Abstract

This article aims to show that (one of) the main argument(s) against the presuppositional account of person is not compelling if one makes appropriate assumptions about how the context fixes the assignment. It has been argued that unlike gender features, person features of free pronouns cannot yield presupposition failure (instead, can yield only falsity) when they are not verified by the referent. The argument is flawed, however, because the way the referent is assigned is not made clear. If it is assumed to be the individual that the audience can recognize as the referent intended by the speaker, the argument is reversed.

Since Cooper 1983, gender features on pronouns have been standardly analyzed as presuppositions (e.g., Heim and Kratzer 1998, Sauerland 2003, Heim 2008, Percus 2011). Cooper’s analysis of gender features has been extended not only to number, but also to person features (e.g., Heim and Kratzer 1998, Schlenker 1999, 2003, Heim 2008, Sauerland 2008), thus replacing the more traditional indexical analysis of first and second person pronouns (Kaplan 1977 and its descendants). All pronouns are thereby interpreted as variables, and all ϕ-features are assigned uniform interpretive functions; that is, they introduce presuppositions that restrict the value of the variables.

However, empirical arguments have been provided that cast doubt on the presuppositional nature of person features (e.g., Stokke 2010, Sudo 2012). In particular, it has been claimed that when the person information is not verified by the referent, the use of a first or second person pronoun does not give rise to a feeling of presupposition failure, as is the case when the gender information does not match the referent’s gender: a person mismatch, unlike a gender mismatch, yields a plain judgment of falsity. This suggests either that the indexical analysis, according to which first and second person pronouns directly refer to the speaker and addressee of the context, should be rehabilitated (but several other arguments militate against it1), or that a different, new analysis of person features should be provided (e.g., Stokke 2010, Sudo 2012).

The goal of this article is to show that such empirical arguments provided against the presuppositional analysis of person features are not compelling if one makes appropriate assumptions about how the context fixes the assignment. Thus, a presuppositional account of person is (maybe2) not to be discarded, after all.

In section 1, I will review the examples taken to argue against a presuppositional analysis of person features. In sections 2 and 3, I will explain the fallacy of this argument for first person, providing more controlled examples that in fact support the presuppositional account of person. In section 4, I will extend the argumentation to second person. Section 5 concludes.

1 The Argument against the Presuppositional Account of Person

Under a presuppositional account of ϕ-features, all pronouns (including so-called indexicals, i.e., first and second person pronouns) contain variables that are interpreted via an assignment function. The ϕ-features of pronouns, which combine with the variable, introduce presuppositions. Specifically, ϕ-features denote partial identity functions, as illustrated in (1) for masculine and feminine, and in (2) for first and second person (number will not be discussed here). When the ϕ-features combine with the variable, this has the effect of constraining the possible values for the variable. Note that the exact formulation of such presuppositions is debated,3 as is the necessity of postulating presuppositions for third person,4 but these issues are orthogonal to my argumentation.

(1)

  • a.

    graphic
    masc
    graphic
    = λxe: x is male. x

  • b.

    graphic
    fem
    graphic
    = λxe: x is female. x

(2)

  • a.

    graphic
    1st
    graphic
    c = λxe: x is the speaker of c. x

  • b.

    graphic
    2nd
    graphic
    c = λxe: x is an addressee of c. x

According to this type of theory, gender and person features do not contribute to assertive meaning. Instead, the result of using these features is to impose a condition: the information they encode about the pronoun’s referent must be assumed to be already true, or at least uncontroversial, in the conversational context.5 In contexts where this condition is not met, it is predicted that a feeling of presupposition failure will arise—a feeling that the speaker has started from wrong assumptions about gender or person information. This feeling is not sufficient to trigger a judgment of falsity, or a judgment of undefinedness for that matter: the sentence will neither be judged false on this basis, nor necessarily be judged undefined, to the extent that an assertive meaning can be derived by stripping the presupposition-inducing features from the sentence. That this indeed happens is illustrated in (3) for gender.

(3)

  • Context: A baby, who is a boy, is sleeping. Mary thought he was a girl, and said the following:

  • She is sleeping.

  • (Sudo 2012:142)

In that context, what Mary asserts is true (the baby is sleeping), but she is wrong about the gender of the baby (she thinks it is a girl, while it is in fact a boy). Crucially, the sentence is not judged false in that case; instead, it is judged true, and the mismatch between the referent’s gender and the pronominal gender results in a feeling of “squeamishness” (the term is used by Sudo (2012), who borrows it from von Fintel (2004)). The truth of the sentence can thus be assessed independently of the truth of the gender information, and the falsity of the latter does not affect the assertive meaning, but yields a feeling of presupposition failure.

If person features are presuppositional just like gender features, cases similar to (3) involving first person pronouns should be found. However, on the basis of (4) and (5) Stokke (2010) and Sudo (2012) argue that this is not the case.

(4)

(5)

  • Context: David is looking at a shop window. On the other side of the glass, there is a man who looks just like him, and David therefore mistakes the window for a mirror. Suddenly, he notices that the person’s pants are on fire and exclaims:

  • My pants are on fire!

