Who is I?

EROS CORAZZA, WILLIAM FISH and JONATHAN GORVETT WHO IS I? (Received 22 February 2000; received in revised version 23 July 2001) ABSTRACT. Whilst it may seem strange to ask to whom “I” refers, we show that there are occasions when it is not always obvious. In demonstrating this we challenge Kaplan’s assumption that the utterer, agent and referent of “I” are always the same person. We begin by presenting what we regard to be the received view about indexical reference popularized by David Kaplan in his influential 1972 “Demonstratives” before going on, in section 2, to discuss Sidelle’s answering machine paradox which may be thought to threaten this view, and his deferred utterance method of resolving this puzzle. In section 3 we introduce a novel version of the answering machine paradox which suggests that, in certain cases, Kaplan’s identification of utterer, agent and referent of “I” breaks down. In the fourth section we go on to consider a recent revision of Kaplan’s picture by Predelli which appeals to the intentions of the utterer, before arguing that this picture is committed to unacceptable consequences and, therefore, should be avoided if possible. Finally, in section 5, we present a new revision of Kaplan’s account which retains much of the spirit of his original proposal whilst offering a intuitively acceptable way to explain all of the apparently problematic data. In doing so, we also show how this picture is able to explain the scenario which motivated Predelli’s account without appealing to speaker intentions. 1. THE CLASSICAL VIEW First, some background. Pure indexicals, “I”, “here”, “now”, “today”, etc. have linguistic meanings, which are, roughly, conventions telling us how to use them. Its dictionary entry, for example, informs us that “I” stands for the speaker or writer. What the dictionary gives us aims to capture the linguistic rule governing its use. Hence a competent speaker knows, roughly, that “I” is the first person pronoun which stands for the utterer (or writer). Kaplan (1977), following Reichenbach (1947), in his logic of demonstratives argues that we should countenance a distinction between the Philosophical Studies 107: 1–21, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 2 EROS CORAZZA ET AL. content of an indexical (its referent) and its character. The latter aims to capture the linguistic meaning of indexical expressions. On this account, the character of “I” is represented by a function that takes as argument the agent, and gives as value the referent. Kaplan argues that the logic of indexical terms forces the distinction between the utterer, the contextual parameters (agent, time, place and possible world) and the referents (contents). Kaplan argues that, as utterances are features of the world and hence occur in time, it is not possible to provide a semantic evaluation of utterances.1 To overcome this, we need instead to assess the abstract notion of a sentence-in-a-context, and to do this we need to endorse the notion of a context.2 The agent is, therefore, an essentially logical notion, a contextual parameter filling the argument of the character (qua function), and giving us the referent of the indexical, and, as such, is logically distinct from the notion of an utterer. However, Kaplan simply seems to assume that the utterer and the agent will always be identical and hence that the referent of “I” will always be the utterer. This account of indexical reference, therefore, appears to assume two identity statements; that the agent is the referent (from the character of “I”), and that the utterer is the agent.3 This account, therefore, can be represented in the following way: Determination of agent Determination of referent f = (The character of “I”) The utterer is the agent f: agent → referent While the utterer and referent are parts of the material world, the agent is a logical parameter, playing the role of taking us from the language to the world. When these two components are combined, we can see how Kaplan preserves the intuition that “I” refers to the speaker or writer. The agent-utterer identity ensures that, for every token of “I”, the contextual parameter of the agent is identified with the utterer, the individual who uses the token. The character of “I” then completes the task by returning the utterer (= the agent) as the referent of the token. The character of the other indexicals, “here” and “now”, can also be represented in the same way. In each case, the character of the WHO IS I? 3 indexical is a function from a contextual parameter (location in the former case and time in the latter) to the referent of the expression. An intuitively plausible model of these indexicals would also include a parallel of the agent-utterer identity we found in the case of “I”. For “here” we might identify the (contextual parameter of) location with the place of utterance, and for “now”, the (contextual parameter of) time with the time of utterance. In this way we can ensure that the character of these indexicals will always yield, as referent, the place and time of utterance. 2. ANSWERING MACHINES AND OTHER DEVICES In a number of recent papers, the Kaplanian framework for the indexicals “here” and “now” has been criticized. Its critics argue that the framework as it stands is unable to deal with a number of common cases. These include such examples as the outgoing messages on answering machines, messages written on post-it notes and the like. Sidelle (1991)4 introduces us to what he calls the Answering Machine Paradox and asks how, given the linguistic meaning of the indexicals, it is possible that utterances of the form “I am not here now”, which wear the form of a contradiction, can nevertheless be true. Statements of this type might be found on a telephone answering machine – hence the Answering Machine Paradox – and written on a post-it note and stuck on an office door when its incumbent is not in residence. The feelings of paradox arise because it seems to demonstrate that Kaplan is wrong when he says that an utterance of “I am here now” is “universally true. One needs only understand the meaning of [“I am here now”] to