In her ingenious and engaging book, Context and Coherence: The Logic and Grammar of Prominence (2021),Footnote 1 and in several related articles,Footnote 2 Una Stojnić appeals to “mechanisms of discourse coherence” (p. 5) to argue for a radical view of the meaning of context-sensitive language. This meaning is not, as the “dominant tradition” holds, “partially determined by non-linguistic features of utterance situation” but rather “is determined entirely by grammar—by rules of language that have largely been missed” (p. 7). I shall look critically at Stojnić’s novel view from perspectives in the theory of reference that she does not mention.

Demonstratives and pronouns (often briefly, “demonstratives”) like ‘that’ and ‘she’ are paradigms of context-sensitivity. Stojnić devotes a lot of attention to them, and so shall I. I shall argue that her radical view of context-sensitivity is wrong. In the course of doing this I shall criticize her theories of reference for deictic demonstratives and demonstrations, partly on the ground that they are seriously incomplete: her theories do not meet the demands on a theory of reference. I side with one part of the tradition on demonstratives. I shall conclude by arguing briefly that coherence has no place in the theory of meaning.

1 Stojnić on demonstratives with demonstrations

“What determines the referent of expressions like ‘that’ on an occasion of use…?” (p. 1). Stojnić identifies two traditional answers, both of which emphasize the “underde-termination” of meaning. According to the “Intentionalist” view, “it is the speaker’s referential or communicative intentions that play this role” (p. 3). This is, as she says, the “most common answer”. Indeed, Mario Gómez-Torrente thinks that this answer is accepted by “the vast majority of theorists of demonstratives” (2019, p. 27).Footnote 3 “Objectivist” views, Stojnić claims, disagree with Intentionalists, but still insist

on the crucial role of extra-linguistic factors in the resolution of context-sensitivity. For instance,…the non-linguistic features of context—objective, speaker-independent, aspects of the real-world situation in which an utterance takes place—are what ultimately fixes the meaning of context-sensitive expressions. (p. 4)

Intentionalist and Objectivist views of the reference fixing of a deictic demonstrative are indeed traditional. I reject both and so will not be defending them from Stojnić.Footnote 4 But Stojnić has strangely overlooked another sort of traditional view, a “perception-based” view which is, as we shall see, very relevant to an assessment of her own view.

Stojnić argues for a view that differs dramatically from traditional views:

context-sensitivity resolution is a matter of linguistic, rather than extra-linguistic, mechanisms; it is governed by linguistic rules, which determine particular values of contextual parameters on which context-sensitive items depend for their meaning at any given point in discourse, independently of speaker intentions, and non-linguistic features of utterance situation….the resolution of context-sensitivity is entirely a matter of linguistic convention. It is the linguistic rules that determine what ‘that’ or ‘she’ picks out on an occasion of use. (p. 5)

Why have these rules “gone unnoticed”? “[B]ecause their principal domains are entire discourses, not just their constituent words and sentences” (p. 5).

What semantic role does Stojnić have in mind for discourses? She thinks that there are “discourse conventions” and that these

determine the resolution of context-sensitivity. The important set of mechanisms I will draw on are the so called mechanisms of discourse coherence, which provide an implicit organization to the discourse that signals how individual utterances are organized into a coherent whole… (p. 5)

Utterances have not only a truth-conditional but also a “dynamic” meaning which

encodes the effect of discourse conventions, specifying how they manipulate and change parameters of context, in a way that determines the propositional content expressed by the subsequent discourse. (p. 8)

As a result of this, “a context-sensitive item receives its interpretation fully and automatically from the linguistic context” (p. 11). The referent selected for the demon-strative is “the most prominent entity—one at the center of attention—in the current context” (p. 32)

Consider one of Stojnić’s key examples:

(17) A woman came in. She sat down.