  • (Sudo 2012:143, from Stokke 2010:98, adapted from Kaplan 1977 (1989:533))

In both of these contexts, Stokke and Sudo argue, the speaker refers to someone wrongly, thinking that that person is he himself: the information conveyed by the first person feature is thus wrong. Under the presuppositional account of person, this should give rise to a feeling of squeamishness without affecting the assertive meaning. Specifically, sentences (4) and (5) should be judged true (since Napoleon indeed won the Battle of Austerlitz and David’s double’s pants are indeed on fire) and associated with a feeling of presupposition failure. But Stokke (2010) and Sudo (2012) point out that this is not what we observe: sentences (4) and (5) are judged plainly false. This shows, according to them, that person features are not presupposition triggers.

The argument is flawed, however. As I will show in the next section, Stokke’s and Sudo’s conclusion does not hold if one makes appropriate assumptions about how the context fixes the assignment with respect to which a sentence is evaluated.

2 The Fallacy of the Argument

First, let us again compare (3) and (4), which are repeated here. The trickier example (5) will be examined in section 3.

(3) She5 is sleeping.

(4) I6 won the Battle of Austerlitz.

The assertion and the presupposition of these sentences can be represented as follows, where g is the assignment function with respect to which the sentence is evaluated:

(6)

  • a.

    Assertion: g(5) is sleeping.

  • b.

    Presupposition: g(5) is female.

(7)

  • a.

    Assertion: g(6) won the Battle of Austerlitz.

  • b.

    Presupposition: g(6) is the speaker of the context.

These representations make clear that for Stokke’s (2010) and Sudo’s (2012) argument to hold, it must be assumed that g returns the baby boy for 5, and Napoleon for 6. Only then can we argue against the presuppositional theory, which should in both cases predict a judgment of truth associated with presupposition failure.

There is, however, an important contrast between the two sentences in this respect: while it seems uncontroversial to assume that the referent of she is the baby in the context of (3), it is much more doubtful to suppose that the referent of I is Napoleon in the context of (4). The reason is that the baby boy is clearly the salient and intended referent in (3), as opposed to Napoleon—who is not even present in the context—in (4). Crucially, if g is assumed to return André for 6 in (4), the plain judgment of falsity follows from the presuppositional theory (since André did not win the Battle of Austerlitz and he is the speaker), and (4) is no longer an argument against this account.

The issue is thus to specify how the context fixes the assignment: who do we take a given pronoun to refer to? This question is not much discussed in the semantic literature, but it seems reasonable to assume that “the relevant assignment is given by the utterance context and represents the speaker’s referential intentions” (Heim 2008:36; cf. Schlenker’s (2003:51) Compatibility Condition) and, more specifically, that “[the speaker] succeeds in referring to x by uttering referential expression e in context c only if, partly as a result of their recognizing [the speaker]’s intention, [the speaker]’s audience in c are in a position to interpret [the speaker] as referring to x with e” (Stokke 2010:34; also see works by other philosophers, such as Neale 2005).6 In other words, the individual that we (as an audience) take a given pronoun to refer to is the individual that we can and are willing to recognize as the referent intended by the speaker on the basis of contextual clues. As a rule of thumb, I assume that we can take the referent of a pronoun to be the individual that the speaker is pointing at, or that we imagine the speaker would have pointed at if she had had to point.

Thus, the question we need to answer regarding (3) and (4) is who would have been pointed at if gestures were added to these contexts. Clearly, Mary intends to refer to the baby in (3), so if she had to point, she would point at the baby boy. What about (4)? It seems just as clear that André intends to talk about himself since there is no other potential candidate referent in the context. Certainly, André wrongly assigns to himself the property of winning the Battle of Austerlitz—that is, of being Napoleon. But this does not mean that he intends to refer to Napoleon. Even if André and Napoleon are the same person according to André, the individual that the audience, who does not have this false belief, can recognize as the intended referent is André, the speaker: if he had to point, André would point at himself while claiming, “I won the Battle of Austerlitz!”

Example (4) is therefore not comparable to example (3), and we cannot conclude anything from the judgment about it regarding the presuppositional account of person. In order to construct a case more similar to (3) for person, we need to find a scenario where the speaker intends to refer to someone who he thinks is the speaker, while it is in fact not the case. This is tricky given that usually, the speaker obviously knows who the speaker is. But what will work is a scenario that involves not ( just) the intended referent but (also) a representation of that referent. For instance, we can refer to an individual by attending to someone understood as posing as that individual, to a picture or statue depicting that individual, or to a TV or any other device that has recorded images or sounds produced by that individual.7 A simple case (involving a third person pronoun with no gender or person mismatch) is shown in (8).

(8)

  • Context: Mary is watching a documentary about Napoleon on TV and when Napoleon appears on the screen, she exclaims:

  • He was such a brave man!

Here, Mary uses the pronoun he to refer to Napoleon even though Napoleon himself is not present in the context. Only a representation of him (on two levels) is present: the TV screen that Mary could point at while uttering he reproduces the image of an actor posing as Napoleon. This means that the strategy proposed above to identify the intended referent should be made slightly more complex: we should take the referent of a pronoun to be the individual—or the representation of the individual—who the speaker is pointing at or who we imagine she would have pointed at if she had had to point.

The same holds for first person pronouns: there is nothing wrong in principle with referring to oneself, the speaker, as I, while pointing not at one’s own body, but at a representation of oneself, as in (9).

