here now’? Kaplan thought not. We think there are.4 If we focus on conversations where individuals are talking to one another in a normal way, it will seem obvious that there cannot be true utterances of ‘I am not here now’. But there are other forms of communication to take into consideration. As Sidelle 1991 pointed out, it’s common for answering machine messages to include the claim ‘I am not here now’. Assuming that the person speaking really isn’t home when the message is received, this part

4

Since we think this, we think some of Kaplan’s “improper contexts” should be allowed back into semantics. However, we don’t think they all should. In particular, we don’t see any reason to let contexts in where the agent of the context is not at the world of the context. We also don’t see a need to let in context in which the time of the context is prior to the time of the agent of the context’s birth. 6

of the message would express something true. Or as Predelli 1998a and 1998b pointed out, one can leave a note for one’s wife that says ‘I am not here now’, or leave a post-it note outside one’s office that says this, and assuming that it is read when one isn’t at one’s house or office, it seems that these messages also express something true. Thus it seems that an utterance of ‘I am not here now’ can expresses a truth, which means that Kaplan was wrong about ‘I am now here’ being a logical truth. The answering machine, post-it note and letter cases have been used to support what seem to us to be radical claims. For instance, Predelli 1998a and 1998b has proposed an intentionalist position, according to which ‘here’ and ‘now’ needn’t refer to the times and locations within the context of utterance, but can refer to times and locations of a different context, the context the agent intends, which he calls a context of interpretation. We understand this not as being a semantic claim but as a pre-semantic claim about what context to associate with a given utterance. According to Predelli, the agent intends a certain context to be the context her utterance is interpreted with respect to, and this is the context we associate with the utterance. Predelli’s position would be opposed, for instance, by someone who thought you look at the situation in which the utterance was made to find the context to associate with the utterance. And thus the time of the context wouldn’t be the time the agent intends, but the time at which the utterance was made. Romdenh-Romluc 2002, on the other hand, thinks we find the context to associate with an utterance not by looking at the context the agent intends, but the relevant context is the context a competent audience would identify. Considering variants of the post-it note case, Corazza, et al 2002 additionally claim that ‘I’ needn’t refer to the agent of the context, but can refer to someone else. (Note that, in contrast to Predelli’s and Romdenh-Romluc’s claims, this thesis seems to be about the character of ‘I’, and thus is semantic.) But there are also conservative responses to the cases. For example,

7

Stevens forthcoming proposes a treatment of the cases that requires no alteration to Kaplan’s theory at all, and defends his claim that ‘I am here now’ is a logical truth. In this paper we also favour a conservative treatment in which the spirit of Kaplan’s theory of indexicals is retained, whilst denying his claim that ‘I am here now’ is a logical truth. We show that these ‘problem’ cases can be seen to be not all that different from many other forms of communication. Looking at some unusual forms of communication reveals some mundane facts about communication. We think the discussions of the cases as well as Kaplan’s original semantics for indexicals have ignored these mundane facts. We re-examine the cases in light of these mundane facts, and discuss at least one possible greater lesson for the semantics of context sensitivity.

2

More non-standard cases

Let’s return to the case we began the paper with involving the famous physicist Stephen Hawking. In order to speak, Hawking sends inputs with a hand to a computer, and the computer responds by producing a digitized voice in accordance with his inputs. Of course most of us just open our mouths and talk. This difference between the way Hawking verbally communicates and the way most people do is irrelevant semantically speaking. A theory of linguistic communication and meaning should pay no attention at all to the fact that Hawking uses an unusual mechanism to make his utterances – his utterances would express the same propositions if he spoke in the way most of us do. 5 All of this seems to be obvious. The way Hawking communicates and the way we do are just a few of the possibilities. Hawking’s computer happens to verbalize words, but it could have produced

5

Throughout this paper we use the term mechanism in a slightly technical sense. A mechanism need not be a piece of machinery like Hawking’s computer or an answering machine; it may be another person (a potential agent) but, as we will argue later, we should not let this detract us from the fact that they are simply part of the mechanism, working on behalf of the agent. 8

