1 MEANING SHIFT AND THE PURITY OF ‘ I ’ Edison Barrios

In this paper I defend the “Standard View” of the semantics of ‘I’ – according to which ‘I’ is a pure, automatic indexical – from a challenge posed by “deferred reference” cases, in which occurrences of ‘I’ are (allegedly) not speaker-referential, and thus non-automatic. In reply, I offer an alternative account of the cases in question, which I call the “Description Analysis” (DA). According to DA, seemingly deferred-referential occurrences of the 1st person pronoun are interpreted as constituents of a definite description, whose operator scopes over an open sentence Rxy – where R is a contextually selected relation ranging over pairs of people and objects. The role of intentions is thus limited to the determination of R, which is posterior to the fixation of the reference of ‘I’. In support of the DA I present evidence that, in the cases in question, the (Determiner) phrase containing ‘I’ behaves in relevant ways like a description. I show that the DA can account for the problematic examples, while preserving the simplicity of the standard semantics of ‘I’. Finally, I examine a rival account of the data, offered by Nunberg (1993), and argue for the superiority of the DA. 1. – The Standard View and a Challenge The most widespread view of the semantics of the first person singular pronoun contains, among others, these two theses: a. Speaker Referentiality: every utterance u of ‘I’ refers to the speaker of u.1 b. Purity: ‘I’ is a pure indexical. Pure indexicals, such as ‘I’, are “automatic”. That is, their designation “is fixed given its meaning and public contextual facts” and “no further intention, [beyond the intention of using it with its ordinary meaning] is relevant to determining the referent” (Perry 2006, 320, italics mine). In this respect, pure indexicals – and ‘I’, in particular – are the opposite of “discretionary” indexicals (such as demonstratives), whose designation depends partly on the speaker’s intentions. In consequence, the second thesis entails that intentions can have no role in fixing the reference of an occurrence of ‘I’. 2 1 For instance according to Kaplan, “In each of its utterances, ‘I’ refers to the person who utters it”. Likewise, for Perry (2006, 318), “a given utterance u of ‘I’ refers to the speaker of u”. 2 “The linguistic rules which govern their use fully determine the reference for each context. No supplementary actions or intentions are needed” (Kaplan, 1989, p. 491). Also, on Kaplan’s view, ‘I’ stands in sharp contrast to “impure” indexicals, such as demonstratives, whose reference cannot be fixed unless some demonstration or intention is taken into account. 2 Let us call this view of the semantics of ‘I’ the “Standard View”. Despite its intuitive appeal the Standard View has had its share of doubters. Among the most recent is Mount (2008), who argues that all indexicals, including ‘I’, are discretionary (i.e. non-automatic). This claim occurs in a more general discussion, aimed at showing that the distinction between “pure” and “impure” indexicals is spurious, since speaker intentions invariably have a crucial role in the determination of reference, indexical or otherwise. Mount’s objections to the Standard View are driven by a series of cases in which the designation of (occurrences of) the first person singular pronoun seems to have shifted away from the speaker, thus challenging presumptions of speaker-referentiality and thus of automaticity. This phenomenon – let us collectively refer to its instances as “Agent Shift” cases – is illustrated by the following utterances:3 (1) a. [Bill Clinton is at a celebrity wax museum, with Hillary. After a quick search, he finally finds the figure that represents him, which stands between Charlie Chaplin’s and Groucho Marx’s figures. However, Hillary hasn’t spotted it yet. To help her find his figure, Bill says to her:] I am next to Chaplin. b. [We are playing Monopoly, and I just made a move. I say to my wife, who has just arrived:] I am on the purple square (adapted from Mount, 2008, p. 200). c. [A man bets on a horse at a particular race, and the horse happens to finish last in the race. He says to his friend:] I finished last in the race (Adapted from Smith, 1989, p.183). d. [There is a summit of world leaders. To ensure the safety of the participants, each politician has been assigned a body double. One morning, Vladimir Putin saunters into the room and, upon seeing his double, he addresses him thus:] How am I doing today? (Adapted from Akman, 2001, q.b. Perry, 2006). e. [Borges, the argentine writer, is asked to comment on his own work, but he refuses. When the journalists inquire about the reason, Borges says:] I haven’t read myself in a long time. Prima facie, agent shift examples seem to be instances of metonymy, brought about by what has been called “deferred reference”.4 Deferred reference is a process whereby an expression gets to refer to an object B, where such an expression does not conventionally refer to B but to an object A that is related to, but distinct 3 A sentence like my example (1b) appears in Mount’s paper, the rest do not. 4 The phenomenon of deferred reference has been widely reported in the literature. An important source in linguistics is Nunberg (1977), and in philosophy discussion goes back at least as far as Quine’s (1968) discussion of “deferred ostension”. 