(9)

  • Context: André has run a marathon. He is watching the video of that marathon on TV and when he appears on the TV screen, he points at himself on the screen and exclaims:

  • Look how fast I6 can run!

Here, André does not point at himself speaking; rather, he points at a representation of himself on the TV screen. Still, the use of I is perfectly felicitous: if the speaker is attending to a recording or an image (or any other kind of representation) of himself, then we can naturally understand the first person pronoun he uses to refer to himself.8

Given this, we can construct a scenario where the speaker intends to refer to someone who he thinks is the speaker, but who the audience cannot recognize as such because the speaker is in fact mistaken, as in (10).

(10)

  • Context: André has gone mad and thinks he is Napoleon. He is watching a documentary about Napoleon on TV and when Napoleon appears on the screen, he points at him and claims:

  • I6 won the Battle of Austerlitz!

In this context, just as in (4), André wrongly attributes to himself the property of being Napoleon. But this time, he does not point at himself, but at the representation of Napoleon on the TV screen. The individual that the audience can recognize as the intended referent is therefore not André, but Napoleon. In that case, the presuppositional theory of person features makes the following prediction: the assertion in (10) is true (Napoleon indeed won the Battle of Austerlitz), but the presupposition is false (Napoleon on the screen is not the speaker). Crucially, the sentence is accordingly judged to be true in this case, but is associated with a feeling of squeamishness, which is consistent with the presuppositional account of person features.

Note that in both (4) and (10), the speaker André thinks that Napoleon and he himself are the same person, so we could at first glance think that by referring to one, he equally refers to the other. This is so from his perspective, but this is crucially not so from the audience’s perspective: the audience needs to recognize who the intended referent is, but does not have the false belief that André and Napoleon are one and the same person. This is what introduces a crucial difference between (4) and (10): in the former case, André is clearly talking about himself (we can imagine he would point at himself ), thinking that he is Napoleon; in the latter case, André is clearly talking about Napoleon (he points at the representation of Napoleon), thinking that Napoleon is he himself. Under the presuppositional theory of person features, this means that the presupposition is violated only in (10), while the assertion is false only in (4). The judgments of these sentences do match these predictions. This example thus confirms that the referent of a pronoun is taken to be the individual recognized as the referent intended by the speaker, even if the referent does not satisfy the presupposition. This means that an audience is willing to ignore a presupposition and instead follow the speaker’s referential intention if the audience recognizes the presupposition to be wrong because of the speaker’s mistake.

Conversely, if we construct a scenario comparable to (4) for gender features, we get the same judgment of falsity as in (4). To this end, we need to imagine a situation where the speaker is also mistaken about the identity of the referent, but where this time that referent does not verify the first person feature (the referent is the speaker), but instead verifies the feminine gender feature (the referent is female): instead of taking himself, the speaker, to be Napoleon, he takes a salient woman to be so. This is illustrated in (11).

(11)

  • Context: André has gone mad. He has gotten it into his head that Napoleon is female and that Mary, who is next to him, is Napoleon. He points at her and claims:

  • She7 won the Battle of Austerlitz!

(12)

  • a.

    Assertion: g(7) won the Battle of Austerlitz.

  • b.

    Presupposition: g(7) is female.

Just as in (4), the information conveyed by the ϕ-features of the pronoun is satisfied in (11): the intended referent, Mary, is female, just as the intended referent in (4), André, is the speaker. However, the assertion is false: Mary, just like André, did not win the Battle of Austerlitz. Crucially, a judgment of plain falsity also obtains for (11), as the presuppositional theory predicts.

Conversely, the counterpart of (10) with respect to gender features gives rise to the same judgment of truth associated with a feeling of unease.

(13)

  • Context: André has gone mad and thinks that Mary, who is next to him, is Napoleon. He is watching a documentary about Napoleon on TV and when Napoleon appears on the screen, he points at him and claims:

  • She6 won the Battle of Austerlitz!

Just like (10), (13) creates a feeling of squeamishness because the speaker André has started from wrong assumptions (since he thinks that Mary is Napoleon): the individual whose representation on TV André is pointing at is not a woman as predicted by the gender feature of the pronoun, but a man. Nevertheless, the audience can ignore this presupposition violation (which gives rise to a feeling of unease) and judge the sentence true.9

In sum, the argument against the presuppositional theory presented in section 1 is flawed, because it makes an unwarranted assumption about the referent of the pronoun. But the argument disappears (and is even reversed10) once we clarify how the context fixes the assignment: the referent of a pronoun can be reasonably assumed to be the individual that the audience can recognize as the referent intended by the speaker on the basis of contextual clues, that is, the individual (or the representation of the individual) who the audience can imagine the speaker would point at if she had to point.

Admittedly, it is much trickier to test the presuppositional account for person features than for gender features by constructing what is predicted to be a presupposition failure: the first person information (the referent is the speaker) is much harder to violate than the feminine gender information (the referent is female). This is so because the choice of referent should correspond to the speaker’s intentions, and while the speaker can easily be mistaken about the gender of another individual, it is much harder to find cases where the speaker could be mistaken about whether an individual is the speaker. Thus, person presupposition failure necessarily involves mistaken identity, while gender presupposition failure solely requires mistaken gender. Examples and (11) nevertheless demonstrate that the two kinds of features can be tested in a parallel fashion. Moreover, cases of mistaken identity can be tested more easily because the speaker can use a pronoun to refer to an individual by attending not only to that individual, but also to a representation of that individual. In particular, we can imagine that the speaker using I would in most cases point directly at himself pronouncing the sentence, but in some cases he could point at a salient and relevant representation of himself.