text on a screen instead. And there are more possibilities. Instead of speaking or using a computer, he could have a bag full of signs with various sentences written on them. Instead of verbalizing the sentence S, he could hold up a ready prepared sign that has S written on it. Or he could tap on the arm of his assistant in morse code, who would then write the sentence down on a piece of paper, and then hold it up for all to see. Or instead of writing the sentence down, the assistant might just say the sentence out loud. Instead of tapping on the assistant’s arm, Hawking might whisper into the assistant’s ear what he wants him to say more loudly to his audience (maybe he can’t speak loud enough). There are doubtlessly many, many other possible forms of communication. And for all of the ways we have mentioned, if someone did communicate in such ways, that would be semantically irrelevant – it wouldn’t change what propositions the sentences uttered expressed. As far as semantics is concerned, if someone were to engage in such modes of communication, she might as well have just spoken her sentences in the normal way. And often when we use one of these non-standard ways of communicating there are indexicals involved. Hawking may have a ready prepared sign that says ‘I am hungry now’ that he holds up when he wants some food. If Hawking’s girlfriend has a sore throat he may lend her his sign so she can hold it up to express the proposition that she is hungry. When on our holiday we may send a postcard home saying ‘The weather is warm here now’ and successfully express the proposition that the weather is lovely at the time and location of writing. Yet, I may on returning home use the same postcard and add ‘so wear light clothing’ to leave a note for my children, expressing the proposition that the weather is lovely at this different time and location. When I write ‘now’ in my diary, it refers to the time of writing, regardless of when it is read. On the other hand, when writing your will you may write ‘Now I am dead and gone’, clearly expressing a proposition about the time of reading, not writing.

9

Clearly the mechanisms of communication are both numerous and pliable, with straightforward speech being only one of them. We have claimed that the fact that these unusually forms of communication are being used is irrelevant to semantics. What does this mean? Recall that in our Kaplanian framework, semantics provides functions from sentence-context pairs to contents, and from word-context pairs to referents. We shall soon be discussing the complications that arise from the use of such non-standard forms for the issue of what context to pair with the uttered sentence. So they definitely complicate the question of what the content of the utterance is. However, we think they only complicate things at the pre-semantic phase, while leaving what happens in the semantic phase untouched. In this respect it’s like the word ‘bank’ (the word that means a financial institution). The fact that this word shares a spelling and pronunciation with another English word doesn’t affect its semantics, though this fact can make it difficult sometimes to determine what the right sentence-context pair to associate with a certain utterance.

3

Distinction between agent, mechanism of production, and sentence tokening

Communicating in one of the non-standard ways mentioned above involves using a ‘middle man’, someone or something acting on behalf of the agent. The agent kicks into action a process that produces a tokening of the sentence the agent utters. In order to account for contexts of utterance involving such forms of communication, we must distinguish between the agent of the context, the mechanism of production of the sentence the agent utters, and the tokening of the sentence the agent utters. The agent kicks the mechanism of production into action, it churns out the sentence tokening, a process which may take time to accomplish

10

its goal, and then at last the product is produced, that is, the sentence is tokened. All these stages are involved in the agent’s making an utterance.6 It’s possible that the event of the agent’s kicking the mechanism of production into action will occur at a different time and/or location than the event of the sentence tokening that results from the mechanism of production doing its work. To take a simple case, an agent could type a message into a computer at t1, but the sentences in the message aren’t tokened for the agent’s audience until the later time t2. (This often happens with e-mail, for instance.) Similarly, the agent might do the typing at location l1, but the sentence is tokened in some other location l2. THE STAGES OF UTTERANCE PRODUCTION7 STAGE 1 The agent of the context kicks the mechanism of production into action. time, location = t1, l1 STAGE 2 The mechanism of production works towards its end, namely producing a sentence tokening. STAGE 3 The sentence is tokened, and the agent’s message is now broadcast to her audience. time, location = t2, l2

It should be uncontroversial that linguistic communication can and sometimes does involve Stages 1 – 3 above. We think that noting these different stages sheds light on some controversial questions about indexicals. Let’s revisit Kaplan’s argument that ‘I am here now’ is a logical truth and ‘I am not here now’ a logical falsehood. We think Kaplan, in assuming that ‘I am not here now’ could never be uttered truly, was assuming at least two things: (i) ‘I’ refers to the agent of the
6

The distinctions we make in our account of utterance production benefit from Perry’s distinction between utterances, as intentional acts of communication, and tokens, as the effects (physical events or objects) that are the effect of such acts. As Perry notes, tokens can often be re-used with no semantic hangover from one use to the next. (See Perry, 2001: 38-9, and 2003: 376-7.)
7

We don’t intend anything temporal by our use of stage 1, 2 and 3 – they could just as effectively be thought of as elements of utterance production. 11

context, ‘now’ to the time of the context, and ‘here’ to the location of the context, and (ii) the agent must be at the time and location of the context when she utters her sentence. But we have seen that we must take into account the above three stages in order to account for certain forms of communication. And we have seen that Stage 1 and Stage 3 may be at different times and locations.

