1.Introduction

The debate over the referents of indexicals has relied heavily on the use of counterexamples. One theorist presents a rule for identifying the referent of an indexical in context, a second presents a counterexample in which that rule seems to take us to the wrong referent and presents an alternative rule, a third presents an counterexample to that rule, and so on. This paper will argue that reliance on counterexamples faces a significant problem. Counterexamples invite us to agree that some theory misidentifies the referent of an indexical as used in a particular context. Even when we are inclined to agree, however, we may well be mistaken.

The following section will present Kaplan’s semantics for indexicals. Section 3 then introduces Mount’s notion of perspective and shows how it can be deployed to respond to counterexamples, illustrating with several cases from the literature. To make progress by the method of counterexamples, we need some way to tell when this response is appropriate. Section 4 will consider several tests for semantic content and argue that none of them can be used to tell when appeals to perspective are admissible. The problem, therefore, remains open. In lieu of a replacement test, Section 5 will then consider how we should respond if we become convinced that there is no way to identify the referents of indexicals in difficult cases, and so no way to identify the characters that determine their referents. I offer some reasons to prefer a nihilist response over epistemicism or pluralism.

2.Kaplan’s semantics for indexicals

Kaplan (1989a)Kaplan, David 1989a “Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstrative and Other Indexicals.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 481–563. New York: Oxford University Press. associates linguistic expressions with two levels of meaning. First, their content or referent. The referent of an indexical like ‘I’, ‘here’, or ‘now’ depends on the context in which it is used. When I use ‘I’, it refers to me; when you use ‘I’, it refers to you. Second, their character. The character of an expression determines its content relative to context. It is (roughly) the character of ‘I’ for example that it refers to the speaker.11.Why ‘roughly’? First, Kaplan’s semantics applies also to inscriptions and other modes of communication. Second, even in cases of spoken language, the one speaking might not be the one communicating, as when a message is being relayed through an interpreter. For further discussion, see Dodd and Sweeney (2010Dodd, Dylan, and Paula Sweeney 2010 “Indexicals and Utterance Production.” Philosophical Studies 150: 331–348. , 344–347). The notions of character and content are not restricted to indexicals but the character of a non-indexical expression determines the same content in all contexts. Nor are content and character restricted to sub-sentential terms. The content of a sentence, for example, is the proposition it semantically expresses or, in Kaplan’s terminology, what is said by uttering the sentence.

Kaplan represents the character of an indexical as a function from context to content. Contexts are represented as ordered tuples of contextual parameters 〈a,l,t,w〉, where a is an agent (roughly, the speaker), l is the location of their utterance, t is the time of their utterance, and w is the world of their utterance. The parameters a, l, and t function, respectively, as the referents of ‘I’, ‘here’, and ‘now’. w may function as the referent for modal indexicals like ‘actually’ but it also restricts the admissible values of a, l, and t. For Kaplan, indexicals have referents only relative to proper contexts, where a is in l at t in w.22.Kaplan (1989aKaplan, David 1989a “Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstrative and Other Indexicals.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 481–563. New York: Oxford University Press., 528) also talks about including a sequence of demonstrations to determine the referents of demonstratives like ‘this’ and ‘that’ and, in his Afterthoughts, of contexts involving “whatever parameters are needed” to determine the content of indexicals (Kaplan 1989bKaplan, David 1989b “Afterthoughts.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 565–614. New York: Oxford University Press., 591). This restriction is imposed to accommodate the intuition that ‘I am here now’ is always true when uttered.

We can extract two related notions of character from Kaplan’s work. These different notions of character imply different individuation conditions. On the first notion, a character simply assigns referents relative to contexts. The character of ‘now’, for example, can be represented by a function from a context c to the time of the c. If that’s all a character is, the character of an indexical is not sufficient to determine the content of its utterances. The world is full of contexts and we need some further rule that associates utterances with contexts, which can then determine content as a function of character. For Kaplan, the relevant context is the context in which the indexical is uttered. Others disagree with Kaplan’s selection of this context. Predelli (1998aPredelli, Stefano 1998a “ ‘I Am Not Here Now’.” Analysis 58: 107–115. , 1998bPredelli, Stefano 1998b “Utterance, Interpretation and the Logic of Indexicals.” Mind & Language 13: 400–414. ), for example, argues that content should be determined relative to the context intended by the speaker. Despite disagreeing about the contexts against which indexicals should be evaluated, he takes himself to retain the “customary characters” for indexicals, e.g. ‘now’ as represented by a function from c to the time of c.33.Note, however, that Kaplan’s restriction to proper contexts is built into Kaplan’s character assignments, even on this conception of character. Predelli recognises this in a later paper (Predelli 2011Predelli, Stefano 2011 “I Am Still Not Here Now.” Erkenntnis 74: 289–303. ), saying that his view of indexicals must, in virtue of denying the restriction to proper contexts, “inevitably disagree with Kaplan’s assignment of characters”.

As I shall use the term ‘character’, however, the character of an indexical is whatever is needed to fill the gap between an utterance and its content. On this alternative notion, the character of an indexical does not just pick out a parameter from a given context: it picks out the relevant context. In disagreeing with Kaplan about the contexts against which characters should be assessed, Predelli disagrees with Kaplan’s assignment of characters. I take it that both notions are available in Kaplan’s work but I take this second notion to make the best sense of Kaplan’s (1989bKaplan, David 1989b “Afterthoughts.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 565–614. New York: Oxford University Press., 598) claim that character is “a function from possible contexts of use”. Here, Kaplan seems to take characters to identify, not only a parameter from any given context, but the context against which an utterance of an indexical should be evaluated.44.Of course, any readers who disagree about this interpretation of Kaplan should simply take me to be using ‘character’ in a slightly different sense from Kaplan’s original.