know that it cannot be uttered falsely” (Kaplan, 1977, p. 509). When played on an answering machine however, it is apparent that such an utterance might well turn out to be false. A productive way of approaching this so-called “paradox” is to view it as a puzzle concerning the referent of “now”. The problem arises from a tension between the pre-theoretic intuition that the utterance of an answering machine message occurs when the person records the message, Kaplan’s dictum that the referent of “now” is the time of the utterance and the fact that, for the answering 4 EROS CORAZZA ET AL. machine to serve any purpose, “now” must refer to the time at which the message is heard. Sidelle solves this with the proposal that an utterance can be deferred.5 On this account, the utterance takes place when the message is heard by the listener (the decoding time) not when it is recorded (the encoding time). In this way the Kaplanian view that the character of “now” directs us to the time of the utterance is preserved. The utterance occurs when someone makes a telephone call and the message is played. If the individual who recorded the message (the referent of “I”) is not at the place the machine is located (the referent of “here”) at the time the call is made (the referent of “now”) then that utterance of “I’m not here now” can, pace Kaplan, express a truth. We can see that the deferred utterance method of resolving this “paradox” leaves the Kaplanian framework untouched as the characters of the indexicals remain the same. “I” continues to refer to the agent, and “here” and “now” to the contextual parameters of location and time. Kaplan’s identities also remain, thereby ensuring that “I” ultimately refers to the utterer, “here” to the place of utterance, and “now” to the time of utterance. However, it is important not to assume that, in every situation where there is a delay between the production of the message and its being interpreted, an utterance always occurs at the decoding time. For example, if you receive a post card saying “I’m having fun here now”, we require “here” and “now” to refer to the place and time the post card has been written (the encoding time), not the time at which it is read (the decoding time). In this case, unlike the answering machine case, we require the relevant contextual parameters to be the time and place the message is produced. We suggest that our varying responses to utterances at a distance is something which should be explained by a satisfactory account of indexical reference. 3. THE PROBLEM OF “I” So Sidelle’s solution proceeds by leaving scope for there to be a temporal distance between the act of writing a note or recording a message, and the moment at which an utterance is made. However, we suggest that once it is allowed that utterances can occur at a temporal distance from the encoding act, situations can arise in WHO IS I? 5 which there is an intriguing and important question of just how the reference of “I” is determined. In the version of the puzzle just sketched, reference seems simple to secure. “I” refers to the utterer who, in the post-it note case, is intuitively seen to be the individual who writes the note and attaches it to her office door. However, there is a variation of the puzzle which brings the problems into focus. Joe is not in his office one day and Ben notices that a number of students keep approaching his door and knocking. They then stand around and look bemused for a while before leaving. Taking pity on these poor souls wasting their time, Ben decides to attach his “I am not here today” note to Joe’s door. The trick works; the students, instead of knocking and waiting, take one look at the note and then leave. This scenario differs from the previous one in that, in this case, Joe has no knowledge of, and has played no role in, deferring these utterances. At the moment one student looks at the note, and an utterance is made, the expressions “today” and “here” successfully refer to the day and place the note is read, but to what does “I” refer? As everything is working so well it would seem strange to deny that it refers in precisely the same way as it refers in the standard scenario – it refers to Joe. After all, this is certainly what the audience of the utterances, the students, take it to refer to. However, given that the character of “I” states that it refers to the agent, and, on Kaplan’s account, the agent is identical with the utterer, Joe must be the utterer if he is the referent. However, this seems to force the extremely counter-intuitive claim that one can make utterances one has had absolutely no role whatsoever in the production of. Joe didn’t write the note, he didn’t place it on his door, and he didn’t ask for it to be placed on his door. In fact, Joe has absolutely no idea that the note is on his door at all. For these reasons we suggest that it is implausible to suppose that Joe is the utterer in this situation. To highlight the unattractive consequences of this claim, imagine if, as a prank, Ben had instead attached a note that stated, “I think the Chancellor is a fool.” Would Joe still be prepared to accept this as his utterance? Note that we are not claiming that being unaware of being a referent is counter-intuitive – it happens to everyone all the time, even after death: How many times has Aristotle been referred to without him being aware of it? Rather, it is being deemed to be 6 EROS CORAZZA ET AL. the utterer of utterances that one has had absolutely no role in the production of that is counter-intuitive. If Joe is not the utterer, the only other candidate for this role would be Ben. So, following Kaplan’s identity of agent and utterer, the character of “I” would therefore yield Ben as the referent. Does “I” then, in this case, refer to Ben, the brains behind the note? We would suggest not. Firstly, the mere fact that Ben wrote the note doesn’t secure that reference. Joe could have, had he thought about it, attached Ben’s note to his door himself and thereby used it to make deferred utterances. Secondly, it would be decidedly odd to claim that “I” refers to Ben merely in virtue of his being the