4

Applying our theory to some cases

In the discussion of answering machine message in the literature, much of the focus has been on ‘now’. For, with respect to old-fashioned answering machines in one’s home, the Stage 1 location – the place where one first speaks one’s message into the device – and the Stage 3 location – where the message is broadcast from – is the same, one’s home. And this is intuitively the value of ‘here’. In answering machine messages, ‘now’ does not refer to the time at which the agent speaks the sentence ‘I am not here now’. The agent speaks this message into his machine long before any caller hears it. But if an agent’s use of ‘now’ could somehow pick out the Stage 3 time, while also avoiding being at that location at that time, s/he could truly utter ‘I’m not here now’. Then Kaplan’s would be wrong about ‘I am now here’ being a logical truth. We propose that this is what is happening in answering machine and post-it note cases. In the answering machine case, the Stage 1 time is the point of time at which the agent speaks into the recording device. The Stage 3 time is when someone calls and hears the agent’s voice saying ‘I’m not here now’. The mechanism in this case is the answering machine itself, and whatever telephone equipment and cable makes it possible for the caller to receive this message. When the agent speaks into the recorder, s/he’s kicking the mechanism into action, and we’re at Stage 1. The answering machine saves the message, and sends it out to someone else’s phone when they call. This is the mechanism doing its work. Stage 3 doesn’t take place until the message is

12

finally sent out to a caller (alternatively, when the caller receives the message – strictly speaking this happens slightly later). We posit that answering machine communication is just another non-standard mechanisms of communication. As we saw with those cases, the fact that the agent of the context is temporally distant from the point at which her message is delivered to her audience doesn’t prevent the agent from making a genuine utterance and expressing a proposition with it. And we saw that the agent is able to make ‘now’ refer to times and locations that differ from the Stage 1 time and location. ‘here’ and ‘now’ can refer to either the Stage 1 or Stage 3 location and time, but it doesn’t follow that either word is ambiguous. Contrary to what Smith 1989 claims, the cases under discussion do not motivate positing an ambiguity in either word. That is because whether or not ‘here’ or ‘now’ refers to the Stage 1 or Stage 3 location or time depends on what happens at the pre-semantic phase of associating a sentence-context pair with an utterance; in particular, it depends on what context we associate with the utterance. To say that, for instance, ‘here’ is ambiguous is to say it has distinct characters (as Smith agrees). But we’re not positing distinct characters for that word – its character always takes you to the location of the context. Note that the reason we don’t find ambiguity is because in our Kaplanian framework, semantics takes one from sentence-context pairs to contents. If our framework was one in which semantics took utterance-context pairs as arguments, we would have to posit ambiguity.8 Post-it note cases provide another counterexample to Kaplan’s claim that ‘I am here now’ is a logical truth. Say I call my friend and tell him to put a note on my office door that says ‘I’m not here now’. In this case, my calling my friend is my kicking the means of production into action. My friend’s actions of writing the note and pasting it on my door are

8

It seems to us that Smith is assuming a framework where semantics takes utterance-context pairs as inputs. He talks of the character of indexical taking one from a “use” to a referent (1989: 167). 13

all part of the means of production. When someone reads the note on my door, at that point in time we’re at Stage 3. This might seem objectionable. After all, the above chart of The Stages of Utterance Production says that Stage 3 happens when the sentence is tokened, and the sentence is tokened, it seems, when my friend writes the sentence down on the post-it note (or perhaps puts the note on the door), not later when someone comes to my office and reads it. But the key here is that Stage 3 happens when the message is broadcast to the agent’s audience. Imagine that instead of a post-it note we have a computer screen on our office door that, when switched on, flashes up the pre-programmed notice ‘I am not here now’ whenever someone steps on the carpet outside the office door. Clearly ‘now’ refers to the time of the displaying, not the programming. Or compare the post-it case with the case mentioned previously where one has several sentences written down on different signs, and communicates by holding up a sign to one’s audience for them to read it. The key point in time here is not when the sentence was first written, but the point at which one holds the sign up for people to read – when one’s message is broadcast to one’s audience. And it’s not until an audience reads the post-it note on my door that my message is broadcast. What the note says is that I’m not at my office at the time at which someone reads the note. Thus the ‘now’ and ‘here’ I utter through the note refer to the Stage 3 time and location, just like in the answering machine case. In determining the context to associate with the utterance at the presemantic phase, we find one such that the time of the context is the time at which the message is broadcast. It may seem that the view we are offering is just the same as Predelli’s view. That this is not so will become clear in consideration of further cases below. We do agree with Predelli about the truth conditions of the sentences uttered in the answering machine and post-it note cases. But we disagree with his explanation of why the utterances have these truth conditions. We favour a more conservative explanation than his. We think that in these cases there are