3 from B.5 Deferred reference has been mostly studied with respect to demonstrative and personal pronouns, where the demonstratum can be different from the referent, e.g.: (2) a. [Restaurant patron to valet, holding up a car key:] This is parked out back. (Nunberg 1995, ex. 1). b. [Pointing to a portrait of Saul Kripke:] That is the author of Naming and Necessity. In (1a) the demonstratum is the car key, while the referent is the car. Likewise, in (2b), Kripke himself is being referred to, though the portrait is the object of the demonstration. Other kinds of expressions can also receive deferred readings, such as personal pronouns, proper names6 and definite descriptions.7 According to Mount (2008) the first person singular pronoun is also susceptible to this general phenomenon, and this would show that ‘I’ is not invariably speaker-referential and thus that the Standard View is false. 5 This definition is put forward with an eye to the discussion of cases such as (1), which would appear to be examples of “referential metonymy” (see Stallard, 2003; Warren, 2006; Bezuidenhout, 2008). However, deferral phenomena have not always been discussed in connection with metonymy. In philosophy, the phrases “deferred reference”, “deferred ostension”, and similar ones have frequently been deployed in discussions in which figurative or nonstandard uses are not the main concern. These include debates about the role of deferral phenomena in indexical reference and about ostension as a method of definition and reference-fixation, as well as arguments about the nature of reference and their attending philosophical implications, such as the inscrutability of reference and ontological relativism (Quine, 1968, 1990). In linguistics Nunberg (1979, 1993) took deferral to be an essential component of all indexical/demonstrative reference. A consequence of this is that the import of deferral is not exhausted by its role in figuration. For instance, the interpretation of indexicals, according to Nunberg, involves two stages. The first is the recovery of a semantically indicated element in the context, called the “index” (e.g. the speaker, in ‘we’); the second is the identification of the actual referent or interpretation (e.g. a group appropriately related to the speaker, in the case of ‘we’), which depending on the case may or may not be identical to the contextual element. So, we would have a whole kind of cases in which the argument of the deferral function is not itself a referent, either standard or nonstandard. The characterization in the text could be modified to accommodate these features, but it, as it stands, has the virtue of being both straightforward and sufficient for the goals of this paper. 6[Looking at the Chancellor’s personal art collection. No physical demonstration is involved:] He must be rich! [At a wax museum, a visitor asks the clerk at the information booth about the location of the president’s wax statue, and the clerk responds:] Obama is on the 2nd floor, near the elevator. In many cases, though, deferral for proper names is blocked. See Nunberg (1993) and Powell (1998). 7 a. [Server to co-worker in a diner] The ham sandwich is at Table 7 (Nunberg 1995, ex. 19). b. [Physician to head nurse, at a hospital] The ruptured spleen in 250 needs a nurse. Nunberg (1995) proposes an alternative analysis, in which the element that shifts meaning is the noun phrase (‘ham sandwich’, ‘ruptured spleen’), rather than the whole determiner phrase. 4 To answer this challenge, the defender of the Standard View must provide a plausible alternative analysis of agent-shift cases that secures reference to the speaker while at the same time accounting for the apparent reference shift. This is the task I take on in this paper. To this end, I provide an account of agent-shift cases, which I call the “Description Analysis” (DA). On the DA, seemingly deferred-referential occurrences of the 1st person pronoun are indeed speaker-referential, but they are interpreted as if they were constituents of an implicit definite description, whose denotatum is an object other than the speaker. The net result is that while the (determiner) phrase – i.e. a DP – in which ‘I’ occurs does not designate the speaker,8 pronoun ‘I’ keeps its ordinary reference.9 In the next section I will introduce a bit of handy terminology. Subsequently, I will motivate the DA (3.1) and provide arguments for it (3.2). The case for the DA mainly consists in showing that agent shift sentences exhibit a description-like behavior – such as scope taking and attributive readings – which follows straightforwardly from the DA, but cannot be accommodated under a deferred-reference view of agent shift. In (3.3) will offer some clarificatory remarks on the nature of the proposal. Finally, in section 4 I will discuss Nunberg’s “Predicate Transfer” approach, and will argue that the DA ought to be preferred. 