one who intended to use the note. After all Ben’s intention was for “I” to refer, not to himself, but to Joe. To claim that “I” refers to Ben regardless would force the implausible claim that a voice-over artist, recording pre-recorded messages for answer-phones, is the referent every time the sentence “I’m not in right now” is played. Another possibility would be to claim that “I” fails to refer, but again, working back through Kaplan’s account, this would force the claim that there is no contextual agent, and therefore no utterer. But how can there be an utterance if there is no utterer, and if there is no utterance, then how can we make sense of the idea that there is any proposition expressed whatsoever? Moreover, it seems odd to claim that, so long as the note states, “I am not here” the note fails to make any meaningful statement, but if the note were to say, “Joe is not here” it would succeed. Either note would work equally well and hence it seems poorly motivated to be forced to claim that the term does not refer at all. The only remaining option seems to be to claim that, although “I” refers to Joe, Joe is nevertheless not the utterer. However taking this line explicitly contradicts Kaplan’s plausible account of the linguistic meaning of “I” where “I” always refers to the utterer.6 Effectively, then, we appear to have four alternatives. If we want to retain Kaplan’s intuitively plausible account of the indexical “I” (in other words, we preserve the identities between on the one hand the agent and the utterer and, on the other, between the referent of “I” and the agent) we must either accept that Joe is the referent and can make utterances that he is unaware of, or that Ben is the utterer and therefore the referent. Neither of these options appear WHO IS I? 7 at all attractive. Alternatively, if we want to retain the intuitively plausible claim that, whilst Joe is the referent, he is nevertheless not the utterer, we are forced to reject one of Kaplan’s two identity statements. We must, to take this option, either reject Kaplan’s account of the character of “I”, a function from agent to referent, or reject the further identity of agent and utterer. We now turn to the discussion of this latter possibility. 4. THE INTENTIONALIST STANCE In two recent papers (1998a, 1998b), Stefano Predelli introduces a further variant of the answering machine case which calls into question Kaplan’s account of the indexical “now”. Predelli proposes to deal with these cases by denying the identity relation between the contextual parameter of time and the time of utterance. In effect, Predelli can be seen as opting, for the case of “now”, for the latter of our four options and for this reason, Predelli’s picture may well be applicable to the data we have highlighted. Predelli asks us to consider the following scenario. Before leaving home at 8.00 AM Joe writes the following note to his partner: “As you can see I am not at home now. Please meet me in six hours in my office” (where “in six hours” is short for “in six hours from now”). Joe, expecting his partner to return at 5.00 PM, intends for her to meet him at 11.00 PM. The reason Joe does not express his intent in a more explicit way, i.e. by saying “meet me at 11.00 PM”, need not concern us. If “now” in Joe’s note refers to the time at which it is read, then Joe’s partner will pick out the time at which she reads the note. If, as expected, Joe’s partner comes home at 5.00 PM and reads the note, everything will work as planned as, on Kaplan’s picture, “now” will pick out 5.00 PM and Joe’s partner will meet Joe in his office at 11.00 PM. Imagine, however, that Joe’s partner is unexpectedly delayed and doesn’t return home until 7.00 PM. According to Kaplan’s picture, “now” in Joe’s note should pick out the time of utterance – 7.00 PM. But Predelli argues that Joe’s partner, being aware that she was expected home at 5.00 PM, will not meet Joe at 1.00 AM the following day, but will meet him, as Joe expects, at 11.00 PM. Predelli’s moral is that “now” in Joe’s note does not always refer 8 EROS CORAZZA ET AL. to the time Joe’s message is read, it rather picks out the time Joe intended to be picked out. Predelli’s contention is that, in order to explain this variant of the answering machine “paradox”, we have to appeal to the notion of intentional contexts. Predelli’s contention is that “an adequate explanation [of indexical reference] ought to take into consideration the co-ordinates intended . . . as relevant for the[ir] semantic evaluation” (1988a, p. 112, our emphasis). On Predelli’s account, therefore, “now” does not always refer to the time of utterance, for “now” can be sensitive to an intentional context which may well differ from the context of utterance. However, in taking this line, Predelli manages to retain half of Kaplan’s account. Predelli rejects the claim that the contextual parameter of time is always identical with the time of utterance, claiming instead that this contextual parameter may well be an intentionally specified parameter, but retains Kaplan’s account of the character of “now”. Even on Predelli’s account, the character of “now” is represented by a function that takes the (now intentionally determined) contextual parameter of time as argument and yields the referent. When applied to “I”, Predelli’s picture promises to solve our puzzle by rejecting the role of the utterer in determining the agent, instead allowing for the context, and hence the contextual agent, to be determined intentionally.7 If Ben creates an intentional context with Joe as the agent, the character of “I” will, as before, yield Joe as the referent. This way, Predelli would be able to retain Kaplan’s account of the character of “I” (a function from agent to referent), claiming that Joe is the (intentionally determined) agent and hence referent. The similarities and differences between Kaplan’s and Predelli’s accounts are illustrated in the table below. Kaplan Predelli Determination of agent Determination of referent The utterer is the agent Intention determines agent f: agent → referent f: agent → referent Although at first sight Predelli’s