The distinction between content and character has been widely accepted, but there is significant disagreement as to the correct analysis of indexicals’ characters. Arguments for a particular analysis of character often rely on counterexamples designed to show that alternative analyses associate indexicals with the wrong referents.55.See, for example, Åkerman (2015Åkerman, Jonas 2015 “The Communication Desideratum and Theories of Indexical Reference.” Mind & Language 30: 474–99. , 2017Åkerman, Jonas 2017 “Indexicals and Reference-Shifting: Towards a Pragmatic Approach.” Philosophy and Phenomenological Research 95: 117–152. ), Cohen (2013)Cohen, Jonathan 2013 “Indexicality and the Puzzle of the Answering Machine.” Journal of Philosophy 110: 5–32. , Corazza et al. (2002)Corazza, Eros, Willaim Fish, and Jonathan Gorvett 2002 “Who is I?” Philosophical Studies 107: 1–21. , Dodd and Sweeney (2010)Dodd, Dylan, and Paula Sweeney 2010 “Indexicals and Utterance Production.” Philosophical Studies 150: 331–348. , Egan (2009)Egan, Andy 2009 “Billboards, Bombs and Shotgun Weddings.” Synthese 166: 251–279. , Gómez-Torrente (2020)Gómez-Torrente, M. 2020 Roads to Reference: An Essay on Reference Fixing in Natural Language. Oxford: Oxford University Press., Michaelson (2014)Michaelson, Eliot 2014 “Shifty Characters.” Philosophical Studies 167: 519–540. , Michaelson and Cohen (2013)Michaelson, Eliot, and Jonathan Cohen 2013 “Indexicality and the Paradox of the Answering Machine.” Philosophy Compass 8: 580–592. , Montminy (2010)Montminy, Martin 2010 “Context and Communication: A Defense of Intentionalism.” Journal of Pragmatics 42: 2910–2918. , Mount (2008Mount, Alison 2008 “The Impurity of “Pure” Indexicals.” Philosophical Studies 138: 193–209. , 2015Mount, Alison 2015 “Character, Impropriety, and Success: A Unified Account of Indexicals.” Mind & Language 20: 1–21. ), Nunberg (1993)Nunberg, Geoffrey 1993 “Indexicality and Deixis.” Linguistics and Philosophy 16: 1–43. , Perry (2003Perry, John 2003 “Predelli’s Threatening Note: Contexts, Utterances, and Tokens in the Philosophy of Language.” Journal of Pragmatics 35: 373–387. , 2006Perry, John 2006 “Using Indexicals.” In The Blackwell Guide to the Philosophy of Language, ed. by M. Devitt, and R. Hanley, 314–34. Malden, MA: Blackwell. ), Predelli (1998aPredelli, Stefano 1998a “ ‘I Am Not Here Now’.” Analysis 58: 107–115. , 1998bPredelli, Stefano 1998b “Utterance, Interpretation and the Logic of Indexicals.” Mind & Language 13: 400–414. , 2002Predelli, Stefano 2002 “Intentions, Indexicals and Communication.” Analysis 62: 310–316. , 2011Predelli, Stefano 2011 “I Am Still Not Here Now.” Erkenntnis 74: 289–303. ), Radelescu (2018)Radelescu, Alexandru 2018 “The Difference Between Indexicals and Demonstratives.” Synthese 195: 3173–3196. , Romdenh-Romluc (2002Romdenh-Romluc, Komarine 2002 “Now the French are Invading England!” Analysis 62: 34–41. , 2006Romdenh-Romluc, Komarine 2006 “ ‘I’.” Philosophical Studies 128: 257–283. , 2008Romdenh-Romluc, Komarine 2008 “First-person Thought and the Use of ‘I’.” Synthese 163: 145–156. ), Sidelle (1991)Sidelle, Alan 1991 “The Answering Machine Paradox.” Canadian Journal of Philosophy 21: 525–539. , Smith (1989)Smith, Quentin 1989 “The Multiple Uses of Indexicals.” Synthese 78: 167–191. , Stevens (2009)Stevens, Graham 2009 “Utterance at a Distance.” Philosophical Studies 143: 213–221. , Voltolini (2008)Voltolini, Alberto 2008 “Fiction as a Base of Interpretation Contexts.” Synthese 153: 23–47. , Weatherson (2002)Weatherson, Brian 2002 “Misleading Indexicals.” Analysis 62: 308–310. . The next section will introduce the notion of perspective and show how it can be deployed to undermine a selection of counterexamples from the literature. This strategy can be utilised to defend many different analyses of character. For the purpose of explaining the strategy, however, I will deploy it primarily to defend a relatively traditional interpretation of Kaplan’s original theory. This paper is not intended as an endorsement of that interpretation, however, and it should be kept in mind that the general strategy is in no way wedded to this traditional interpretation. The notion of perspective is not only intended to pose a problem for counterexamples to Kaplan’s analysis of character but a more general problem for the use of counterexamples in the debate.