14

two times/locations that could be the times/locations of the context, namely the Stage 1 and Stage 3 times/locations. We argue that, in virtue of the chosen mechanism of communication, a speaker can force the Stage 1 and Stage 3 times/locations apart. In that sense we are at least open to the possibility that a speaker’s intentions place an important role in determining what the context is – whether it’s a context with a Stage 1 or a Stage 3 time/location. But when we talk of an agent intending a Stage 3 time/location we do not mean that the agent can select the time and location of Stage 3. Predelli allows the agent’s intentions to do something rather dramatic. He thinks agent’s intentions can force us to look at a context of interpretation – the context the agent intends for us to evaluate her utterance with respect to – for the referents of ‘now’ and ‘here’. We think that this is to give intentions too much power. And we suggest that there is no need for this radical proposal simply to get the right truth conditions of the relevant utterances. So, although Predelli and us agree about the truth conditions of answering machine messages, we are in a position to offer a more conservative explanation of these truth conditions. And, as the cases below will show, we don’t always agree about truth conditions. Consider Predelli (1998b: 402)’s note to his wife (‘I am not here now . . . Meet me in six hours at the Hotel Cabo Real.’) written at 8am, intended for his wife to read at 5pm, but not actually read by her until 8pm. Predelli thinks ‘now’ refers to the time at which he intends his wife to read the note, 5pm. He thinks that the context that should be associated with this utterance is the context he (the agent) intended. But on our model, ‘now’ will refer to 8pm.9 What context is associated with the utterance is determined by the time at which the message is broadcast, i.e., the Stage 3 time. Thus ‘now’ refers to the time his wife reads the message, 8pm. And, since ‘meet me in six hours’ means six hours from now, we also give different results about when he is telling his wife to meet him (2am, not 11pm). Predelli has the
9

We’re assuming it doesn’t refer to the Stage 1 time in this case.

15

intuition that the ‘now’ refers to 5pm, so he will claim that his theory is making correct prediction about this case, and ours an incorrect one (see his 1998a: 110).10 We don’t think so. Intuitions regarding this case vary according to the background information. For example, if we are told that it is common knowledge between Predelli and his wife that she consistently arrives home at 5pm then we are more likely to expect his wife to latch onto his intended communication and meet him at 11pm. But this has nothing to do with speaker’s intentions. What we are interested in is what is semantically expressed, not what is pragmatically conveyed. We think our theory actually makes the right prediction, namely that the proposition Predelli literally expressed is that he wasn’t home at 8pm (when she red the note), and what he literally told her to do is to meet him at 2am, not what he intended to tell her, namely to him at 11pm. We think our theory’s prediction becomes plausible when we compare Predelli’s note case to other cases of communication. Consider the case where Tom types into a computer with a slow processor. His message is displayed on a screen five minutes later, and when answering the question ‘What time is it’ he says ‘You can find out what time it is now by looking in the upper right hand corner of this reader board’, referring to the clock on his reader board. ‘Now’ refers to the time at which the message appears, not when Tom typed it. Say that for some reason the computer takes twice as long as usual to process his message, so that it doesn’t appear until ten minutes after he typed it. He intended the ‘now’ to refer to the time five minutes after he typed it so, according to Predelli, that is what it would refer to. We find Predelli’s theory’s prediction here to be implausible. Say Tom intended and expected the message to appear at 8:53, but it actually appeared at 8:58. The utterance, on Predelli’s view, is false. Since ‘now’ refers to 8:53 and the message doesn’t appear until 8:58, you can’t tell
10

Sidelle 1991 shares Predelli’s intuition here, and amends his theory accordingly (see p. 537, cited by Predelli 1998a: 111). 16

what time is ‘now’ by looking in the upper right hand corner of the reader board. It says 8:58 up there, but ‘now’ refers to 8:53! We find it more plausible that that the utterance is true. And it is on our theory, according to which ‘now’ refers to whatever time the message is broadcast, i.e., read, and that time is identical to the time displayed in the upper right hand corner of the reader board. So this is a case where our theory makes the right prediction, and Predelli’s seems to make the wrong one. We don’t see any relevant difference between the case just described and Predelli’s note case. We think that they should receive the same semantic treatment. Just as ‘now’ refers to the time at which Tom’s message is read rather than the time Tom intended ‘now’ to refer to, so ‘now’ should refer to the time at which Predelli’s wife reads his message, not the time Predelli intended ‘now’ to refer to. To conclude, we agree with Predelli against Kaplan that ‘I am here now’ is not a logical truth, and we also agree with him about the truth conditions of answering machine messages. But our proposed explanation is more conservative than his. And when generalized to other cases, we give different truth conditions than he does.

5.

Some problem cases

Although most of the focus in the answering machine and post-it note cases has revolved around the value of ‘now’, we can consider some interesting cases involving the value of ‘here’. What if there is no answering machine in my home, but instead I speak into a receiver, my voice is recorded into a centralized data processing centre owned by my phone company and located far away from my house, and it is this processing centre that sends my message ‘I’m not here’ to the caller. Now it seems that the Stage 3 location is the processing centre, and not my house, and yet the ‘here’ refers to my house. Thus it doesn’t refer to the Stage 3 location. Nor does it refer to the Stage 1 location. This case is a problem for our model.