picture might appear convincing, if we return to more normal examples, we can see that Predelli appears to be committed to an implausible consequence. If we allow Predelli’s appeal to an intentional agent, we must accept that “I” refers to Joe solely on the grounds that Ben intends it so to refer. WHO IS I? 9 However, if we accept that Ben, purely in virtue of his having the intention to do so, can use “I” to refer to Joe, why can he not use “I” to refer to pretty much anybody? Moreover, if Ben had, instead of affixing his note to Joe’s door, simply leant out of his own office and said to the students, “I am not here today,” (intending thereby to use “I” to refer to Joe) surely he would have failed in this intention and merely referred to himself. To claim otherwise is a Humpty Dumpty picture and we suggest that we should not be committed to such a thesis for indexicals. Similar reasoning can be used to enervate any claim of reference being secured by simulation, where Ben succeeds in using “I” to refer to Joe as long as Ben is playing the part of, or pretending to be, Joe. It appears that, for this type of simulation to work, the participation of the audience must bear some of the weight – they would have to be, in some sense, in the know. If the audience is not required to be aware of the simulation, the resultant picture appears committed to claiming that all that is required to secure reference is the intention of the speaker (in this case, to be playing a part) and, as such, suffers from the same kind of implausibility as Predelli’s picture. However, to go one step further and claim that reference is determined solely by the beliefs/intentions of the audience is, in its own way, as implausible as saying that reference depends purely on Ben’s intentions. If Ben refers to himself saying, “I am tired,” and his class believes him to be playing the part of Aristotle, “I” doesn’t thereby refer to Aristotle, but to Ben – the class have just made a mistake. If the only plausible version of a simulation picture requires some level of audience participation, it is of no use in solving our puzzle as, in this case, the audience are entirely unaware of any involvement on Ben’s part. Of course, there may be ways in which Intentionalists such as Predelli may attempt to temper their theory so as to avoid these implausible consequences and we shall consider (and reject) one such possible move below in section 6. For the time being, however, we shall move on to show instead how to solve the puzzle outlined above without relying on a potentially problematic appeal to speaker intentions. In explaining this thesis, we shall also show just why Predelli’s appeal to intentions in the note-leaving example looks plausible. Finally, it is worth pointing out that, whilst Predelli’s 10 EROS CORAZZA ET AL. account might be faithful to (some of) the letter of Kaplan’s picture, it is disloyal to its general spirit.8 Kaplan’s overall approach to language is consumerist whereby words and their meanings are mind independent. We are consumers of words and therefore of their meanings. We have an intention in so far as we choose which words we use. However, Kaplan does not hold that these intentions determine the meanings of the words, but that the meanings are independent of our intentions.9 The picture we shall propose below is not only faithful to the letter but also to the spirit of Kaplan’s original proposal. 5. THE (UN)CONVENTIONAL STANCE Let us recap. Kaplan’s account of the indexical “I” has two aspects. The first component is the claim that the character of “I” can be represented as a function which takes as input the contextual parameter, the agent, and returns as value the referent. The second element is the claim that, for any given utterance of “I”, the contextual agent will be identical with the utterer. As we saw in section 3, however, this account is not flexible enough to cope with certain possible uses of “I”. Kaplan’s original account suffers from the rigidity of the claim that, in every situation, the agent, and hence the referent, will be identical with the utterer. Predelli’s account offers to solve this problem by denying agentutterer identity and allowing us to determine the agent intentionally. However, as we saw in section 4, replacing the agent-utterer identity with an intentionally determined agent yields an overall picture which is not restrictive enough. So Predelli goes too far in the opposite direction. In a nutshell, Kaplan’s original account is not flexible enough to deal with certain problem cases, whereas Predelli’s position is so unrestricted it yields counter-intuitive consequences. What we require is a position which avoids these extremes and is flexible enough to account for the problem cases, but is not so loose that it becomes implausible. We propose to expand Kaplan’s original framework to account for the variegated scenarios in which indexicals can be successfully used. Our position is similar to Predelli’s in as much as we retain Kaplan’s account of the character of “I” as a function from agent 11 WHO IS I? to referent, and deal with the problem cases by rejecting the agent– utterer identity. We differ from Predelli, however, in that we reject his Intentionalist approach to semantics, and replace the agent– utterer identity with a more objective method of determining the agent. This way our picture is much closer to the spirit, as well as the letter, of Kaplan’s original account. Our proposal is that, for any use of the personal indexical, the contextual parameter of the agent is conventionally given – given by the social or conventional setting in which the utterance takes place.10 For instance, with “now”, the setting or context in which it is used changes the time that the term refers to: if “now” is heard on an answering machine, we take the relevant time to be the time at which it is heard, and we arrive at the referent accordingly. In contrast, if we read “now” on a postcard (“the weather is beautiful now”), the change in context or setting determines that the message refers differently. In the case of a postcard, unlike the case of an answering machine