As Stojnić notes, “the traditional reaction” to examples like this was to claim that “demonstrative pronouns are ambiguous” (p. 33). ‘She’ might be anaphoric, depend-ing for its interpretation on the indefinite ‘A woman’ or it might be deictic, referring to a particular female identified in the extra-linguistic context. And in, “If a woman comes in, she sits down”, ‘she’ might be bound by ‘a woman’. Furthermore, on the traditional view, not only is the pronoun type ambiguous but any token of it may be linguistically “incomplete” (p. 38). For, consider deictic tokens: their interpretation demands, as many say, “saturation” in context. On this view, Stojnić emphasizes, “the mechanisms whereby a particular interpretation is determined…are left implicit in the logical form” (p. 38); thus, the logical form of (17), on a deictic reading, does not explicitly identify the female that ‘she’ refers to.

Stojnić rejects this traditional story unequivocally:

The meaning of a pronoun is simple, uniform, and unambiguous; as a first pass, a pronoun selects the most prominent candidate interpretation—what is “at the center of attention” at the point in discourse in which it occurs….the resolution of a pronoun requires no extra-linguistic supplementation. It is linguistically determined, through and through: by its standing linguistic meaning, and the linguistically set up context. (p. 40)

On Stojnić’s account, the pronoun ‘she’ in (17) “is naturally interpreted as bound by and co-varying with ‘A woman’” unless ‘she’ “is uttered in tandem with a pointing gesture” (p.42). Stojnić claims:

It is obvious that in this case the pointing gesture affects pronoun resolution. One intuitive way to describe this effect is to say that the pointing gesture induces another shift in attention; it makes a new entity prominent, demoting what was previously at the center of attention. (p.44)

How does the pointing gesture, a demonstration, do this?

we should represent the contribution of a demonstrative gesture, a pointing, explicitly, and interpret it independently of the interpretation of a pronoun, because there are good reasons to understand its contribution as underwritten by conventions that specify a particular semantic contribution….it is…a part of utterance, an expression along others, with its own conventionally specified contribution, that of an attention-shifting update. (p.46)

Stojnić sees her break with tradition like this. Suppose that the reference of a deictic demonstrative token is to X. According to tradition, this reference is “not underwritten by grammar” (p. 51): it is not explicit in the logical form; perhaps the representation in the logical form is just a ‘that’. In contrast, on Stojnić’s view, a demonstrative’s reference is explicit in the logical form (Stojnić et al., 2017, p. 142). First, whether it is deictic or not, its reference is always to “the most prominent candidate” at its stage of the discourse. Second, the demonstration that accompanies this particular deictic token is itself in the logical form and explicitly represents the relation to X, thus making X the most prominent candidate.

A minor puzzle. Stojnić requires that the referent for a demonstrative be made antecedently prominent by the discourse: “demonstrative pronouns require an antecedently set context, one in which a particular interpretation has been made prominent” (p. 173); “the context has to be antecedently set up correctly” (Stojnić et al., 2013, p. 509). Yet surely if a demonstrative has a pointing gesture then that gesture is typically simultaneous with the demonstrative rather than prior to it; the demonstrative and the gesture are, as she says, “in tandem” (p. 42). So, the referent is not made antecedently prominent.

Stojnić makes the reference of a deictic demonstrative dependent on a demonstration (though not always; see Sect. 6). That prompts the objection that she has simply moved the non-linguistic determination of reference from its traditional place with demonstratives to a place with pointing gestures. So “context-sensitivity resolution” is not always “a matter of linguistic, rather than extra-linguistic, mechanisms”. Stojnić anticipates this objection and responds. I shall assess her response in Sect. 3. But first we need to consider the reference fixing of demonstrations.

Stojnić’s discussion of demonstratives began appropriately with the question: “What determines the referent of expressions like ‘that’ on an occasion of use…?” (p. 1). This is appropriate because such a question is the basic one about reference for any term. We have seen that Stojnić’s answer makes demonstratives dependent on other terms in a way that “can be boiled down to just this: no heavy lifting for pronouns” (Stojnić et al., 2020, p. 143). In the case of a deictic demonstrative, its referent is (often) dependent on that of a demonstration, which Stojnić regards, rightly in my view (2004, p. 291), as a referring device in its own right. So, what determines the referent of the demonstration (what “makes a new entity prominent”)? What is Stojnić’s answer to the basic question for demonstrations?