2. – P(hrasal) Shift vs. R(eferential) Shift In this preliminary section I will introduce a couple of distinctions that are necessary to appreciate the differences between rival accounts of agent shift cases, in terms of the possible sources of meaning shift. Limiting ourselves to (complete) declarative sentences, we can understand ‘the ATM swallowed my card’, ‘the White House is worried’ or ‘the ham sandwich wants his check’ as involving a main predicate or relation, 8 I’m using the verb ‘designate’ as a general term for semantic relations through which an expression picks an individual, thus it includes both reference and denotation (in the Russellian sense) and is neutral between them. 9 There are other apparent counterexamples to the Standard View. Among these are examples based on answering machine messages and post-it notes that contain utterances (including both spoken and written ones) where ‘I’ does not refer to the utterer or writer. These cases arise when there is a discrepancy between the context in which the message was coded and that in which is decoded, or because the transmitter or deliverer of the message is not the same as its originator. In general, these cases can be accommodated within the Standard View, and there are several proposals about how to do it, which may be mutually equivalent. One way is by distinguishing between speaker/utterer and agent/author, and tying the reference of ‘I’ to the second. Another strategy is we can call “context shift” the context relevant for the interpretation is different from the context U of utterance, that is, from the context of material production of the utterance. (For a variety of solutions along these lines see Bianchi, 2001; Corazza, Fish and Gorvett (2001); Romdenh-Romluc, 2006; Dodd and Sweeney, 2009). In any case the essence of the character of 'I' is preserved, since the changes are peripheral, or in any case pre-semantic (see Predelli, 2005). Other interesting prima facie counterexamples are (among others) so-called “descriptive uses” (such as the example of the condemned prisoner: ‘I am traditionally allowed to order whatever I like for my last meal’. See Nunberg, 1993) and impersonal uses of the first person singular pronoun, observed, for instance, in languages such as English and German (Zobel, 2010). These cases merit attention, but I won’t discuss them here, as this paper focuses on “deferred reference” cases. 5 which can be expressed as an open sentence, such as x swallowed y’s card, x is worried , x wants y, as well as a set of arguments (The ATM, The White House, The ham sandwich, x’s check), including subjects and objects. Meaning shift can occur in the main predicate or in some of the arguments, or both.10 If the shift takes place in the predicate, we have a case of Predicate Shift. Suppose, for instance, that I utter the sentence, ‘John is at Harvard now’, while my interlocutor and I both mutually know that at the moment John is no more than 20 ft away from us (and that we are in California). In this case it is plausible to say that x is at Harvard, but not ‘John’, underwent a shift in its meaning, where the relation x is at y is interpreted as something like x works at y or x is associated with y. In contrast, the following sentences seem to contain cases of argument shift, where the italicized term is the most obvious shifter: (3) a. Washington fears that the enrichment activities, some of them initially conducted in secret, could service a clandestine bomb program (Time, 8/21/2010). b. This is parked out back [holding up a car key]. The most widely held assumption about Agent Shift cases – which is needed for them to constitute counterexamples to the Standard View – is that they are examples of argument shift. I will go along with this assumption. However, Nunberg (1993, 1995) does offer an alternative analysis in which the meaning transfer observed in agent shift cases is due to predicate shift, not argument shift. I will discuss this view in section 4. Within the category of argument shift, there are further distinctions to make. A particularly important one has to do with the syntactic locus of the shift within the phrase that realizes an argument. Some expressions, which we could call “subphrasal terms” (or simply “terms”), are “atomic” from a syntactic point of view. Let us assume that referential expressions such as names and pronouns are in this category. Thus, a shift due to a change in a term’s reference is an instance of ‘R(eferential)-shift’. Deferred reference is the clearest example of R-shift: for instance, in the example of the car keys (number) the reference of the demonstrative term ‘this’ has shifted away from the demonstratum (the keys) towards a suitable related car. Phrases, of course, are composed out of terms or other phrases. P(hrasal)-shift occurs when the transference originates at the phrasal level, and cannot be traced back to any of the individual terms composing the phrase. Thus, argument shift is not exhausted by R-shift, since phrases constituted by expressions that are not typereferential, such as quantifier phrases, can also shift meaning. For instance in a sentence such as ‘The ham 10 In some cases, such as deferred equatives (‘I am the pad thai’), the shift occurs in the copula, according to Ward’s (2004) analysis. 