message, we take the relevant time to be the time at which the words were written. Hence we get a different referent in each case. This notion of a setting is part of the context of the utterance and as such plays a role in determining the contextual parameters of agent, location and time which are then utilized as argument for the character of the indexicals. The relevant factors which Kaplan allows to influence the contextual parameters of agent, location and time are very narrow. In determining these parameters he is only concerned with the who, where and when of the utterance, and does not allow anything else to influence the issues. In contrast, the notion of a setting of an utterance allows us to cast our net much more widely and include, among other things, the language being spoken, the physical environment and other factors as relevant to determining our contextual parameters. In the terms of our explanation of Kaplan above, our account differs from Kaplan and Predelli as follows: Kaplan Predelli The Conventional Account Determination of agent Determination of referent The utterer is the agent Intention determines agent Convention determines agent f: agent → referent f: agent → referent f: agent → referent 12 EROS CORAZZA ET AL. This position can be illustrated with Perry’s distinction between semantic and pre-semantic uses of context. As Perry claims (1997, p. 593) “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 pre-semantic 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.” We would rather say that, in the case of indexicals, context is used both semantically (narrow context) and pre-semantically (broad context).11 On our account, it is a matter of the setting or broad context whether a given word, for instance, is used with such or such a conventional meaning. In short, features such as speaking English, belonging to a given community, hearing an answering machine, sarcastically imitating someone, acting in a piece of theater, etc. are not part of what one says and usually aims to communicate. These features are better understood as aspects of the setting on which the linguistic interchange takes place. The fact that one speaks German, for instance, when one says ‘Ich’ has no semantic role – it is not part of what one says. Our German speaker merely refers to herself using the first person pronoun ‘Ich’. If she were to speak English she would have used ‘I’ to refer to herself, for if she had uttered ‘Ich’ she would probably have expressed disgust. Thus facts about the setting such as these do not determine what the terms used mean, more which terms are used. In this example, the mere fact that one speaks German is the setting upon which the linguistic interchange takes place. This setting is conventionally governed. If we bring back these points to our initial worries, we can argue that our social practice with regard to the way we use answering machines and post-it notes, is conventionally ruled in such a way that it allows someone to use a token of the first person pronoun produced by someone else to refer to herself. This convention is illustrated by the fact that we do not have any difficulty in coping with someone’s pre-recorded answerphone message. It is a part of our conventions determining the use of answer-phones that someone can buy a tape of a Woody Allen imitator reading a message and use it on their answer-phone to refer WHO IS I? 13 to themselves. Without such settings and conventions we would be unable to successfully use and manipulate answering machines and other similar devices. For reasons such as this, we should see that the context or setting of a linguistic interchange plays a role in determining how the agent is determined. In other words, the agent of “I”, like the relevant contextual parameters such as the time and place, is best understood to be the conventionally determined agent, and the agent determined by convention may well, as we have seen, be distinct from either the utterer or the producer of the token of “I”. We suggest that this attention to conventions goes so far as to explain how the deferred utterance method of resolving the original answering machine “paradox” gets off the ground. It succeeds because we are aware of the conventions governing the use of answering machines and the fact that the purpose of such devices is to inform the caller of the state of affairs at the time the call is made. Paying attention to conventions in this way also enables us to explain why we don’t treat other superficially similar examples of utterances at a distance (such as postcards) in the same way. The conventions governing the use of postcards have it that their usual purpose is to inform the recipient of the state of affairs at the time of writing. It is only by paying attention to the differing conventional roles of these devices that we are able to get these explanations off the ground. To explain our example in this framework; when Ben sticks his own post it note saying “I am not here today” to Joe’s office door with the intent to inform the eventual readers that the usual occupier of the office is not in, Ben can be seen to be exploiting the conventional setting of using notices in this way. The convention is that “I” on a notice on someone’s office door refers to the office’s usual incumbent. However, the example we formulated illustrates that, on occasions when the office’s occupant and the utterer come apart, the power of our linguistic conventions serves to ensure that “I” nevertheless refers to the inhabitant of the office. Moreover, even if Ben intended the note to refer to himself (say he left in a hurry and attached the note to the wrong door by mistake), the conventions would override the intentions thus ensuring that the note still referred to Joe.12 It is for this simple reason that it is better to under- 14 EROS CORAZZA ET AL. stand the agent qua contextual parameter to which the character of “I” directs us to be the conventional agent. It goes without saying, of course, that in most of the cases the contextual agent is the writer and or speaker. In certain situations, however, it may not be the same. 6. ISSUES