3.Reference and perspective

We begin with a counterexample to Kaplan’s semantics for indexicals proposed by Mount (2015)Mount, Alison 2015 “Character, Impropriety, and Success: A Unified Account of Indexicals.” Mind & Language 20: 1–21. , whose terminology I will soon adopt. Mount’s central case involves Cora, who avoids a company party by pretending to attend a cousin’s wedding in Seattle. On the day of the party, Cora is at home in Chicago. She receives a call from a colleague who asks if she is enjoying the wedding. Cora responds:

According to Mount, Cora falsely claims that she is in Seattle. If Kaplan’s semantics is correct, however, ‘here’ refers to the location at which it is uttered (Chicago) and Cora says truly that she is in Chicago. Mount concludes that Kaplan’s semantics is unable to explain the example.

Mount proposes that indexicals are sensitive to the mutually accepted perspective of the utterance.66.Or, more precisely, the perspectives that are mutually accepted, given that Mount appeals to independently shiftable perspectives for agent, location, and time. Mount often talks about the perspective of a discourse. Presumably, the agent perspective is associated with individual utterances, rather than discourses, given that the referent of ‘I’ shifts depending on who is speaking. Indeed, the locational perspective presumably varies throughout the discourse when interlocutors occupy different locations. When we speak on the telephone, we will each refer to the other’s location as ‘there’ rather than ‘here’. If the discourse is extended over a significant time, what was previously referred to as ‘now’ will later be referred to as ‘then’. I therefore prefer Mount’s formulation in terms of utterance perspectives. A perspective serves as a stable point from which reference can be determined. In the simplest cases, the mutually accepted perspective is simply the context of utterance and indexicals are associated with the referents that Kaplan identifies: ‘I’ refers to the speaker, ‘here’ to the location of utterance, and ‘now’ to the time of utterance. The mutually accepted perspective of an utterance can, however, be shifted from the context of utterance in various ways. Cora’s audience mistakenly believes that Cora is speaking from Seattle, leading them to interpret from the perspective of Seattle. Because this is also the perspective that Cora intends, Seattle is the mutually accepted perspective and therefore the referent of ‘here’, explaining the intuition that Cora has falsely claimed to be in Seattle. Note that adopting Seattle as the perspective of the utterance does not require Cora to believe that she is speaking from Seattle. As Mount (2015Mount, Alison 2015 “Character, Impropriety, and Success: A Unified Account of Indexicals.” Mind & Language 20: 1–21. , 6) puts it, “accepting a perspective for the purposes of a discourse is a matter of speaking (or being disposed to speak) as if certain things are true”. Cora’s audience is disposed to speak as though Cora is in Seattle because that is what they believe. Cora is disposed to speak that way because she is being deceptive.

Mount’s notion of perspective is an extremely important contribution to the literature on indexicals77.Perhaps it has its roots in some comments by Kaplan (1989aKaplan, David 1989a “Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstrative and Other Indexicals.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 481–563. New York: Oxford University Press., 525–526) noting “the importance of perspective”. but it significantly complicates Mount’s argument against Kaplan’s analysis of character. Even if Kaplan’s analysis is correct, we should expect the content communicated by Cora to be affected by the mutually accepted perspective. Suppose that Kaplan’s analysis is correct and a token of ‘here’ refers to the location at which it is uttered. Cora’s audience will attempt to assign Cora’s location as the referent of ‘here’. Because they believe that Cora is in Seattle, however, they will interpret from the perspective of Seattle and assign Seattle as the referent. This is, of course, precisely what Cora intends and she thereby communicates the falsehood that she is in Seattle. In such cases, where the audience interprets with respect to a context other than that which determines the referent of an indexical, I will describe the interpreter as adopting a shifted perspective.

Assuming that Kaplan’s account is correct, we should expect the Cora example to play out just as stipulated. Let me be absolutely clear to avoid any potential confusion: the foregoing was not intended as an endorsement of Kaplan’s analysis of character. Indeed, this paper does not endorse any particular analysis of character. The point was to demonstrate how Kaplan could appeal to perspective shifts to offer a very plausible explanation of the purported counterexample. We will soon see how perspective shifts can be used to defend alternative analyses of indexicals. It is this broad applicability of the strategy that poses a general challenge to the use of counterexamples in the debate over indexicals’ characters.

Let’s first examine another proposed counterexample to Kaplan’s analysis:

It’s Daniel’s office hour and he has gone to lunch. In order to save Daniel’s students some time, Frank writes ‘I am out to lunch’ on a scrap of paper and attaches it to Daniel’s office door. When Daniel’s students arrive, they recognise that Daniel is out to lunch and promptly leave.

Corazza et al. (2002Corazza, Eros, Willaim Fish, and Jonathan Gorvett 2002 “Who is I?” Philosophical Studies 107: 1–21. , 4–5) take this case to raise “an intriguing and important question of just how the reference of ‘I’ is determined”. They argue that the indexical refers to Daniel; after all, Daniel is “certainly what the audience of the utterance, the students, take it to refer to” (2002Corazza, Eros, Willaim Fish, and Jonathan Gorvett 2002 “Who is I?” Philosophical Studies 107: 1–21. , 5). As they assume that Kaplan’s theory of indexicals identifies the referent of ‘I’ with its utterer, they take Kaplan’s theory to “force the extremely counter-intuitive claim that one can make utterances one has had absolutely no role whatsoever in the production of” (ibid.). They conclude that ‘I’ does not always refer to the utterer, suggesting instead that the referent depends on the conventions associated with the medium of communication; in this case the conventions associated with office-door notes.