Before considering Stojnić’s answer, it is helpful to describe the sorts of answers that might be given to the basic question for any term. Description theories have been traditionally popular: the reference of the term is fixed by certain descriptions that speakers associate with the term. But description theories cannot be true for all terms because they “pass the referential buck” to other terms (Devitt & Sterelny, 1999, p. 60), ultimately to terms that depend for reference on some sort of “direct”, non-descriptive, relation to reality.Footnote 5 The basic question for those terms demands what we might call an “ultimate” explanation of reference (Devitt, 2014).Footnote 6

Turn now to Stojnić on the basic question for demonstrations. Since demonstrations are doing the “heavy lifting” for many demonstratives, we badly need her answer. Yet, surprisingly, she has little to say. The question is not raised in her book, so far as I can see, despite the question having been pressed by Ethan Nowak and Eliot Michaelson (2020, p. 125) in a criticism of two of her earlier papers, coauthored with Matthew Stone and Ernie Lepore (Stojnić et al., 2013, 2017). In response to that criticism, the coauthors, hereafter “SSL”, do address the question, but the treatment is cursory (Stojnić et al., 2020). SSL discuss proper names, including a mention of Kripke’s (1980) “influential direct reference account of names” (p.144),Footnote 7 and then say this about the reference determination of demon-strations:

We see pointing, like naming, as a form of direct reference. We regard it as a psychological fact that pointing is directed at entities in the world. Just as with names, we conceive of this directedness in terms of causal and historical explanation, not in terms of referential intentions as typically conceived. (p. 145)

This talk of a “causal and historical explanation” and “directedness” implies that SSL have an ultimate explanation in mind as an answer to the basic question for demonstrations. Indeed, demonstrations seem very good candidates for an ultimate theory. And the “causal and historical” talk is a promising beginning to such a theory. But it is no more than that: we need the details; causal-historical links are everywhere. Which one determines the referent of the pointing? SSL do not say. So, we are left with a very incomplete explanation of the reference of a demonstration (of how it makes an entity prominent). So, insofar as deictic demonstratives really depend on demonstrations for reference, we are short of a full explanation of their reference too. This charge of incompleteness is one of my main criticisms of Stojnić’s views.

The incompleteness of Stojnić’s “causal and historical explanation” of the reference of demonstrations is a sign that she does not appreciate the demands on a theory of reference. There are more signs to come.

I noted earlier that Stojnić had overlooked one sort of traditional view of demonstratives, a perception-based view. I shall introduce this view in the next section. It provides a way to complete Stojnić’s explanation of demonstrations.

2 Causal-perceptual theories of demonstratives and demonstrations

According to traditional perception-based views, a deictic demonstrative’s reference is fixed, at least partly, by the direct perception of an object. As I pointed out recently (2022),Footnote 8 this sort of view goes back at least to Husserl (1900/01)Footnote 9 and has been urged and criticized by many. I cited some examples: McGinn (1981), Davies (1981), Evans (1982), Pendlebury (1984), Neale (1990), Almog (2012), Hawthorne and Manley (2012), and referred to papers in a volume, Singular Thought, edited by Jeshion (2010). I have urged such a view myself (1974, 1981a, b, 1985, 2004, 2014), which I sum up:

the reference of a person’s deictic referential demonstrative is fixed in the object in mind by a causal link between the person and the object when it is, or was, the focus of that person’s perception. This is what I call a “grounding”. (2022, p. 997)Footnote 10