17

The case isn’t a problem for Predelli’s intentionalism. The speaker can intend a context of interpretation, and ‘here’ will refer to a location of that context. If she intends her house as the relevant context, then ‘here’ refers to her house. Thus this case motivates Predelli’s view. But perhaps things aren’t so clear cut. Admittedly, there is a problem for our view here, but problems are lurking for Predelli’s too. There seem to be many kinds of problem cases and peculiarities around the area. Below are some cases and considerations, which are worth taking into account in considering what to say about ‘here’ in these non-standard cases: Mobile phone (Consideration) It seems peculiar, at least to us, that someone would record ‘I’m not here now’ as one’s message for one’s mobile phone. Instead one should say something like ‘I’m not currently available’. Office-Home You’re going on holiday. You have a system where you can hook your office phone up to your home answering machine so that it answers for both your home and office. Whether it’s your home or office that someone calls, the same recording is played, the recording of your voice, which you originally intended only for your home. Your home and office are miles apart. Someone calls your office and gets the message. Intuitively, ‘here’ refers to your office. Office-Home Mixup It’s just like Office-Home, only what happens is due to an accidental and bizarre mix up of wires. You’re not on vacation; in fact you don’t want calls to your office going to your home answering machine. But that’s what happens. It seems that in this case, when someone calls your office and gets your home answering machine as a result, the message says that you’re not in your office at that time. ‘here’ refers to your office. You wish it didn’t, though. It seems to us that Predelli’s theory plainly gets this case wrong. In no sense did you or anyone ever intend the ‘here’ in your answering machine messages to refer to your office. Yet when someone calls your office, that’s just what the ‘here’ in the message refers to. Vacation Letter You live in Scotland, but you’re on vacation in Barbados. You’re writing a letter to your friend, not even mentioning anything about your vacation, but carrying on about your life in Scotland. You say ‘Things are going well for me here’. You mean that things are going well for you in Scotland. You’re not talking about Barbados. ‘Here’ clearly refers to where you live in Scotland. But the surroundings may influence the reference. You’re talking about Scotland. The return address you put on the envelope is your Scottish address. But what if you

18

sent a tourist postcard with a picture of Barbados on it? What if you’re staying at your vacation home in Barbados, and put your Barbados address as the return address? Changing the surroundings like this make it harder for ‘here’ to refer intuitively to Scotland, regardless of your intentions. The details we were considering in Vacation Letter and Office-Home Mix-up make us think there’s much more to be said regarding what determines the referent of ‘here’ in these cases than anyone’s intentions. Office-Home Mix-up strongly suggests that with answering machines, ‘here’ must automatically refer to the location being called. Now consider this case:

Yard You’re out in your garden. Someone calls your house. You don’t hear the phone ring and your answering machine picks up the call saying ‘I’m not here now’. Was that true? Well, you weren’t in your house. But we think what was said was false. You were there. Say the person calling is your neighbour, and she sees you out in your garden. A natural thing for her to say is, “Well, actually he is there, I can see him!” We think that Yard suggests that ‘here’ refers to the address associated with the phone’s location. This thesis explains Mobile. Mobile phones, unlike home phones, aren’t associated with an address, which is why it would be strange to say ‘I’m not here’ on one’s mobile phone message.11 The case we started with, that of the centralised answering system, doesn’t seem to be a problem for Predelli’s intentionalism. The speaker can intend a context of interpretation, and this is the context we associate with her utterance. ‘here’ will refer to whatever the

11

Could it be that a convention has developed down through the years whereby ‘here’ in an answering machine message refers to the address at which the phone is located? Here’s how it might have happened. Back in the old days before mobiles and such, phone numbers were simply associated with an address. When you called the number, and heard ‘I’m not here now’, the point was that no one was at the address listed for that phone number. Thus ‘here’ as used in answering machine messages has come to have this reference by convention. Perry (2003: 384) makes a similar point, stating that ‘here’ refers to ‘the place where the phone is or is expected to be, not the place where the answering machine is.’ 19

location of that context is. If she intends a context with her house as the location, then ‘here’ refers to her house. Thus as Predelli points out, this case motivates his view. But, as we have illustrated, there are similar cases that don’t seem to be good for his view. In particular, his theory struggles with Vacation Letter, and seems plainly to get Office-Home Mixup wrong. We think these cases are a mess. Apparently no one’s theory gets them all right.