We can resist the claim that Kaplan’s analysis identifies the referent of ‘I’ with its utterer.88.See Footnote 1. For the purposes of their argument, however, I will accept it. Even so, we should expect the students to interpret ‘I’ as referring to Daniel if Kaplan’s view is correct, because their mistaken beliefs will lead them to shift perspective. If the students follow Kaplan’s semantics, they will attempt to interpret ‘I’ as referring to the author of the note. A first-personal note attached to an office door would usually be written by the regular occupant of the office, so the students can be expected to falsely believe that Daniel authored the note, explaining why they mistakenly believe that Daniel is the referent. As the students’ mistake is predictable, Frank can exploit it to communicate that Daniel is out to lunch.99.Montminy (2010Montminy, Martin 2010 “Context and Communication: A Defense of Intentionalism.” Journal of Pragmatics 42: 2910–2918. , 2915) suggests something similar, saying that Daniel is the speaker’s referent in Kripke’s (1977)Kripke, Saul 1977 “Speaker’s Referent and Semantic Referent.” Midwest Studies in Philosophy 2: 255–276. sense. Using this terminology, we can say that the interpreter is able to identify the speaker’s referent because Daniel would be the semantic referent if the interpreter’s mistaken beliefs were accurate.

Perspective significantly complicates attempts to construct counterexamples. If perspective shifts were only ever the result of mistaken beliefs, they could be avoided by stipulating that the speaker and hearer have no false beliefs about any facts that might be semantically relevant, but once we admit that it is possible to shift perspective by accident, it is plausible that we can shift perspective intentionally when we believe that the speaker intends us to. Consider the following case:

Frank has made a cake for a party he and Daniel are attending. He leaves it to cool on the table, next to a note which reads ‘I am a cake for the party. Please don’t eat me!’ On reading the note, Daniel recognises that the cake on the table is a cake for the party and the cake survives uneaten.

Unlike the previous case, Daniel needn’t have any belief as to who wrote the note and certainly needn’t falsely believe that it was written by the cake. Still, we needn’t follow Corazza et al.’s (2002)Corazza, Eros, Willaim Fish, and Jonathan Gorvett 2002 “Who is I?” Philosophical Studies 107: 1–21. reasoning and conclude that the deceptively simple term ‘I’ is associated with a surprisingly complex character that allows for reference, not only to the author of a note, but also to a cake prepared by the author of a note. Even supposing that ‘I’ strictly refers to Frank, the note might lead Daniel to imagine a (mildly) amusing context in which the cake itself is the author, which goes some way towards explaining why such notes are (mildly) amusing. If the cake had written the note, then ‘I’ would refer to the cake, according to Kaplan’s analysis. Daniel can recognise that Frank intends him to interpret as though the cake authored the note, insofar as he interprets ‘I’ as referring to the cake.

In this case, the perspective shifts, not as a result of a mistaken belief, but as a result of an interpretive pretence: the pretence that the cake wrote the note. Voltolini (2008)Voltolini, Alberto 2008 “Fiction as a Base of Interpretation Contexts.” Synthese 153: 23–47. , following Recanati (2000)Recanati, François 2000 Oratio Obliqua, Oratio Recta. Cambridge, MA: MIT Press. , incorporates pretence into a semantics for indexicals. The point here is that we can employ pretence non-semantically, to explain otherwise problematic cases.

There is a natural explanation of these cases in terms of perspective that coheres with Kaplan’s analysis of character. Let me stress again, however, that I am not endorsing Kaplan’s analysis. My goal here is only to demonstrate how the possibility of perspective shifts presents a challenge for the construction of counterexamples, not to defend any particular analysis of character. In fact, other analyses of indexicals can use the same general strategy to explain away purported counterexamples. Predelli (1998aPredelli, Stefano 1998a “ ‘I Am Not Here Now’.” Analysis 58: 107–115. , 1998bPredelli, Stefano 1998b “Utterance, Interpretation and the Logic of Indexicals.” Mind & Language 13: 400–414. ) presents the following case in an attempt to dislodge the view, proposed by Sidelle (1991)Sidelle, Alan 1991 “The Answering Machine Paradox.” Canadian Journal of Philosophy 21: 525–539. and more recently endorsed by Cohen (2013)Cohen, Jonathan 2013 “Indexicality and the Puzzle of the Answering Machine.” Journal of Philosophy 110: 5–32. , that a token of ‘now’ refers to the time at which it is interpreted. According to Predelli’s view, the referents of indexicals are determined by the speaker’s intentions: ‘now’ refers to whichever time the speaker intends:

Consider the anecdote of Jones, who expects his wife to come home at six, and writes ‘I am not here now’ at four, with the intention of informing Mrs. Jones that he is away from home at six – or, if you prefer, imagine that he records ‘I am not here now’ on a tape, expecting his wife to activate the tape-recorder upon her arrival. However, Jones’s wife is late, and she only reads the message (or turns on the tape-recorder) at ten. Clearly the vicissitudes of Mrs. Jones do not affect the content of Jones’s message. Intuitively, such content is to be established with respect to the time of intended decoding, and not with respect to the time when decoding actually took place.(Predelli 1998aPredelli, Stefano 1998a “ ‘I Am Not Here Now’.” Analysis 58: 107–115. , 110)

Predelli’s claim that Mrs. Jones’s lateness cannot affect the content of the note appears to beg the question against his opponent. Perhaps, however, Predelli’s intention is not this abrupt denial of the opposing view but the very plausible claim that Mrs. Jones’s lateness will not (and should not) affect how Mrs. Jones interprets ‘now’. Yet, we can explain this fact without denying that the reference of ‘now’ is determined by the context in which it is interpreted. If ‘now’ refers to the time at which it is interpreted, then it refers to ten o’clock, but Mrs. Jones, knowing when Mr. Jones expected her home, will interpret the indexical from the perspective of six, as Mr. Jones intended.