However, most of these examples involve very uncommon scenarios, where the speaker is taken to be mad: only the case of gender presupposition failure is illustrated with an example from everyday life (see (3)). To further show that person and gender features exhibit the same behavior, let me conclude this section by mentioning two more common and parallel examples involving violation of gender and person information.

(14) Context: Tom is listening to amateur recordings of songs that he and his friends performed.

  • a.

    The first song is sung (too fast) by Tom’s friend Carl, but because of the bad quality of the recording, Tom mistakes Carl’s voice for his own voice and says:

    I sang very fast.

  • b.

    The second song is sung (too slowly) by Liz, but because of the bad quality of the recording, Tom mistakes her voice for Carl’s and says:

    He sang very slowly.

In both cases, the intended referent is clear: it is the person whose voice is heard on the recording11 (we could imagine that Tom would point at the recording device12). Moreover, in both cases the ϕ-feature information is wrong for the same reason: Tom mistakes a voice for another voice. The two examples are thus fully parallel. Crucially, the judgment that obtains is also uniform: both sentences are judged true and are associated with a feeling of squeamishness. This again supports the presuppositional theory of ϕ-features, which predicts presupposition failure of the first person feature in (14a) and of the feminine gender feature in (14b), given that the feeling of squeamishness is arguably a feeling of presupposition failure.

3 The Case of the Window Mistaken for a Mirror

In section 2, I have shown that under closer scrutiny, it is not adequate to make the case against the presuppositional account for person on the basis of (4). What about (5), repeated here?

(15)

  • Context: David is looking at a shop window. On the other side of the glass, there is a man who looks just like him, and David therefore mistakes the window for a mirror. Suddenly, he notices that the person’s pants are on fire and exclaims:

  • My8 pants are on fire!

  • (Sudo 2012:13, from Stokke 2010:98, adapted from Kaplan 1977 (1989:533))

(16)

  • a.

    Assertion: g(8)’s pants are on fire.

  • b.

    Presupposition: g(8) is the speaker.

The judgment is not so clear in (15): while Sudo (2012:143) reports the sentence to be “unambiguously judged false,” Stokke (2010:99) claims that the intuitions are not as stable as in (4): even if “it is likely that most people will judge the utterance false, . . . some might judge the utterance true.” Stokke’s observation corresponds to my own intuitions and my consultants’: the sentence can be judged plainly false, and the audience could answer, “No! Your pants are not on fire, look!”; or it can be judged true but be associated with a feeling of squeamishness, and the audience could answer, “Well yes, this person’s pants are on fire, but look, it’s not you; it’s a window, not a mirror!”

The scenario is complicated by the presence of a window mistaken for a mirror. Importantly, this implies that from the speaker’s perspective (i.e., David’s), there is only one candidate referent (i.e., himself ), but from the audience’s perspective, there are two (i.e., David and his double). In that situation, who should be taken to be the referent of my in (15)?

At first glance, this situation seems similar to that in (10) involving the speaker André and what he thinks is his representation on TV (it is in fact Napoleon). But the nature of the mirror complicates the question of who David would point at if he had to point: crucially, both options are possible in (15) as opposed to (10): that is, David could point either at his own body or at his supposed mirror image. This difference is due to the specificity of a mirror image: unlike the recorded image of a person when it is shown on TV (unless it is broadcast live), a mirror representation is not displaced in time, so that what can be seen to happen to the person herself can also be seen to happen simultaneously to her mirror image. Therefore, while saying “My pants are on fire!” to talk about what is happening to him at the moment of utterance, David could point either at his supposed reflection in the mirror (since the mirror is where he has been able to realize this first) or at his own body—since what he can see happening in the mirror happens in the real world simultaneously, and this is where he can do something about it (and in this case especially, it is more urgent to worry about his real body). This double possibility can be confirmed by cases from everyday life: if I notice in a mirror that I have a stain on my shirt, I can either look further at it in the mirror—and point at it on the mirror (especially if I cannot see it otherwise—say, if it is on my back) or instinctively look at the shirt itself and point at my own body.

Since David’s pointing is ambiguous in scenario (15), we correctly predict that the judgment of the sentence is unstable. If the audience understands that David mistakes the window for a mirror and imagines that David would point at what he thinks is his mirror image, which is in fact his double, the sentence will be judged true and associated with a feeling of unease: since the referent of I is taken to be David’s double, the person presupposition is violated, but the fact that the intended referent’s pants are on fire is true. If, however, the audience imagines that David would point at his own body (whether David’s mistake about the window is understood or not), the sentence will be judged plainly false with no presupposition failure: David is indeed the speaker, but David’s pants are not on fire. In sum, such an example again supports the presuppositional account of person once the complication introduced by the window mistaken for a mirror has been clarified.

In fact, we can construct a comparable scenario for gender features and get the same unstable judgment.