6

‘I didn’t say that!’: ‘I’ is a pure indexical

Is ‘I’ a pure indexical? A pure indexical is an indexical such that the semantic rules governing it are alone sufficient to fix its referent. The character of the term doesn’t need to be supplemented by things in the context like a demonstration, the agent’s intentions, etc. For example, ‘that’ is not a pure indexical, because it doesn’t have rules governing it that are alone sufficient to fix the referent. On the other hand, it is plausible that ‘I’ is a pure indexical. The semantic rule governing it says that when ‘I’ is used in a meaningful sentence, it refers to the agent of the context. Many, including Kaplan, have thought that ‘I’ must always refer to the agent of the context. We agree. In what follows we shall argue that there has been no serious challenge in the literature to the initially plausible thought that ‘I’ is a pure indexical. In particular, no post-it note cases pose a serious challenge to this claim. Again, we stress that the key to understanding post-it note cases lies in considering them in light of the three elements of utterance production outlined above. It’s clear that a person (other than the agent) can be part of the mechanism of communication. For instance, say Peter dictates a paper to Mary his secretary, who then writes it up for him. Although there is a natural sense in which it’s Mary who writes the paper, there’s another sense in which it’s really Peter. Mary wrote up the paper – she helped produce it – but it’s Peter’s paper she is producing, not her own. If Peter said some offensive things in his paper and we tried to hold Mary accountable she’d speak truly if she said, “I

20

didn’t say those things, Peter did.” That’s because Peter is the one speaking to his audience through the paper, the one who is communicating, while Mary is merely part of the mechanism of communication. In Kaplan’s terminology, Peter is the agent of the context and Mary is not. And in terms of our table above, Mary appears only in the middle column, only as part of the mechanism. It is true that there might not be a communicative act for us to semantically analyse if it weren’t for Mary, but our analysis won’t pay notice to her any more than it would pay notice to the computer she typed the paper on, or the printer which printed it. All we just said about Peter and Mary would also be true if instead of printing the paper Mary merely memorized the message Peter wanted her to later deliver orally to his audience. In such a case, although Mary might be the speaker of Peter’s message, Peter would be the agent of the context – he would be the one really doing the communicating. And again, Mary only features in the second column as part of the mechanism. When a king sends his sentry to another king, who delivers his message ‘I surrender, let’s discuss terms’, it may be the sentry who is speaking, but the king is the agent of the context and the one who is really communicating with the receiving king. That’s why the receiving king doesn’t think it’s the sentry who is surrendering, but the enemy king. Kaplan’s claim about ‘I’, recall, is that it always refers to the agent of the context. It’s essential to distinguish this claim from the claim that it refers to the speaker – the one who actually verbalizes (speaks) or writes down the message. The case of the king and his sentry is a counterexample to the latter claim, since as we saw the ‘I’ refers to the king and not the sentry. But it’s not a counterexample to Kaplan’s original claim.12

12

Note that in this case the sentry could have said ‘I’m not here now, but back in my camp. But I surrender . . .’ Thus we have another counterexample to Kaplan’s claim that ‘I’m here now’ is a logical truth.

21

Recall our earlier discussion of the post-it note case. Let’s consider some variants that can be found in the literature. Say that, rather than put the note on her door herself Sarah calls her colleague Mike and tells him to put a note on her office door saying ‘I’m not here today’.13 When Mike writes the message on a post-it note and puts it on Sarah’s door, he is in effect receiving her dictation and placing Sarah’s message where she said to place it. He is part of the mechanism of communication. Like the king’s sentry, he may be the one who produces the sentence tokening, and this respect he’s a sort of speaker. But he’s not the agent of the context. Just as it would’ve been correct for Mary the secretary to say ‘I didn’t say that’ to Peter’s paper, it would correct for Mike to say of the post-it note ‘I didn’t say that [pointing to the note he put on Sarah’s door]’. So this post-it note case also doesn’t present a counterexample to Kaplan’s claim that ‘I’ always refers to the agent of the context. Say Mary didn’t just stick to dictating Peter’s paper, but put some of her own content in it too through her own initiative. Say it was this stuff that Mary added on her own that people found offensive. Now it would be Peter who could correctly complain ‘I didn’t say that!’ When one is part of the mechanism of communication of someone else’s utterance, one must stick to the message one is meant to deliver, and not add one’s own content. When one doesn’t stick to the message of the person who is supposed to be the one communicating, but delivers a message with content one added through one’s own initiative, in delivering this added content one is not delivering something the other says. The other could correctly say ‘I didn’t say that’ to this added content. When Mary takes the initiative of putting her own content into Peter’s paper, does she say those things? Do we have a mixed message so to speak: there’s the stuff Peter told her to say which she is faithfully transcribing Peter as saying, and that’s mixed in with stuff she put