6 sandwich wants his check’, the locus of shift is the whole descriptive phrase, since none of its constituents seems to have undergone a shift in meaning.11 The distinction between P-shift and R-shift serves to differentiate my position from those of other authors who agree that agent shift cases are characterized by Argument Shift, but who classify them as instances of “deferred reference”. In particular, the deferred reference challenge proposed by Mount and others assumes that the locus of shift is the pronoun ‘I’, and thus that the cases at issue are examples of R-shift. In this paper I reject this assumption, and argue instead that the shift is localized at a purely phrasal level (P-shift). More specifically, I will argue that the shift originates, not in ‘I’, but in [DP I], a phrase that contains the occurrence of the pronoun. The justification of this claim depends on the Description Analysis of agent shift cases, which will be introduced in the next section. Before I finish this section, I would like to make a terminological parenthesis. As mentioned earlier, the examples in (1) are prima facie cases of metonymy, so for the rest of the discussion it will be convenient to borrow some nomenclature from the literature on this phenomenon. In particular, I will adopt part of Fauconnier’s (1985) terminology (though not his analyses). So, in the case of a particular phrase P used metonymically: the trigger is the conventional interpretation of P, i.e. some object, property, relation or event A; the target is the intended interpretation of P, i.e. some object, property, relation or event B distinct from, but related to A. In (1), the trigger is invariably the agent of the context, i.e. the speaker of the utterance, whereas the target and trigger-target relation vary from case to case. 2.0 THE DESCRIPTION ANALYSIS 2.1 Agent-Shift Cases and Possessives There are striking similarities between the interpretation of agent-shift utterances and the interpretation of utterances containing first-person possessives in argument positions. For every example in (1), a speaker can express the same idea by substituting the first person pronoun (the trigger) with a DP of the form My x, where x stands for an NP containing a relevant predicate satisfied by the target.12 E.g.: 11 In this example the shifting phrase contains a noun-noun compound, but this is not the only configuration in which P-shift can occur, as we’ll see later. According to certain analyses (Nunberg, 1993, 1995; Sag, 1981) the shifter in this case is the NP ‘ham sandwich’, rather than the whole DP. In this case, we have an instance of “common noun shift”, which can be regarded as a shift in a predicate that sits within the argument. 12 The sentences could be paraphrased in different, equally acceptable ways, but at this stage of the discussion this should not be a worry. 7 (1a) I am next to Chaplin. (1a') My statue is next to Chaplin. (1b) I am on the purple square. (1b') My piece is on the purple square. (1c) I finished last in the race. (1c') My horse finished last in the race. (1d) How am I doing today? (1d') How is my double doing today? (1e) I haven’t read myself in a long time. (1e') I haven’t read my own work in a long time. This systematic correspondence suggests that both classes of phenomena are candidates for receiving uniform treatment. A phrase such as My x can be interpreted as picking something (the “possessee”) which stands in a relevant relation R to the possessor – the speaker, in this case. In these phrases the possessee is marked as definite.13 This can be clearly seen in this non-agent shift example (taken from Barker, 1995, p. 78 ff): (4) a. I saw a child. b. I saw the child. c. I saw John’s child. In (a), the complement is an indefinite phrase, and there is no presumption that the child in question would be identifiable or contextually unique. However, in (b), the definite version, the child must be contextually unique or identifiable. Sentence (c) patterns with (b) rather than (a) in this respect, so possessives seem to be closely associated with definiteness (compare, for instance ‘my child’ with ‘a child of mine’). In fact, possessives have traditionally been analyzed as definite descriptions,14 so that a sentence such as (5') is interpreted in the manner of (6), where R is a relation holding between the possessor and the possessee: 13 That is, as involving uniqueness relevant to the context (or familiarity/salience/givenness/accessibility, etc.). Feel free to substitute with your favorite view about the nature of definiteness. 8 (5) [To a parking lot attendant]: (5') My car is parked out back. (6) [The x: x is a car and x stands in R towards me] x is parked out back. This makes it explicit that there is a car that uniquely stands in a relation R towards the speaker. Let us reconsider, then, the sentences in (1). Take (1a'), repeated here for convenience: (1a') My statue is next to Chaplin. According to the definite description view of possessives (1a') is equivalent to: (1a'') The statue that stands in R to me is next to Chaplin. Now, since (1a) – in the intended interpretation – and (1a') are interchangeable, and since (1a') has the same truth conditions as (1a''), then an utterance of (1a) – again, with the intended reading – will be interpreted as an equivalent of (1a''). Suppose that the value of R in this case is something like ‘x represents y’ or ‘y is represented by x’, then the interpretation of (1a) would be expressed by (7) or the more English-like (8): (7) [The x such that x is a statue and I am represented by x] x is next to Chaplin. (8) The statue that represents me is next to Chaplin. Similarly, as pointed out earlier, (1b) and (1b') are interchangeable in the context at issue: (1b) I am on the purple square. (1b') My piece is on the purple square. And since (1b') can be interpreted as a description, as in (1b''), the same goes for (1b): 14 See, for instance, Russell’s (1905, p. 484) analysis of the denotation of “my only son”, as well as Donnellan’s (1966) famous discussion of cases like “Smith’s murderer” and “her husband”.