(17)

  • Context: David is looking at a shop window, and Mary is 30 feet away from him. On the other side of the glass, there is a man who looks just like Mary, and David therefore mistakes the window for a mirror. Suddenly, he notices that the person’s pants are on fire and exclaims:

  • Her3 pants are on fire!

(18)

  • a.

    Assertion: g(3)’s pants are on fire.

  • b.

    Presupposition: g(3) is female.

Here, the speaker David is also mistaken about the identity of the referent, which does not affect the person information this time, but the gender information: Mary’s double is not female, but male. Just as before, we can imagine either that David would point at the mirror image while uttering the sentence, or that he would quickly turn to Mary herself and point at her. In the former case, the sentence is judged true with a feeling of presupposition failure; in the latter case, it is judged plainly false.

There is nevertheless a slight difference in judgment between (15) and (17): we are more prone to judge the sentence false in the first case than in the second. Crucially, this is not due to a difference in status between person and gender information; rather, it is due again to the specificity of the mirror case. When I am looking at myself in a mirror and notice something unusual, I tend to look at myself to do something about it, rather than at my mirror image (if I can—and clearly, this is possible in the above scenario since I can look at my pants). In fact, this is used in the mirror test (developed in 1970 by psychologist Gordon Gallup Jr.) as a measure for self-recognition in children: if an experimenter surreptitiously places rouge on the nose of a child and then places him in front of a mirror, the child is assumed to understand self-awareness if he touches his own nose. It is therefore expected that the audience should be more prone to imagine that David would point at himself in the scenario in (15) and accordingly judge the sentence false. The situation is different if I look at someone else’s reflection in a mirror. First, since in principle I can easily look at the person herself, I need a specific reason to look at the mirror image instead. This may be because I cannot see the person herself for some pragmatic reason—in that case, I cannot point at her—or because it just happens that I have noticed something unusual about the mirror image first. Second, since I myself am not the person concerned, there is less instinctive urge to turn to the real person instead of the mirror image. It is thus expected that in (17), the audience will be more prone to imagine that David would point at Mary’s double (who he thinks is Mary’s reflection) and accordingly judge the sentence true with presupposition failure.

In sum, the tricky mirror scenario confirms that we take the referent of a pronoun to be the individual that we recognize to be the referent intended by the speaker, and a practical strategy to identify this individual in murky cases is to imagine who (or what) the speaker would have pointed at if he had had to. If the judgment is unclear in the mirror scenario, it is because David’s pointing can be thought of as ambiguous in that case.

Other cases that can give rise to ambiguous judgments are those where the contextual clues are unclear so that the audience does not have enough evidence to recognize the speaker’s intentions (which are, by their very nature, not directly accessible to the audience). This was not the case in the previous scenario, where the resemblance between David and his double and the possible mirror function of a window were taken as sufficient contextual clues showing that David mistook the window for a mirror (if these clues were not recognized by the audience, sentence would simply be judged false). However, this is the case in (19), which focuses on gender information.

(19)

  • Context: Mary enters a room where a baby boy is sleeping and a baby girl is playing next to him. While seemingly looking in the direction of the baby boy, Mary says:

  • She9 is sleeping.

(20)

  • a.

    Assertion: g(9) is sleeping.

  • b.

    Presupposition: g(9) is female.

Just as in (15)/(17) and (10)/(13), there are two candidate referents and thus two competing options here: the first one makes the sentence true but violates the presupposition (g(9) is the baby boy); conversely, the second one satisfies the presupposition but makes the sentence false (g(9) is the baby girl). The judgment of the sentence thus depends on which baby the audience can recognize as the speaker’s intended referent. If Mary’s gaze was totally unclear, this would depend on what the audience thinks about Mary’s state of knowledge. To the extent that the audience thinks it likely that Mary is aware of which baby is the boy and which baby is the girl, the audience would guess that Mary means to refer to the baby girl: the presuppositional content would be used as a clue for identifying the referent. But if the audience does not know anything about Mary’s state of knowledge and thinks that one baby is quite obviously asleep and the other baby is quite obviously awake, the audience would probably guess that Mary means to refer to the baby boy. In the latter case, the audience would thus take the referent to be the referent that makes the sentence true: interestingly enough, this means that in the absence of clearer contextual clues, the assertive meaning itself can serve to identify the speaker’s referential intentions. In scenario (19), where Mary seems to be looking at the boy, but her gaze is not totally clear, the judgment is as unstable as in (15): the audience could imagine she would point at the boy and then tell her, “This baby is indeed sleeping, but in fact it’s a boy!,” thereby taking the sentence to be true with presupposition failure; or the audience could imagine she would point at the girl and exclaim, “No, look, she’s playing!,” thus taking the sentence to be false. Note that the audience could also simply judge the sentence to be undefined and ask Mary who she intends to refer to, just as André or David could be asked, “But who are you talking about?” All these judgments are consistent with the presuppositional theory of ϕ-features and in fact support it.