This view of deictic demonstratives rests on three other views. (I), on the view that the singularity of a singular thought is explained by its being causally-perceptually grounded in the object of thought. (II), on the view that a demonstrative is a conventional way of expressing the singularity of a singular thought, where expressing is another causal process.Footnote 11 (III), given the deep Gricean truth about the priority of thought that underlies the wise saying, “language expresses thought”,Footnote 12 the grounding that fixes the reference of the thought is crucial in fixing the reference of the deictic demonstrative.Footnote 13

Turn now to demonstrations. It is a commonplace to say that demonstrations often accompany demonstratives. But it is important to note that demonstrations often stand alone. Thus, consider a woman confronted by a line-up and asked the question, “Who mugged you?” She points wordlessly to a man, Harry. She has referred to Harry, and not simply speaker-referred to him. Her demonstration in these circumstances is a conventional expression of the thought that Harry mugged her, just as much as would be her responding, “Harry”, when asked that same question at the scene of the crime. There is a convention of using a gesture toward an object in mind as part of an expression of a singular thought about that object. Demonstrations, like demonstratives, are conventional devices for referring to a particular object in mind. Inspired by the above theory of demonstratives, I proposed that the reference of a demonstration is determined in

the same causal-perceptual way as the reference of a demonstrative….A person’s use of ‘he’ refers to a male that is, or was, the focus of her perception. Similarly, her use of a demonstration refers to an object in the gestured area that is, or was, the focus of her perception. (2022, p. 1002; see also 2021b, pp. 285–286)

David Kaplan once remarked: “We might think of the demonstration on the model of a term in apposition to the demonstrative” (1989, p. 582, n. 35). That should be our model for demonstrations accompanying demonstratives. If the woman points to Harry in the line-up and says “He mugged me”, she has referred twice to Harry just as she would have had she said at the crime scene, “Harry, the man living above the restaurant, mugged me.” The logical form of each utterance contains two terms that, in context, refer to Harry.

So, this causal-perceptual theory is my suggested completion of Stojnić’s “causal and historical explanation” of the reference of a pointing gesture.Footnote 14 She may reject the suggestion, of course. but she needs something like it to explain the direct relation to reality that constitutes the reference of a demonstration and the singularity of the thought that the demonstration expresses. That’s what the theory of reference demands.

Any such completion raises an obvious worry for Stojnić. If a completed, perception-based, theory of demonstrations is good, why wouldn’t a similar perception-based theory of deictic demonstratives be good? (Note that my suggested completion for demonstrations was inspired by a similar one for deictic demonstratives.) The worry for Stojnić is that this is not the theory of demonstratives that she proposes; in particular, this theory, unlike hers, does not make the reference of any demonstratives dependent on demonstrations. This worry is very real, as we shall see in Sect. 6.

3 The context-sensitivity of demonstrations

Stojnić’s earlier-quoted radical rejection of the traditional view of context-sensitive language is not restricted to demonstratives but is quite general: “context-sensitivity resolution is a matter of linguistic, rather than extra-linguistic, mechanisms…the resolution of context-sensitivity is entirely a matter of linguistic convention” (p. 5). Yet, in making the reference of a demonstrative dependent on a pointing gesture, Stojnić seems to have simply moved the non-linguistic determination from one place, demonstratives, to another place, pointing gestures. As noted in Sect. 1, Stojnić anticipates this objection:

And didn’t we just then replace an intentionalist account of pronoun resolution with an intentionalist account of demonstrative gestures? That is, in interpreting a gesture, isn’t it ultimately the speaker’s intention that fixes its demonstratum…? (p. 53)

Stojnić rightly points out that even if she concedes that the grammar thus underspecifies the demonstratum of a pointing gesture, she can nonetheless maintain that “the contribution of demonstrations should still be represented separately in the logical form” (p. 53). But she does not concede (p. 54).