The audience may shift perspective for a variety of reasons. They may attempt to interpret an indexical relative to the actual context and interpret relative to another context due to mistaken belief (as Cora’s audience plausibly does), or they may consciously shift from the perspective of the actual context to accommodate the intentions of the speaker, whether because they recognise that the speaker is mistaken about the actual context (as plausibly happens in the case of Mrs. Jones), or because they recognise that the speaker invites a certain pretence (as in the case of the cake).

Perspectives can also be made explicit, as in some cases of the historical present, like the following utterance made in 1996: “It is 1796. Napoleon, now commander of the French troops in Italy, defeats the Sardinian forces and turns against Austria.”1010.Predelli (1998aPredelli, Stefano 1998a “ ‘I Am Not Here Now’.” Analysis 58: 107–115. , 112). See also Predelli (1998bPredelli, Stefano 1998b “Utterance, Interpretation and the Logic of Indexicals.” Mind & Language 13: 400–414. , 406). Predelli notes that the author is “not conveying the false information that Napoleon is the commander of the French troops in 1996”. Certainly not, but we needn’t think that ‘now’ will or should be interpreted as referring to 1996, even if that is its actual referent. The speaker introduces a new perspective, explicitly signalled by ‘It is 1796’. When ‘now’ is taken to refer to 1796, the utterance conveys the (I presume) true information that Napoleon is the commander of the French troops in 1796.

How is this perspective shift achieved? One possibility is that ‘It is 1796’ is a monstrous operator, which semantically shifts the time parameter of the context. Kaplan (1989aKaplan, David 1989a “Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstrative and Other Indexicals.” In Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, 481–563. New York: Oxford University Press., 510) argued that English has no such operators and “such operators could not be added to it”.1111.For more on monstrous operators, see Rabern and Ball (2019)Rabern, Brian and Derek Ball 2019 “Monsters and the Theoretical Role of Context.” Philosophy and Phenomenological Research 98 (2): 392–416. and the papers cited therein. Instead, I suggest that the phrase leaves the parameters of context untouched, but introduces a perspective shift. ‘It is 1796’ can be paraphrased as ‘The present year is 1796’. If the indexical ‘the present year’ is taken to refer to 1996, the content is contradictory. To avoid falsity, interpreters shift their perspective to the year 1796. That leads to the trivial content that 1796 is 1796, but the phrase introduces a new perspective from which the indexical ‘now’ is interpreted, delivering the non-trivial content that, in 1796, Napoleon was the commander of the French troops.1212.In the terminology of Rabern and Ball (2019)Rabern, Brian and Derek Ball 2019 “Monsters and the Theoretical Role of Context.” Philosophy and Phenomenological Research 98 (2): 392–416. , ‘It is 1796’ is neither a formal monster nor a content monster. It does not change what is said, or the context parameter that determines what is said, but it affects the context that the interpreter will use to interpret indexicals.

I expect several of the examples discussed so far to be somewhat contentious. To know what to make of these cases, we need some way of telling when appeals to perspective shifts are legitimate. We could do so if we had a way to identify the referent of an indexical in a given case. In the following section, I consider several extant tests for semantic content and conclude that none of them can play this role.

4.Identifying semantic content

These counterexamples are intended to disprove some theory of indexicals by showing that the theory associates indexicals with the wrong referents. The previous section showed how counterexamples are often plausibly undermined by appealing to a perspective shift. How can we tell when this explanation is appropriate? We could answer this question if we had a test for the reference of an indexical in a given case. The bluntest test is something like this

This is probably the most utilised test in the literature. A case is presented featuring the use of an indexical and its author opines as to that indexical’s referent. Sometimes the test is bolstered by appeal to what is intuitively said, stated, or claimed, the assumption being that the referent of an indexical is whatever delivers the intuitive content.1313.See Mount (2015Mount, Alison 2015 “Character, Impropriety, and Success: A Unified Account of Indexicals.” Mind & Language 20: 1–21. , 3): “she [Cora] claimed to be at a wedding in Seattle, when in fact she was in Chicago”. Similarly, theorists often appeal to intuitions about truth-value: The referent of an indexical is one that determines a proposition that matches the intuitive truth-value of the utterance.1414.See, for example, Weatherson (2002Weatherson, Brian 2002 “Misleading Indexicals.” Analysis 62: 308–310. , 309): “It seems to me that what I wrote on the note was true”; Mount (2015Mount, Alison 2015 “Character, Impropriety, and Success: A Unified Account of Indexicals.” Mind & Language 20: 1–21. , 3): “Intuitively, Cora said something false”; and Sidelle (1991Sidelle, Alan 1991 “The Answering Machine Paradox.” Canadian Journal of Philosophy 21: 525–539. , 528): “Some utterances of ‘I am not here now’ are true at the time of utterance, e.g. those presented by answering machines (in their non-screening use)”.

If it strikes you as a totally inadequate test, then you have already been convinced of this paper’s central claim, but I’ll briefly outline the problems raised by shifts in perspective. Intuitions about the referents of indexicals, about what is said, and about the literal truth-values of utterances vary significantly between competent speakers, as is amply demonstrated by the literature on indexicals. What seems to be happening is that at least some of these theorists are shifting perspective and so misidentifying the referent. The challenge is to tell which. Even when all competent speakers agree on an intuitive referent, we have no guarantee that the intuitive referent is the actual referent. If the intuitive referent of an indexical is whatever we are most naturally disposed to take the indexical to refer to, then we should expect that, when we are naturally disposed to shift perspective, reliance on intuition will lead us to misidentify the referent. We may be naturally disposed to assign Seattle as the referent of ‘here’ as uttered by Cora, for example, but this is exactly what we should expect if Kaplan’s semantics is correct and the context compels us to shift our perspective.