In sum, we have seen in this section that even trickier examples with unclear judgments do not militate against the presuppositional account of person; on the contrary. There are two main reasons for this. First, person and gender information give rise to the same effects. Second, when the person information does not match the referent’s person (the referent is not the speaker), this yields a feeling of squeamishness; some work needs to be done by the audience to deal with the mismatch, which is a hallmark of presupposition failure. These two arguments for the presuppositional account of person can be made once we make appropriate assumptions about how the context fixes the assignment: the audience uses contextual clues and what is known about the speaker’s state of knowledge to recognize the speaker’s referential intentions. Unclear judgments arise when the contextual clues are ambiguous and contradictory; the person feature is one of these clues—besides gestures, salience, world knowledge, and specifics of the situation, as well as the assertive meaning in some cases—but does not override the others since it is the speaker’s referential intention, not the presupposition, that primarily matters for the purposes of assignment.

4 Second Person

Examples involving second person have been provided to make the same point against the presuppositional account of person. It has been argued that just as a first person pronoun necessarily refers to the speaker (there cannot be any presupposition failure), so a second person pronoun necessarily refers to the addressee. The only example given by Sudo (2012) is (21), parallel to (15).

(21)

  • Context: Saul and David are sitting on a sofa. In front of the sofa is a glass pane. Opposite is another sofa. The sofas look exactly the same and two men dressed like David and Saul are sitting in the seats opposite to them. So, from where David and Saul are sitting, the pane looks like a mirror. Suddenly, David realizes that the pants of the person sitting opposite Saul are on fire. So he turns to Saul and warns him:

  • Your3 pants are on fire!

  • (Sudo 2012:143, from Stokke 2010:99)

(22)

  • a.

    Assertion: g(3)’s pants are on fire.

  • b.

    Presupposition: g(3) is the addressee.

Similarly to (15), Sudo (2012:143) judges the sentence in (21) to be plainly false, rather than true with a feeling of presupposition failure, and Stokke (2010:100) finds it likely that “most people will deem David’s utterance false.”

In fact, the judgment is clearer in (21) than in (15): I agree that this sentence is judged false. But this does not make a better point against the presuppositional account of person: the difference in judgment follows from a difference in the situation. This can be made clear if we add gestures to the context: in (21), David would have to point at Saul since he is said to “turn to Saul”; but in (15), we saw that it would be possible for David to point at what he thinks is his mirror image (in fact, his double). This means that while the intended referent in (15) could be either David or David’s double because of the ambiguous pointing, the intended referent in (21) has to be David’s addressee, Saul; it cannot be Saul’s double. If David was not explicitly said to turn to Saul, the judgment would be as unstable as in (15) and (17), and the same argumentation could be made, showing that this example does not argue against the presuppositional account of person once we are precise about how the context fixes the assignment.

In fact, all the points made in the previous sections also hold for second person. Thus, the contrast between (23) and (24) is parallel to the contrast between (4) and (10).

(23)

  • Context: André has gone mad and thinks his friend Sam is Napoleon. When Sam tries to calm him down, he retorts:

  • You won the Battle of Austerlitz!

(24)

  • Context: André has gone mad and thinks his friend Sam is Napoleon. He is watching a documentary about Napoleon on TV and when Napoleon appears on the screen, he points at him and exclaims:

  • You won the Battle of Austerlitz!

While (23) is judged false because the referent of you is clearly Sam, André’s addressee, (24) is judged true with a feeling of presupposition failure, because the referent that the audience recognizes as the intended referent is Napoleon on the TV screen. Thus, the second person feature can be taken to be wrong, and in that case, the same feeling arises as feelings typically associated with presupposition failure.

Note that (24), just like (10), should also be coupled with (25), which is parallel to (9) and shows that, like a first person pronoun, a second person pronoun can be used to refer to the addressee not only in the presence of the addressee, but also in the presence of a representation of the addressee.

(25)

  • Context: André is proud of his friend Sam, who has run a marathon. André and Sam are watching the video of that marathon on TV and when Sam appears on the TV screen, André points at him on the screen and exclaims:

  • Look how fast you can run!

Finally, more natural examples yielding presupposition failure can also be constructed with second person; this is shown by (26), parallel to (14).

(26)

  • Context: Tom and Sam are listening to amateur recordings of songs that they and their friends performed. The first song is sung (too fast) by Tom’s friend Carl, but because of the bad quality of the recording, Tom mistakes Carl’s voice for Sam’s voice and says:

  • You sang very fast.

Here, the intended referent is clearly recognized as the person whose voice is heard on the recording. The sentence is thus judged true, with a feeling of presupposition failure: the hearer Sam can accept the truth of the sentence and signal the presuppositional mistake by answering, “This person indeed sang very fast, but you’re mistaken; it’s not me.”

All these examples buttress the point already made by the previous examples involving first person: far from making a case against a presuppositional account of person, they in fact support it by illustrating that the person information can be taken to be wrong without affecting the assertive meaning, and this yields a feeling of squeamishness just like other presuppositions.

5 Conclusion

In this article, I have argued that (one of) the main argument(s) provided (by, e.g., Stokke (2010), Sudo (2012)) against the presuppositional account of person, which is based on examples involving free readings of first and second person pronouns, is not compelling. These examples are meant to show that first and second person pronouns directly contribute to the truth conditions of the sentence—that is, that person features, unlike gender features, are not presupposition triggers. But in fact, the argument is flawed because the way the context fixes the assignment is not made explicit. Once we make appropriate assumptions about this, we can see that such examples do not make the point they are supposed to make, and other, more controlled examples actually make the opposite point: just like third person pronouns with respect to gender, first and second person pronouns can give rise to presupposition failures, a result that supports the presuppositional theory of person features.