ARISING There are, of course, a number of issues arising from our treatment of these cases which need to be addressed. We shall begin with a general concern. Given that we have shown Kaplan’s account to be flawed, why should we retain any of it? Why not instead just reject the entire picture. We want to retain certain aspects of Kaplan’s theory for a number of reasons. Firstly because it is a powerful account which enables us to explain a number of facets of indexical reference. Secondly, as Kaplan says, the logic of demonstrative expressions requires us to introduce the notion of a contextual agent. Thirdly, and most importantly, in retaining the idea that every token of “I” has the same character, we can respond to the objection that, in different scenarios (pace Quentin Smith), we are not using the same terms, but linguistically distinct homonyms. On the contrary, every utterance of “I” regardless of how employed, is an utterance of the same type: in each case the character of the term remains the same. Moreover, when extended to “here” and “now”, this treatment enables us to elegantly explain the fact that these terms seem to have two distinct uses – as pure indexicals and as demonstratives. We can give an account of these facts as follows. In every use of the word “here”, the character of the indexical is a function from the contextual parameter of location to the referent. It is just that different conventional settings determine the location differently. Most of the time, the location is, as Kaplan suggests, identical with the place of utterance. However, in some of its conventional uses this does not hold. Take indicating places on maps as an example. In this convention-governed use of “here”, the character of the indexical remains the same – it is a function which returns the contextual parameter of location as referent. What changes is the way in which the contextual parameter of location is determined – in this case it is determined by the additional demonstration. In this way we can WHO IS I? 15 give a powerful and elegant account of the way in which certain indexicals seem to have these two distinct uses. It remains to give an account of how our position addresses some of the other problematic cases we have considered. As we have accepted that, in our case, Joe is the referent of “I”, then how should we explain the scenario where the note Ben attaches to Joe’s door states “I think the Chancellor is a fool”? We might diffuse any concerns about this case in two ways. The first would be to remember that, although Joe is the referent of “I”, he is not, thereby, the utterer. This response would enable us to claim that this scenario is no different for Joe than if Ben had attached a note saying “Joe thinks the Chancellor is a fool”. In both cases the proposition expressed is the same, and in neither case does Joe have any responsibility for the expression of that proposition. Alternatively we might respond that joke notes like this constitute a conventional setting all of their own. For example, if we see someone wearing a sign on their back which states “kick me”, we do not assume that they desire a blow with the foot as we do not, in this case, assume that the purported referent of “me” is responsible for the sign. Reading such a sign is not akin to hearing someone speak and assuming that they are telling the truth (as in normal communicative interaction), it is more like hearing somebody tell a joke or a tall story – normal interpretations are withheld. Two related issues which need addressing are the following. Firstly there is the thought, following our rejection of the possibility that “I” might fail to refer, that if an utterance is made, then there must be an utterer. If this is not Joe then who is it? There is also the related concern of how to cope if the note attached to Joe’s door found its way there by random means (say Ben threw it away but a gust of wind picked it up from his bin and deposited it on Joe’s door). In response to the first concern, we might follow our intuitions as regards Ben’s using a note which states “Joe is not here today”. In this scenario, there is no basis for denying that Ben is the utterer and Joe the referent. In the “I am not here today” case, we suggest that the same considerations should apply. Ben is the utterer (but not the referent) and Joe is the referent (but not the utterer). To cope with the second anxiety we should remember that, in order for a proposition to be expressed, there has to be a commu- 16 EROS CORAZZA ET AL. nicative intention which is sufficient for utterances to occur. In our original scenario the communicative intention was on Ben’s part, whereas in the random variant, there was no communicative intention and hence no utterances occur. It should be noted, however, that asserting the necessity of a communicative intention is not equivalent to the Intentionalism we saw in Predelli. Actually we have to recognize two distinct kinds of intention – the communicative (consumer) intention and the intention a speaker has to identify/speak about a given item (the individuative intention).13 We reject, pace Predelli, that there is any semantic role for the second type of intention. We accept, on the other hand, the necessity of the first type of intention. After all, if one is to play a (language) game, one ought to have the intention to follow the rules of that game. Finally, we need to explain how our position accounts for Predelli’s note example where Joe’s partner, reading the note at 7 PM takes “now” to refer to 5 PM. The convention exploited in this case is that of leaving notes in a common context. It is only because Joe’s partner is aware that Joe would have expected her to come home at 5 PM (either because of prior arrangements or because that is the time she usually comes home) that she is able to grasp that Joe must have intended “now” to refer to 5 PM. If there were no social convention between the couple regarding the time of her arrival home (either normally, or on that specific occasion), then the convention would decree that some other time be used as the parameter, regardless of the intentions