Stojnić’s reason for not conceding is ingenious but, I shall argue, misguided. She starts with the claim that proper names are “ambiguous” (p. 54). Many different objects get to be dubbed by a name yielding many conventions linking it to many referents; as SSL point out, “each such convention gives rise to a distinct entry in the mental lexicon of a speaker acquainted with it” (Stojnić et al., 2020, p. 144). I think that this is dead right (1976).Footnote 15 But then Stojnić claims that demonstrations “are akin to names” in being ambiguous (p. 54). SSL explain:

just as saying what name is used says whose name is used, saying what the pointing was involves saying what was pointed at. That makes pointing no more context-sensitive, and no less ambiguous than a name…we propose a semantic representation where the pointing gesture contributes its referent to logical form. (Stojnić et al., 2020, p. 145)

I find this claim about pointings frankly baffling.

To see why, we should say more about ambiguity. An ambiguous term features in more than one convention, each one yielding a distinct entry in a competent speaker’s mental lexicon. When a speaker uses the term in an utterance, its meaning on that occasion obviously depends on which entry she is exploiting. The term, with the conventional meaning that it has in that entry, will then appear in the logical form of the utterance. So, if, according to that meaning, the term refers to X, then that reference to X will be part of the logical form. It follows that there is some sort of context-dependency in the meaning of an ambiguous term on an occasion of use: its meaning varies from context to context depending on which of its lexical entries a speaker is exploiting. But, Stojnić insists (p. 55), and I very much agree (2021b, pp. 204–205), this dependency is quite different from the context-sensitivity that she rejects. For, she rejects that there is a term in the logical form, say ‘that’, which requires extra-linguistic supplementation to deliver a referent. Consider the proper name, ‘Aristotle’. There is a convention of speakers using ‘Aristotle’ to refer to a certain famous ancient philosopher, and another convention of speakers using ‘Aristotle’ to refer to a certain shipping magnate who was the second husband of Jacqui O. And there are many other such conventions for ‘Aristotle’. So, ‘Aristotle’ is ambiguous. Each one of its conventions determines a meaning and yields a distinct lexical entry. Suppose someone utters, “Aristotle was fond of dogs”, exploiting the lexical entry that refers to the great philosopher. Then the logical form of her utterance will include a token of ‘Aristotle’ that refers to that philosopher. There will be no need to look outside the language for the determination of reference.

So, if demonstrations were “akin to names” in being ambiguous, then Stojnić would be right not to concede to the objection. But it is prima facie extraordinary to say that demonstrations are thus akin to names. We quoted Stojnić’s claim (p. 46) that a demonstration, independently of a demonstrative, is governed by convention as a result of which the demonstration makes a semantic contribution; “A pointing gesture is a bona fide linguistic expression, not an extra-linguistic parameter of the context” (Stojnić et al., 2020, p. 140). Clearly, I agree (Sect. 2). But what conventions govern demonstrations? Surely, there is only one for the demonstrations that concern us.Footnote 16 I have just made a proposal about that convention in the process of completing Stojnić’s explanation of a demonstration’s reference (Sect. 2). I proposed, in effect, that the convention is of people using a demonstration to refer to an object in the gestured area, an object that is, or was, the focus of the person’s perception; it is a convention of using a demonstration to express a thought about an object in which the thought is causally-perceptually “grounded”. Now there is of course room for argument about whether this is the right theory of the direct link between a demonstration and its referent, required for an ultimate explanation. So, let’s abstract from my theory and describe the convention really vaguely: it is the convention of people using a demonstration to refer to an object in the gestured area, an object that is directly linked to the demonstration somehow or other in the context. This vague description suffices for our key point. Whereas there are many conventions for ‘Aristotle’, each of which alone links the name to a particular object, there is just one convention for demonstrations, which does not alone link the demonstration to a particular object. Rather, that convention is of a way of linking the demonstration to an object in the context; the conventional meaning of a demonstration demands saturation by whatever object in the context is linked in that way.Footnote 17 And that is the meaning of the demonstration expressed in the logical form of the utterance. So, the logical form alone does not identify a particular referent; it directs us outside the language to find the referent.

Why does Stojnić think otherwise? The answer is to be found in her discussion of

(17) A woman came in. She sat down.