In the course of the argumentation, two other points have also been clarified. First, I have demonstrated that Stokke’s (2010) intention-sensitive principle about reference is adequate: the referent assigned to a given pronoun is the referent that the audience can recognize as the referent intended by the speaker using that pronoun; a concrete strategy to identify it is to imagine who the speaker would have pointed at if she had had to point on the basis of the available contextual clues. Unstable judgments arise when the contextual clues are unclear or complex so that the audience cannot unambiguously decide between two candidate referents. Second, I have shown that first and second person pronouns, just like names or third person pronouns, can be used to refer to the speaker and the addressee not only when attending to the corresponding individuals themselves, but also when attending to representations of those individuals, displaced in location (in cases involving mirrors or live broadcasts) and/or in time (in cases involving videos or pictures). This implies that first and second person pronouns can be used even if they are accompanied by a gesture that does not point at the speaker or the addressee’s body, but instead points at some representation of it. Methodologically, this leaves room for a way to construct presupposition failures of person almost just as easily as presupposition failures of gender, even if it is intrinsically easier for the speaker to be mistaken about the gender of another person than about the identity of the speaker or the addressee.

Notes

1 In particular, the ability of first and second person pronouns to be bound in English focus constructions (e.g., Heim 2008) and to shift their reference in attitude contexts in some other languages (e.g., Schlenker 2003) challenges Kaplan’s (1977) fixity thesis (see Sudo 2012 for a review of the arguments for and against the indexical and variable binding analyses of first and second person pronouns).

2 To fully support the presuppositional account of person, arguments that have been provided against it based on failure of bound readings in examples such as (i) should also be reviewed. This is not the goal of this article, which only aims at refuting the argument based on free readings. Importantly, this argument has been claimed to be “potentially more serious” (Sudo 2012:142). The lack of a bound reading in (i) could indeed be explained by adding an independent constraint—for instance, the prohibition against using a third person expression to refer to the speaker (Heim 2008:40).

  • (i)

    #Exactly one student did my homework (namely me). (Sudo 2012:140)

3 For instance, the following presuppositions have been proposed by Heim (2008), among others, to accommodate plural pronouns:

  • (i)
    • a.

      graphic
      1st
      graphic
      c = λxe: x includes the speaker of c. x

    • b.

      graphic
      2nd
      graphic
      c = λxe: x includes the addressee of c and excludes the speaker of c. x

4Sauerland (2003) proposes that third person does not denote the partial function in (ia) as is often assumed, but that it instead denotes the unrestricted identity function in (ib).

  • (i)
    • a.

      graphic
      3rd
      graphic
      c λxe: x excludes the speaker and the addressee of c. x

    • b.

      graphic
      3rd
      graphic
      c λxe. x

      For Sauerland, the use of third person to refer to individuals other than the speaker and addressee comes about because of the principle Maximize Presupposition, which says that a speaker must always choose the person feature with the strongest presupposition she can felicitously make.

5 Since Stalnaker 1973, 1974, two closely related concepts have been discussed under the term presupposition, that is, semantic and pragmatic presuppositions. On the type of theory that I am discussing here, ϕ-features are directly associated with a semantic presupposition, but this in turn gives rise to a pragmatic presupposition.

6Stokke (2010) argues for an intention-sensitive semantics in Essay 2, “Indexicals and Intentions.” Despite this, he does not specifically examine the consequences of this theory for examples (4) and (5), which he mentions in Essay 4, “Descriptive Meaning, Presupposition and Interpretation.” Instead, he concludes that such examples demonstrate that wrong person information never yields what we could recognize as presupposition failures (and more generally that ϕ-features do not behave like standard presupposition triggers).

7 For simplicity, I refer to all these cases as representations even if the relation between the individual and the object (or individual) involving that individual is not uniform. For instance, pictures and statues visually depict an individual, while audio recordings reproduce sounds previously performed by the individual. These differences are interesting in themselves, but are irrelevant to the argument.

8 But we do not have to, as noted by an anonymous reviewer: in such scenarios, the pronouns can also refer to the representation itself. This is the case if I say “I am fuzzy” while pointing at my fuzzy image on TV (cf. Nunberg’s (1993: app.) examples in which I seems to be able to refer to a car (“I am parked out back”) or to any other object associated with the speaker). The same holds for names, pronouns, and reflexives; see Jackendoff ’s (1992) statue rule and other reference shifters.

9 The audience does not need to understand that André takes Mary to be Napoleon. The judgment of truth associated with presupposition failure arises because there is a mismatch between the gender of the pronoun and the gender of the intended referent, which is clearly Napoleon here (given the gesture of pointing at the TV screen where he is represented). Whether the audience thinks this mismatch is due only to mistaken gender (André thinks that Napoleon is a woman) or also to mistaken identity (André thinks that Mary is Napoleon) does not matter. But the introduction of mistaken identity in this example is methodologically important: it makes examples involving gender mismatch similar to examples involving person mismatch.

10 The argument is reversed in the sense that the examples discussed above show that feelings of squeamishness arise exactly as predicted by the presuppositional theory. It is not clear how other theories could make the same predictions.