of the note writer. If the usual time of Joe’s partner’s homecoming is 6 PM, and if she had stipulated that tonight she would, as usual, be home at 6, then surely “now” would have referred to 6 PM regardless of Joe’s intentions. The only reason communication successfully takes place is because Joe’s partner is aware that she usually returns home at a certain time and that this is the time that Joe would have expected her to return home. It might be argued that all this shows is that these facts about the situation entail that Joe could not have intended to use “now” to refer to 5 PM. This proposal would amount to adding a qualification to the Intentionalist proposal such that, whilst the listener is supposed to look for the intended referent, this cannot work unless it is possible (in the context) for the decoder to work out who the referent is.14 WHO IS I? 17 However, once we accept that there is a limit to what we can intend to use a term to refer to, we suggest that the Intentionalist picture comes to assume something like the very account we have been proposing. When faced with this kind of situation we need to establish what it is about a given scenario that makes it the case that it is reasonable, or plausible, to expect the decoder to be able to identify the referent of a given indexical. For example, what makes it the case that Ben’s leaving the note on Joe’s door makes it plausible to expect the decoder to come up with Joe as the intended referent, whereas leaving it on the toilet door does not? Surely what Ben is doing here is exploiting the very type of convention that we are invoking. So, in as much as Ben’s intentions are involved in assigning a referent to “I” in our problem case, it is his intention to exploit the convention of leaving notes on doors that is important rather than his intention that “I” should refer to Joe. When seen in this light, however, it is easy to see how the communicative intention to exploit a convention might have been mistaken for an individuative intention to assign a referent.15 So, we can see then that it is of no account that Joe intended “now” to refer to 5 PM; if Joe had intended for “now” to refer to 11 PM (for whatever reason) it would still only have referred to 5 PM given the context in which the note was left. Joe’s intentions only come in to the picture in as much as he leaves the note with a communicative intention to exploit a certain convention. We do not, however, have to appeal to the speaker and/or audience’s individuative intention to account for the reference of indexical expressions. All we need to appeal to are conventionally given contextual parameters. 7. CONCLUSION To sum up, Kaplan’s story about the character of “I” rests on his plausible identification of the utterer, the agent and the referent. So the character of “I” takes as argument the agent/utterer/referent and gives as value the agent/utterer/referent. Our story, however, shows that we cannot always identify the utterer with the referent in this way. 18 EROS CORAZZA ET AL. Predelli’s account of intentional contexts appears to offer a way of biting the bullet and claiming that the referent/agent of “I” can be picked out by an utterance that she never dreamed of. Predelli effects this by retaining Kaplan’s account of the character of “I” whilst replacing agent-utterer identity with the claim that the agent is determined intentionally. However we have argued that, if this picture is accepted, we must also accept the implausible consequence that one can use “I” to refer to whatever one wants to. Our story, on the other hand, powerfully and elegantly accommodates the data we have been discussing, by introducing the notion of the conventionally determined agent. This position rests on the plausible distinction between broad context (the conventional setting upon which the linguistic act takes place), and narrow context, i.e. the parameters (time, place and agent) qua argument and value of the character of indexicals. ACKNOWLEDGEMENTS We would like to thank Michael Clark, Stefano Predelli and especially the anonymous referee for Philosophical Studies for their comments and suggestions for improvements on earlier versions of this paper. NOTES 1 Kaplan illustrates this concern in the following passages. “Utterances take time, and utterances of distinct sentences cannot be simultaneous (i.e., in the same context). But in order to develop a logic of demonstratives we must be able to evaluate several premises and a conclusion all in the same context. We do not want arguments involving indexicals to become valid simply because there is no possible context in which all the premises are uttered, and thus no possible context in which all are uttered truthfully” (Kaplan, 1977, p. 522). “Utterances take time, and are produced one at a time; this will not do for the analysis of validity. By the time an agent finished uttering a very, very long true premise and began uttering the conclusion, the premise may have gone false. Thus even the most trivial of inferences, P therefore P, may appear invalid” (Kaplan, 1989, p. 584). 2 “[Context] is a package of whatever parameters are needed to determine the referent . . . the context supplies the time and place parameters that determine content for the indexicals ‘now’ and ‘here’ ” (Kaplan, 1989, p. 591). WHO IS I? 3 19 The latter of these assumptions (of agent-utterer identity) explains why Kaplan sometimes appears to treat them as co-extensive terms. For example, “[‘I’] refers to the speaker or writer of the relevant occurrence of the word ‘I’, that is, the agent of the context” (1977, p. 505) or “In each of its utterances, ‘I’ refers to the person who utters it” (1977, p. 520). 4 Although Sidelle is considered to be the locus classicus of this problem, precursors of this problem can also be seen in Kaplan (1977, p. 491ff.) and Vision (1985). 