I earlier noted Stojnić’s claim that ‘she’ “is naturally interpreted as bound by and co-varying with ‘A woman’” unless ‘she’ “is uttered in tandem with a pointing gesture” (p. 42). Stojnić offers the following crucial representation of this utterance when it is accompanied by the gesture:

(23) A woman came in. She [pointing at a female cat, Betty] sat down. (p. 42)

She later comments:

Notice that the deictic gesture in (23) is represented as an act of pointing at a particular demonstratum, Betty, not an act of pointing at x for some contextually determined x. That is, pointing is not semantically interpreted as having a context-sensitive meaning, which given a context (and together with potentially extra-linguistic resources that context makes available), determines a referent. Rather, pointing gestures are ambiguous between multiple possible forms (p. 54; see also Stojnić et al., 2017, p. 529).

Now (23) does represent the pointing as being “at a particular demonstratum, Betty”, but (23) is a misrepresentation of an utterance of (17) accompanied by a pointing gesture! A more accurate representation would be

(23)* A woman came in. She [pointing] sat down.

(23)* better captures the logical form of what is actually uttered. (23)’s introduction of ‘Betty’ into the logical form of the utterance to specify a particular demonstratum is quite gratuitous. For, there is no convention of using the pointing gesture to refer to Betty, or to any other female, in particular. In contrast, there is a convention of using the proper name ‘Betty’ to refer to Betty in particular. It is false that “saying what the pointing was involves saying what was pointed at” (Stojnić et al., 2020, p. 145) if we are simply describing the logical form of (17). Saying what that pointing was involves giving a physical description of this gesture, nothing more. The reference of a demonstration, but not that of a name, depends on an aspect of the current context.

In an earlier work, Stojnić and her coauthors, SSL, “introduce…‘<πc>’, where ‘π’ corresponds to the act of pointing and ‘c’ to some individual being pointed at” (Stojnić et al., 2017, p. 526). SSL then place the symbol ‘<πc>’ in the logical form of utterances that include a pointing. No argument is given for the inclusion of ‘c’. I am arguing that there is no basis for it.

What I have just claimed about pointing conventions seems obvious to me. That’s why I’m baffled by SSL’s claim. Here is some more about conventions in support of my position:

On occasions like baptisms, linguistic conventions are established by people in a community stipulating that a certain form has a certain meaning and the community concurring. However, following Grice (1989) and Stephen Schiffer (1972), I think that the conventions associated with a linguistic form—a sound, an inscription, etc.—in a community typically come from the regular use in the community of that form to convey certain parts of messages. That regular use of it in utterances with a certain speaker meaning, leads, somehow or other, to that form having that meaning conventionally in the language of that community. (2021b, p. 36)Footnote 18

People regularly use a demonstration to refer to some entity or other in the gestured area. They do not regularly use a demonstration to refer to the cat Betty in particular; indeed, they may never have done so until this moment and may never do so again. Furthermore, there has been no stipulation that a demonstration is to refer to Betty in particular. Contrast this with the story for the name, ‘Betty’. ‘Betty’ was likely stipulated to be the name of the cat in question and has likely been used countless times to refer to her.

I conclude that Stojnić should concede: in making the reference of a demonstrative dependent on a pointing gesture she has indeed simply moved the non-linguistic determination of reference from its traditional place with demonstratives to a place with pointing gestures. Even if she did concede this, of course, her account of that dependence might still be right. In Sect. 6, I shall argue that it is not right. But first I have a puzzle about ‘I’.

4 A Puzzle about ‘I’

Stojnić’s rejection of traditional views of extra-linguistic “context-sensitivity resolution” is quite general. So how does it deal with the “pure indexical” ‘I’?