Similar problems arise for Michaelson’s (2016)Michaelson, Eliot 2016 “The Lying Test.” Mind & Language 31: 470–499. Lying Test, when applied to indexicals:1515.Here, I am concerned only with the Lying Test as a means of resolving disagreements about indexical reference. Even if the Lying Test is not useful in this context, it might be very useful for achieving clarity in other areas.

Michaelson suggests two very plausible jointly sufficient conditions for lying, where p is the semantic content of an utterance P: (1) the speaker believes that p is false, and (2) the speaker intends to deceive their audience with respect to p. If p is the semantic content of P, then P is either a lie, in which case A is true, or it is not a lie, in which case one of these sufficient conditions must be unsatisfied, and either B or C must be true. If A, B, and C are false then, by modus tolens, p cannot be the semantic content of P.

The test is deployed as follows. Pick a semantic theory to test and find an utterance to which it applies. The utterance should be one that is not a lie, so that A is guaranteed to be false. Then check to see whether B or C is true. If we can find a non-lie that satisfies neither B nor C, then the semantic theory is falsified.

We might think that we can use the Lying Test to identify the referent of an indexical. Consider the case of Cora. We might argue as follows:

1. If Cora’s utterance of ‘here’ refers to anywhere, then it refers either to Seattle or to Chicago. (assumption)

2. If Cora’s utterance of ‘here’ refers to Seattle then the content of Cora’s utterance is the proposition that Cora is in Seattle. (from the semantics of the sentence uttered)

3. Cora believes that this proposition is false. (assumption)

4. Cora intends to deceive their audience with respect to this proposition. (assumption)

5. Cora’s utterance is not a lie. (assumption)

6. So, by the Lying Test, the proposition that Cora is in Seattle is not the semantic content of Cora’s utterance. (from 3, 4, 5)

7. So, by modus tolens, Cora’s utterance of ‘here’ does not refer to Seattle. (from 2, 6)

8. Therefore, if Cora’s utterance of ‘here’ refers to anywhere, it refers to Chicago. (from 1, 7)

The Lying Test relies on intuitions about whether or not someone has lied (premise 5) and leverages them to falsify theories of semantic content. The test is, therefore, only as reliable as our intuitions about lying. The problem is that our intuitions about whether or not someone has lied by using an indexical are plausibly influenced by our intuitions about the reference of the indexical. If perspective shifts can lead us to mistake the referent of an indexical, therefore, they can plausibly lead us to mistake lies for mere deception and vice versa. Did Cora lie, for example? For me at least, the answer seems to depend on the reference of the indexical. If she referred to Chicago then she said truly that she was in Chicago and did not lie, though she was deceptive. If our intuitions about lying are influenced by our intuitions about reference in this case, then we should not expect our intuitions about lying to be more reliable than our intuitions about reference.

Michaelson (2016Michaelson, Eliot 2016 “The Lying Test.” Mind & Language 31: 470–499. , 484) considers the objection that “judgments about whether or not someone has lied are ultimately driven by judgments about what they said, claimed, or stated”, and demonstrates that there are cases in which we have clear judgments about whether a speaker has lied, without clear judgments about what they have said. The concern, however, is not that all judgments about lying depend on prior judgments about semantic content (Michaelson’s test may be very useful when our intuitions about lying are more reliable than our judgments about semantic content), but that some judgements about semantic content may influence judgments about lying. If our judgments about whether someone has lied using an indexical are influenced by our judgments about the referents of the indexicals, then the former will be no more reliable than the latter.

The key problem for each of these tests is that there is not always a conscious difference between cases in which we shift perspective and cases in which we do not. Interpretations resulting from a shifted perspective are ‘immediate’ in the sense that they are not derived from any prior truth-conditional content.1717.See, for example, Recanati (2004Recanati, François 2004 Literal Meaning. Oxford: Oxford University Press., 17) who says that what is said “corresponds to the primary truth-evaluable representation made available to the subject”. This makes them very difficult to distinguish from what is said.

The results in this section are primarily negative, as I have no replacement test to offer. Yet, I hold out hope that others might be able to generate tests that do better than those considered here, whether linguistic tests or experimental studies.

5.Nihilism, pluralism, and epistemicism

Suppose we become convinced that there is no way to tell when we can appeal to perspective shifts. We are left with a number of cases in which there are multiple candidates for the referent of an indexical, and so multiple viable analyses of the character that determines the referent, each of which explains the data equally well. What should we say about the meanings of indexicals when faced with many viable analyses of their characters? This section considers three possible responses, each of which deals with this proliferation of characters in a different way: epistemicism, pluralism, nihilism. Respectively, they are the view that indexicals have a single unknowable character, the view that indexicals have more than one character, and the view that indexicals have no characters. I will first set pluralism to one side and then argue that we should prefer nihilism over epistemicism.