13

Several of the variants on the post-it note case we will discuss are similar to cases discussed by Corazza, et al 2002, although our analysis of these cases will be diametrically opposed to theirs. 22

in the paper on her own which is stuff she is saying? We think not. This mixed message view doesn’t appreciate the element of deceit involved in Mary’s action. The paper has Peter’s name attached to it. We’re imagining that Mary doesn’t put in the middle of the paper something like ‘By the way – this is just me, Mary, the one to whom Peter dictated this paper talking here, not Peter – I think that . . .’14 Rather she is passing off the content she is adding as something Peter is saying. If Mary did make it clear that she was putting her own content in along with Peter’s in the manner just described – if she made it clear that it was she and not Peter who was talking – we would then have no problem accepting the mixed message analysis of the text. But we’re imagining that that isn’t what happened. Mary, we’re imagining, isn’t conveying Peter’s message, but neither is she telling the reader what she thinks. Rather, she’s putting words in Peter’s mouth. She is presenting Peter as saying something that he isn’t saying. So who is saying these things? We think, no one. Peter certainly isn’t. But neither is Mary. She isn’t asserting something when she puts these comments in; rather she’s trying to get Peter to assert them, so to speak. But she fails. For Peter to assert them he has to take the initiative and be the agent of the communication (although, of course, he need not be the speaker). In the case we began this paper with, the sentence didn't express the proposition that Hawking’s computer hates the addressee. In that case a mechanism of communication was malfunctioning, and as a result it articulates sentences that aren't being used by an agent to express propositions. When Mary tries to put words into Peter's mouth, she is such a malfunctioning mechanism, albeit a human one. In both cases of malfunctioning mechanisms, there's no agent of the context, and no proposition expressed. If Mike, seeing a queue of people outside Sarah’s door, uses his own initiative and puts a note on the door saying ‘Sarah is not here now’ he will express a proposition. Mike is
14

As occurs in Paul’s letter to Romans where, in the midst of transcribing Paul’s message, the scribe introduces himself and delivers his own message, mid-text (see Romans 16:22). 23

the agent of the context in which this proposition about Sarah is produced. And, if Mike sees the queue of people and puts a note on Sarah’s door that says ‘I am not here now’, he has certainly communicated something. He has produced a sentence. He has even communicated something potentially useful. But in this case there is no proposition expressed. When using the personal indexical ‘I’, a potential agent other than the referent can only be involved as part of the mechanism of communication. And the role of the mechanism is to carry out orders, not to have communicative intentions.15 To reiterate, we are not claiming that nothing is communicated in these ‘malfunction’ cases, simply that the literal proposition is left incomplete. One way to make sense of this is to add Perry’s notion of reflexive content to give a pragmatic explanation. The reflexive content retains the content associated with an indexical term when the fact(s) needed for the referential (literal) content are in some sense lacking. As Perry puts it, “If we do not fix the facts about who is the speaker of u, when the word ‘I’ occurs in the sentence in question, the issue of who the speaker of u is will be a constituent of the truth-conditions.” (2003: 380) So, taking the Hawking case as an example, the referential content is something like, ‘The speaker of the context hates the addressee of the context’. That this content is available would certainly account for the information communicated. In the analyses of the cases of Mary’s deceit and Hawking’s malfunctioning computer, there are sentences that are written down or uttered that fail to express a proposition because there is no agent of the context who is their utterer. The post-it note cases that have been the focus of recent theories should receive a similar treatment. Say Sarah doesn’t call Mike, or in any way tell him to put a sign on her door saying that she isn’t there.

15

We analyse a variant of the king-sentry case similarly. If the king sends the sentry to say ‘I surrender’ but the sentry in his role as spokesperson says instead ‘I will eat your liver tonight with fava beans’, the ‘I’ refers to neither the king nor the sentry, for the sentry as a mere means of production is trying to get the king to say something he doesn’t say. Nothing is actually said because there is no real agent of the context to say something. 24

Mike just sees a long queue outside Sarah’s door, and so puts a post-it note on her door saying ‘I’m not here today’. Sarah might be glad that he did that when she hears about it later, but it would still be correct for her to say ‘I didn’t say that’. Mike is trying to get Sarah to say something. Although we don’t deny that Mike is successful in getting the message across to the students in the queue that Sarah won’t be in today, in terms of the semantic analysis of the post-it note Mike left, it is exactly like the other two cases: no proposition is expressed. There is no agent of the context. It would be Sarah if anyone, but Sarah cannot be the agent since Mike acted completely on his own initiative. So, just as with the other cases, ‘I’ doesn’t refer and, to use Perry’s terminology, there is no referential content. There is a final kind of post-it note case from Corazza, et al 2002 we would like to consider; a case they claim shows that ‘I’ does not always refer to the agent of the context. Say Mike takes the post-it note off of Sarah’s door and then puts it on his own door. We think this case, semantically speaking, is like the case of someone who cannot talk but communicates with signs with sentences pre-written on them. It doesn’t matter whether he wrote the sentences on the signs or someone else did. The important thing is that he’s the one who is taking the initiative to use them to express his own thoughts. Thus when Mike takes the post-it note with ‘I’m not here today’ off of Sarah’s door and puts it on his own, he is the agent of the context, and ‘I’ refers to him. It’s semantically irrelevant that the sentence he is using was once on someone else’s door, and it’s also irrelevant whether it was he or Sarah who originally wrote the sentence down.16 In conclusion, we don’t think any of the above cases are most plausibly seen as counterexamples to Kaplan’s claim that ‘I’ always refers to the agent of the context. In some of them, we think it does refer to the agent, even though there may be someone else who