5 See Sidelle (1991). As far as we know Sidelle was the first to propose the notion of deferred utterances as a way of resolving the answering machine paradox. We believe that if Sidelle’s solution is satisfactory, it should apply to the puzzle in the post-it note form as well. 6 Quentin Smith introduces a number of other examples which, he suggests, are occasions when “I” can be used to refer to someone other than the utterer (1989, pp. 182–186). Whilst not necessarily wanting to endorse Smith’s examples, it is instructive to consider his positive proposal. In the light of the notion that “I” can refer to someone other than the utterer, Smith concludes that there are “varying reference-fixing rules that govern . . . indexicals” (p. 170). This position, however, commits one to the thesis that indexicals are ambiguous. A position avoiding this conclusion should be preferred for it is, if nothing else, more economical. As Kripke reminds us: “It is very much the lazy man’s approach in philosophy to posit ambiguity when in trouble. If we face a putative counterexample to our favorite philosophical thesis, it is always open to us to protest that some key term is being used in a special sense, different from its use in the thesis. We may be right, but the ease of the move should counsel a policy of caution; Do not posit ambiguity unless you are really forced to, unless there are really compelling theoretical or intuitive grounds to suppose that an ambiguity really is present” (1977, p. 19). 7 “In all these examples, the notion of utterance and the idea of a context of utterance do not play any semantically interesting role, and the correct results are obtained by anchoring indexical expressions to the intended context of interpretation” (Predelli, 1998a, p. 114). 8 “Suppose, without turning and looking I point to the place on my wall which has long been occupied by a picture of Rudolf Carnap and I say, “[T]hat is a picture of one of the greatest philosophers of the twentieth century”. But unbeknownst to me someone has replaced my picture of Carnap with one of Spiro Agnew. I think it would simply be wrong to argue an ambiguity in the demonstration, so great that it can be bent to my intended demonstratum. I have said of a picture of Spiro Agnew that it pictures one of the greatest philosophers of the twentieth century” (Kaplan, 1978, p. 30). 9 “Words come to us prepacked with a semantic value. If we are to use those words, the words we have received, the words of our linguistic community, then we must defer to their meanings” (Kaplan, 1989, p. 602). 10 This picture might be understood along the lines of something like a Wittgensteinian language game. Our linguistic furniture seems to function in 20 EROS CORAZZA ET AL. different ways dependent upon the social context of a given linguistic act. We might interpret this as saying that our linguistic pieces play a different role when we play different language games with them. 11 Since Perry’s notion of pre-semantic use of context, as we understand it, plays a role in determining the meaning of the terms used it may be misleading to oppose it to a semantic use of context. We thus think it is more accurate to adopt the distinction we are proposing between narrow and broad context. 12 It is worth noting that the position we are defending also allows us to dismiss cases such as the ones when a post-it note is found in a dustbin or on a corridor floor. As the normal function of a post-it note stating “I’m not in now” is to be put on a door, the setting (i.e. a post-it note in a dustbin), prevents us from taking the post-it note as meaning something like the agent is not in the dustbin. 13 “To use language as language, to express something, requires an intentional act but the intention that is required involves the typical consumer’s attitude of compliance, not the producer’s assertiveness” (Kaplan, 1989, p. 602). 14 We would like to thank the anonymous referee for bringing this example to our attention. 15 This also explains why we cannot do away with this multiplicity of conventions in favor of a single, underlying interpretive rule which all these conventions are merely instances of. The kind of rule which might play this role would be something like the qualified Intentionalist rule above. However, as we have seen, this approach only gets off the ground in as much as it assumes the existence of certain constraints which look very much like the conventions it is proposed to replace. REFERENCES Almog, J., Perry, J. and Wettstein, H. (eds.) (1989): Themes from Kaplan, Oxford: Oxford University Press. French, P.A., Uehling T.E. and Wettstein, H.K. (eds.) (1977): Contemporary Perspectives in the Philosophy of Language, Minneapolis: University of Minnesota Press. Hale, R. and Wright, C. (eds.), Companion to the Philosophy of Language, Oxford: B. Blackwell. Kaplan, D. (1977): ‘Demonstratives’, in J. Almog, J. Perry and H. Wettstein (eds.) (1989), Themes from Kaplan, Oxford: Oxford University Press, pp. 481–563. Kaplan, D. (1978): ‘Dthat’, in P. Yourgrau (ed.) (1990), Demonstratives, Oxford: Oxford University Press, pp. 11–33. Kaplan, D. (1989): ‘Afterthoughts’, in J. Almog, J. Perry and H. Wettstein (eds.), Themes from Kaplan, Oxford: Oxford University Press, pp. 565–614. Kripke, S. (1977): ‘Speaker’s Reference and Semantics Reference’, in P. French, T. Uehling and H.K. Wettstein (eds.), Contemporary Perspectives in the Philosophy of Language, Minneapolis: University of Minnesota Press, pp. 6–27. WHO IS I? 21 Perry, J. (1997): ‘Indexicals and Demonstratives’, in R. Hale and C. Wright (eds.), Companion to the Philosophy of Language, Oxord: B. Blackwell, pp. 586–612. Predelli, S. (1998a): ‘I am not Here Now’, Analysis 58(2), 107–115. Predelli, S. (1998b): ‘Utterance, Interpretation, and the Logic of Indexicals’, Mind and Language 13(3), 400–414. Reichenbach, H. (1947): Elements of Symbolic Logics, New York: Free Press, #50, pp. 284–286. Sidelle, A. (1991): ‘The Answering Machine Paradox’, Canadian Journal of Philosophy 81(4), 525–539. Smith, Q. (1989): ‘The Multiple Uses of Indexicals’, Synthese 78, 167–191. Vision, G. (1985): ‘I am Here Now’, Analysis 45(4), 198–199. Yourgrau, P. (ed.) (1990): Demonstratives, Oxford: Oxford University Press. Department of Philosophy The University of Nottingham University Park Nottingham, NG7 2RD, UK E-mail: eros.corazza@nottingham.ac.uk william.fish@nottingham.ac.uk apxjg@nottingham.ac.uk