The pluralist takes indexicals to have multiple characters.1818.For another view that could be labelled as ‘pluralism’, see von Fintel and Gillies (2011)von Fintel, Kai, and Anthony Gillies 2011 “ ‘Might’ Made Right.” In Epistemic Modality, ed. by A. Egan, and B. Weatherson, 108–130. Oxford University Press. . ‘Here’ for example, may be associated with a set of characters, all of which assign the same referent in simple cases, but assign different referents in difficult cases like Cora’s utterance.1919.An anonymous reviewer suggests Gómez-Torrente (2020)Gómez-Torrente, M. 2020 Roads to Reference: An Essay on Reference Fixing in Natural Language. Oxford: Oxford University Press. as an example of this kind of view. Though Gómez-Torrente identifies several different rules for the identification of referents, his view is not pluralist in the sense considered here, as these different rules never identify conflicting referents. We can instead view Gómez-Torrente as associating indexicals with a single complex character. Likewise for Smith (1989)Smith, Quentin 1989 “The Multiple Uses of Indexicals.” Synthese 78: 167–191. , who argues that indexicals are associated with multiple characters but also with a single metacharacter that determines which character is appropriate in each context. The pluralist view raises many difficult questions that I am unable to answer. Suppose that an indexical i is associated with a set of characters C. How does C determine the referent of i? Perhaps x is the referent of i when all members of C agree on x as the referent. When some members of C disagree about the referent, what is the referent of i? If there is a single referent, how is it determined? If i has many referents, then how should we evaluate a sentence in which i appears? Is ‘F(i)’ both true and false when one referent is F and the other is not? Is it sufficient for truth that one referent is F, or is supertruth (truth on all interpretations of the referent) required? And what purpose would such multiple reference have in communication? If, on the other hand, i has no referent when the members of C conflict, why posit multiple characters in the first place?2020.An alternative is that such cases are simply not decided. Such a view is presented by Gómez-Torrente (2020)Gómez-Torrente, M. 2020 Roads to Reference: An Essay on Reference Fixing in Natural Language. Oxford: Oxford University Press., who associates indexicals merely with sufficient conditions that leave various cases undecided. I have a lot of sympathy for this view of character, but it does not avoid the problems raised in this section. This is one viable analysis of character among many. Further, there will be viable analyses of character that disagree about what the undecided cases are. Following the pluralist response to these disagreements, we would then have to associate indexicals with multiple characters, each of which leaves different cases undecided. This raises further questions that I am also unable to answer. When these characters disagree about whether the referent is decidable, is it decidable or not? Why not just associate i with a single character which delivers a single referent when the members of C agree and no referent when the members of C disagree? Perhaps interesting answers can be given to these questions but I am not able to give them.

There is a more pressing problem for the pluralist line, however. Faced with multiple analyses of character that explain the data equally well, the pluralist says that many of them are right. The problem, however, is that there will be multiple pluralist explanations that explain the data equally well. One pluralist view might associate an indexical with the set of characters {C ₁,C ₂,C ₃} while another associates the indexical with {C ₁,C ₂}, appealing to perspective shifts to explain the cases for which the alternative analysis postulates C ₃. How are we to deal with this proliferation of pluralist theories? It won’t help to reapply the pluralist response, associating indexicals with sets of sets of characters, as the problem will only re-emerge at a higher level: which among the viable alternatives is the right set of sets? In the end, it seems to me, we will have to either admit that we cannot know, which amounts to epistemicism, or deny that indexicals are associated with characters, or sets of characters, or sets of sets of characters, and so on, which amounts to nihilism. The pluralist response to the problem therefore reduces to the nihilist and epistemicist responses that we consider next.

On the nihilist view, indexicals have no characters.2121. Sherman (2015)Sherman, B. 2015 “Constructing Contexts.” Ergo 2 (23): 581–605. defends a Context Constructivism that is very closely related to the nihilist proposal. On that view, indexicals are associated with unique characters that specify contextual roles (‘I’, is associated with speaker role, for example), but these roles are assigned by language-users, rather than being objectively real features of the conversational situation. The perspective-shifting view described here has so far agreed with Sherman’s view that roles such as speaker are assigned during interpretation but takes there nonetheless to be objective facts about who fits those roles and so about the referents of indexicals (though I have been completely neutral as to what those facts are). Given this possibility, it is hard to construct convincing counterexamples against analyses of character. My aim in this section is to consider how we should respond if we find ourselves unable to identify the characters of indexicals. There are interesting connections between the nihilist view defended in this section and the Constructivism of Sherman (2015)Sherman, B. 2015 “Constructing Contexts.” Ergo 2 (23): 581–605.. Due to considerations of length, however, I am unable to consider them here. The issue is not that the notion of character is incoherent, inconsistent, or poorly defined. It is a perfectly coherent notion. According to nihilism, however, it is a matter of contingent fact that there are no characters associated with indexicals in a natural language like English. Despite rejecting a key assumption of Kaplan’s framework for indexicals, nihilism requires a reinterpretation of the framework, rather than its total rejection. Instead of taking the characters of indexicals to be determined by the conventions of a language, the nihilist can take characters to be assigned to indexicals by interpreters. This allows for different interpreters to associate indexicals with different character functions, without any interpreter violating the conventions of the language. For example, I might assign a character to ‘I’ such that it refers to Frank in the office-note case, while you might associate the same indexical with a character that assigns Daniel as the referent.