11 These examples show that the intended referent of a pronoun does not have to be present in the context. In this scenario, a representation of the referent (i.e., the recording of her voice) is nevertheless present. Even this is not necessary. Suppose that Sue had a friend over last night. Sue’s neighbor infers that that friend is a girl because most of Sue’s friends are girls. In fact, Sue had Hans over, a German friend of hers. Moreover, Sue’s neighbor saw the German license plate of the friend’s car. The next morning, when Sue tells her neighbor about the friend she had over, the neighbor replies, “Yes, I saw the car, she is German.” Here, she successfully refers to Sue’s male friend even if he is not present in the context, which makes the sentence true despite the presupposition failure. The same holds for the first person pronoun in cases involving representations: the representation of the intended referent (whom the speaker believes to be himself ) does not have to be present in the context. Suppose that there was a documentary about Napoleon on TV last night. In the morning, André, who has gone mad and thinks he is Napoleon, meets his neighbor and says, “Did you watch the documentary last night? I was impressive, right?” Again, as long as the neighbor knows about André’s madness, I successfully refers to Napoleon (even if he and his representations are clearly not present in the context), which gives rise to a presupposition failure.

12 In (14b), we can only imagine that Tom would point at the recording device if he had to point, since the individual he intends to refer to (Carl) is not himself present in the context. Note that if Carl were present in the context, two possible scenarios would arise depending on whether Tom clearly indicates (e.g., by looking or pointing at the recording device) that he intends to talk about the person whose voice is heard on the recording, or whether he overtly signals (by looking or pointing at Carl) that he intends to talk about Carl, who is next to him. From Tom’s perspective, the two options yield the same meaning, but they are crucially different from the perspective of the audience, who may know that the voice on the recording is not Carl’s.

Acknowledgments

I am grateful to Irene Heim and Dominique Sportiche for helpful discussion and feedback on this work. Thanks to three anonymous reviewers for detailed comments on the manuscript. This material is based upon work supported by the National Science Foundation under grants 1424054 and 1424336.

References

Barwise,
Jon
, and
John
Perry
.
1983
.
Situations and attitudes
.
Cambridge, MA
:
MIT Press
.
Cooper,
Robin
.
1983
.
Quantification and semantic theory
.
Dordrecht
:
Reidel
.
von Fintel,
Kai
.
2004
. Would you believe it? The king of France is back! (Presuppositions and truth-value intuitions). In
Descriptions and beyond
, ed. by
Marga
Reimer
and
Anne
Bezuidenhout
,
315
341
.
Oxford
:
Oxford University Press
.
Heim,
Irene
.
2008
. Features on bound pronouns. In
Phi theory: Phi-features across modules and interfaces
, ed. by
Daniel
Harbour
,
David
Adger
, and
Susana
Béjar
,
35
56
.
Oxford
:
Oxford University Press
.
Heim,
Irene
, and
Angelika
Kratzer
.
1998
.
Semantics in generative grammar
.
Oxford
:
Blackwell
.
Jackendoff,
Ray
.
1992
.
Mme. Tussaud meets the binding theory
.
Natural Language and Linguistic Theory
10
:
1
31
.
Kaplan,
David
.
1977
. Demonstratives: An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals.
Ms.
[Published in
Themes from Kaplan
, ed. by
Joseph
Almog
,
John
Perry
, and
Howard
Wettstein
,
481
563
.
Oxford
:
Oxford University Press
,
1989
.]
Neale,
Stephen
.
2005
. Pragmatism and binding. In
Semantics versus pragmatics
, ed. by
Zoltán G.
Szabó
,
165
285
.
Oxford
:
Clarendon Press
.
Nunberg,
Geoffrey
.
1993
.
Indexicality and deixis
.
Linguistics and Philosophy
16
:
1
43
.
Percus,
Orin
.
2011
.
Gender features and interpretation: A case study
.
Morphology
21
:
167
196
.
Sauerland,
Uli
.
2003
. A new semantics for number. In
Proceedings of SALT 13
, ed. by
Robert B.
Young
and
Yuping
Zhou
,
258
275
. https://journals.linguisticsociety.org/proceedings/index.php/SALT/issue/view/96.
Sauerland,
Uli
.
2008
. On the semantic markedness of phi-features. In
Phi theory: Phi-features across modules and interfaces
, ed. by
Daniel
Harbour
,
David
Adger
, and
Susana
Béjar
,
57
82
.
Oxford
:
Oxford University Press
.
Schlenker,
Philippe
.
1999
.
Propositional attitudes and indexicality
.
Doctoral dissertation, MIT, Cambridge, MA
.
Schlenker,
Philippe
.
2003
.
A plea for monsters
.
Linguistics and Philosophy
26
:
29
120
.
Stalnaker,
Robert
.
1973
.
Presuppositions
.
Journal of Philosophical Logic
2
:
447
457
.
Stalnaker,
Robert
.
1974
. Pragmatic presuppositions. In
Semantics and philosophy
, ed. by
Milton
Munitz
and
Peter
Unger
,
197
213
.
New York
:
New York University Press
.
Stokke,
Andreas
.
2010
.
Indexicality and presupposition: Explorations beyond truth-conditional information
.
Doctoral dissertation, University of St Andrews
.
Sudo,
Yasutada
.
2012
.
On the semantics of phi features on pronouns
.
Doctoral dissertation, MIT, Cambridge, MA
.