16

Thus we disagree with Corazza, et al 2002 that this case shows that ‘I’ doesn’t refer to the agent of the context. They wrongly are assuming that the agent is the one who originally wrote the note. 25

plays an essential role in producing the sentence in which it occurs. In others, we think the most plausible thing to say about them is that there is no agent of the context. In those cases the proposition is incomplete, as ‘I’ does not refer. In short, we see no reason to deny Kaplan’s immensely plausible claim that ‘I’ is a pure indexical that always refers to the agent of the context. We don’t think any of the cases discussed in this section require making any changes to Kaplan 1989’s semantics for indexicals. We do, however, recommend that one keep track of the different elements of utterance production presented above.

7

Conclusion

We have suggested that the ‘answering machine’ cases make up just a small amount of the instances in which we communicate via some non-standard method of communication. We argued against Kaplan’s claim that ‘I am here now’ is a logical truth, as have others. That it is a logical truth isn’t plausible once we take into account the distinction between the agent of the context, the mechanism via which the sentence tokening is to be produced, and the tokening of the sentence. Once we have these distinctions it is clear that ‘I am not here now’ can be true, as the agent is no longer tied to the location and/or time of the utterance. We have applied these distinctions to the ‘answering machine’ and ‘post-it note’ cases, and have argued that they can be handled without significant change to Kaplan’s semantic framework. The only change that is needed is that some of the contexts – the ones he called “improper contexts” – which he purged from his semantics, we have let back in. An interesting upshot of our treatment of the answering machine and post-it note cases is that it put us in a position to defend Kaplan’s plausible claim that ‘I’ is a pure indexical, a claim that has recently come under attack.17

17

We are grateful to Andreas Stokke. We wouldn’t have been able to write this paper without our discussions with him of the issues of this paper. We are grateful to audiences in Barcelona, Krakow, and St. Andrews for helpful feedback. We are indebted to feedback from 26

References Bach, K. 1998. Ambiguity. In Craig (ed.), Routledge Encyclopedia of Philosophy. New York: Routledge. Corazza, E., W. Fish, and J. Gorvett. 2002. Who is I? Philosophical Studies 107: 1-21. Corazza, E. 2004. On the alleged ambiguity of ‘now’ and ‘here’. Synthèse 138: 289-313. Egan, A. Forthcoming. Billboards, bombs and shotgun weddings. Synthèse. Kaplan, D. 1989. Demonstratives. In J. Almog, J. Perry and H. Wettstein (eds.), Themes From Kaplan. Oxford: Oxford University Press. _____. 1990. Words. Proceedings of the Aristotelian Society, Supplementary Volume 64: 93119. Perry, J. 2001. Reference and Reflexivity. Palo Alto, CA: CSLI Publications. _____. 2003. Predelli’s threatening note: contexts, utterances, and tokens in the philosophy of language. Journal of Pragmatics 35: 373-387. Predelli, S. 1998a. ‘I am not here now’. Analysis 58: 107-15. _____. 1998b. Utterance, Interpretation and the Logic of Indexicals. Mind and Language 13 (3): 400-14. _____. 2002. Intentions, indexicals and communication. Analysis 62: 310-16. _____. 2008. ‘I exist’: the meaning of ‘I’ and the logic of indexicals. American Philosophical Quarterly 45: 57-66. Romdenh-Romluc, K. 2002. Now the French are invading England! Analysis 62: 34-41. Sidelle, A. 1991. The answering machine paradox. Canadian Journal of Philosophy 21: 52539. Smith, Q. 1989. The multiple uses of indexicals. Synthèse 78: 167-91. Stevens, G. Forthcoming. Utterance at a distance. Philosophical Studies.

Ambròs Domingo, Manuel García-Carpintero, Patrick Greenough, Dan López de Sa, Crispin Wright, Elia Zardini, and an anonymous referee from Philosophical Studies. 27



Add Comment

To add a comment, you need to login or signup (you can signup without leaving this page)