If the assignment of characters were totally arbitrary, then there would be no possibility of successful communication. There would, for example, be no reason to think that a speaker uses ‘I’ to refer to an individual rather than a location. In fact, however, speakers generally agree about the referents of indexicals in all but the more difficult cases, such as those used by philosophers in the formulation of counterexamples. How is this agreement possible if indexicals have no conventional characters that speakers can use to coordinate on referents? The nihilist can explain broad agreement on reference as follows. Children will be introduced to indexicals as members of a community that already agrees about the referents of indexicals in simple cases.2222.How humans came to agree on referents in the first place is not a matter for semantics but for evolutionary linguistics. Through exposure to these cases, they will identify the agreed referents in these cases and, from those referents, reverse-engineer characters. But those simple cases will be consistent with a range of different characters that generate different referents in more unusual cases. Different speakers may therefore end up associating indexicals with different characters, though those differences may only become apparent (if at all) much later in life.

Successful communication is therefore explicable, despite differences in idiolectical character, because the community has been trained on similar sorts of cases. In the simplest cases, the different characters associated with an indexical by different speakers will determine the same referent. In these cases, communication is unhindered. In slightly more complex cases, those characters might determine different referents but speakers can coordinate on the same referent when contextual factors lead one of the speakers to shift perspective. Again, communication can succeed without a hitch. In very unusual cases, speakers might associate an indexical with different characters and with different referents. In these cases, communication is endangered. The significant difference here between the nihilist view and the standard Kaplanian view is that the nihilist needn’t take either speaker to be misidentifying the conventional character of the indexicals. The nihilist can use the Kaplanian machinery to explain communication, even in difficult cases like that of Frank and Daniel, but needn’t worry about identifying unique characters for indexicals.

On the nihilist view, natural language indexicals have no conventional characters. On the epistemicist view, indexicals have unique characters that we are unable to identify. Epistemicism was popularised by Williamson (1994a)Williamson, Timothy 1994a Vagueness. London, Routledge. as a response to the sorites paradox. Consider a series which starts with a contemporary chicken and extends by the mother-of relation. We have our present chicken, her mother, her mother’s mother, and so on. Evolutionary theory tells us that, eventually, we shall reach an ancestor which is not a chicken. At some point, we move from a series of chickens to a series of non-chickens, but every mother is so similar to her daughter that it becomes impossible to pinpoint the last chicken in the series. The epistemicist responds that there is a sharp cut-off between the chickens and the non-chickens – there is a mother-daughter pair such that the mother is the first non-chicken in the series and the daughter is the last chicken – but that we cannot know where that cut-off is. To identify the last chicken requires a knowledge of the meaning of ‘chicken’ that is simply unattainable. According to Williamson (1994bWilliamson, Timothy 1994b “Definiteness and Knowability.” The Southern Journal of Philosophy 33: 171–191. , 184–5) the last chicken is determined by an unknowable function of the unknowable dispositions of the language users.

The epistemicist view posits a realm of unknowable facts about the characters of indexicals, while the nihilist view posits none. Considerations of theoretical and ontological simplicity would therefore seem to mediate in favour of nihilism. Epistemicism might be preferable to nihilism if it offered a better view of communication, yet nihilist and epistemicist explanations of communication will be strikingly similar. Both agree that interpreters cannot know the characters of indexicals, so both have to explain how communication can succeed in the face of multiple viable characters. Such an explanation was offered above.

To endorse epistemicism, we need compelling reason to prefer it to nihilism. Williamson (1994a)Williamson, Timothy 1994a Vagueness. London, Routledge. provides arguments against the global nihilist view that all sorites-susceptible terms are meaningless in that they are necessarily true of nothing. The concern is that global nihilism will spiral out of control. Williamson (1994aWilliamson, Timothy 1994a Vagueness. London, Routledge., 171) argues that global nihilism seems “unlikely to receive true expression, whether in language or thought. Such an expression would be an utterance, inscription, thought, or the like; once again, ‘utterance’, ‘inscription’, ‘thought’ and ‘expression’ itself are sorites-susceptible, and so would be empty if nihilism were true”.

The global nihilist should welcome Williamson’s observation that nihilism is, in a sense, inexpressible. Constructing a sentence with the content of the global nihilist thesis would require the use of sorites-susceptible terms, which are devoid of content. Yet, it may be possible to communicate propositions even when the sentences we utter have no semantic content,2323.See Bowker (2019Bowker, Mark 2019 “Underdetermination, Domain Restriction, and Theory Choice.” Mind & Language 34 (2): 205–220. , 2022Bowker, Mark 2022 “Ineliminable Underdetermination and Context-Shifting Arguments.” Inquiry 65 (2): 215–236. ) for a view on which we can communicate without semantic content. so failure to semantically express the content of the global nihilist thesis does not entail that global nihilism is not pragmatically expressible. Enoch (2007Enoch, David 2007 “Epistemicism and Nihilism About Vagueness: What’s the Difference?” Philosophical Studies 133: 285–311. , 301), for example, suggests that it is possible to communicate the content of the global nihilist thesis through presentation of examples: ‘Tall’ is necessarily true of nothing, ‘heap’ is necessarily true of nothing… and so on. Regardless, nihilism about the characters of indexicals does not entail global nihilism, so Williamson’s argument against global nihilism poses no problem for the limited form of nihilism that concerns us here.

6.Conclusion

I have not sided with any particular analysis of indexicals in this paper. Rather, I have presented a general strategy for responding to counterexamples that can be deployed in defence of many different analyses. To draw any substantial conclusions from these examples, we need some way to tell when this strategy is legitimate. We looked at two extant tests for semantic content and saw that neither of them can help. The issue therefore remains open. Those who would rely on the method of counterexamples in arguing for or against analyses of indexicals must provide reason to think that their examples involve no perspective shift. I then considered how we should respond if we become convinced that it is not possible to identify unique characters for indexicals and argued that a nihilist view should be preferred over pluralism or epistemicism.