Stojnić thinks that demonstratives, on her view of them described in Sect. 1, are like ‘I’:

there are moves in discourse that antecedently, but systematically, set up a context in a way that allows a demonstrative pronoun to automatically select a referent, much as ‘I’ does; (pp. 31–32)

We have seen how this automatic selection works for demonstratives on her view of them: they select “the most prominent entity” determined by the discourse; for example, in (17), the entity is determined by the prior ‘A woman’ or by an accompanying pointing gesture. And this, on her view, is all linguistic selection. But how is that so with ‘I’? As she says, “the meaning of ‘I’ fully determines its referent as a function of context” (p. 2). But that meaning automatically selects a referent only given the context, in particular, given the speaker. What appears in the logical form is just ‘I’, having a conventional meaning that yields a referent only after the identification of a speaker. But then that identification is surely extra-linguistic.

So, reference resolution for ‘I’ seems to be unlike that for demonstratives on Stojnić’s view of them, and to be partly extra-linguistic. And ‘I’ is also surely quite unlike proper names in this respect. Yet she thinks not:

And while it is true that it is a non-linguistic fact that I am speaking when I utter ‘I,’ this doesn’t mean that non-linguistic features of utterance situation fix the meaning of ‘I’ on an occasion of use: as with ‘I,’ one has to look into the world to determine who Mary is when the speaker utters ‘Mary is smart’; that Mary is who she is is a non-linguistic fact, but not the fact that determines that ‘Mary’ refers to Mary. (p. 2, n. 1)

Something has gone wrong here. It is true that non-linguistic features of the utterance situation do not partly fix the meaning of the speaker’s use of ‘Mary’. So, we do not have to “look into the world” to determine that Mary has the property of being the referent of ‘Mary’. That property of being the referent is the only possible part of what constitutes “who Mary is” that is relevant to reference fixing. In contrast, non-linguistic features do partly fix the meaning of Stojnić’s use of ‘I’. So, we do have to look into the world to determine that Stojnić has the property of being the referent of that use of ‘I’. Of course, to determine the many other, much more interesting, properties that constitute who Mary is we would have to look into the world. And so we would too to determine who Stojnić is. Mary and Stojnić are indeed on a par in this respect. But these other properties are irrelevant to our semantic concerns. When it comes to Mary’s property of being the referent of ‘Mary’, and Stojnić’s property of being the referent of ‘I’, Mary and Stojnić are not on a par. We have to look to the world to determine Stojnić’s property but not Mary’s.

I conclude that Stojnić’s radical thesis that “context-sensitivity resolution is a matter of linguistic, rather than extra-linguistic, mechanisms” is not true of ‘I’. In Sect. 3 we saw that the thesis is not true of demonstrations. In Sect. 6, I shall argue that the thesis is not true of deictic demonstratives.

I turn now to the theory of demonstratives.

5 Deictic demonstratives without demonstrations

I shall be concerned only with deictic demonstratives, particularly with discourse-initial ones.Footnote 19 An obvious problem for the view that the reference of a deictic demonstrative is dependent on a “pointing gesture” is that there seem to be examples of demonstratives that are not accompanied by such a gesture. Stojnić is aware of this, of course, and has a response. But before considering the response, I shall demonstrate just how common these demonstratives are.

I start with what should be the most uncontroversial examples. These are demonstratives “referring to an object earlier observed and now remembered”, as Gareth Evans puts it (1982, p. 135). Evans calls these “past-tense demonstratives” but they have since been more aptly called “memory-demonstratives”. The referent of a memory-demonstrative is typically not around to be demonstrated. Thus, consider two friends, Fiona and Anne. They were at the same party and run into each other at breakfast next morning. Fiona says to Anne, “That was a lot of fun last night”, referring to the party. Or, commenting on one notable part of the evening, “That was a moving speech”, referring to the address that started the evening. Or, commenting on a man who got very drunk and behaved badly, “He was really obnoxious”, referring to the drunk. In each case, Fiona had a singular thought and expressed it in a conventional way using a demonstrative without a gesture.

Next, consider examples where the object of thought is around to be demonstrated but the speaker rightly thinks she can convey her message about that object without a gesture. Here are some examples. During the speech at that same party, Fiona says to Anne. “This is moving”; or later, observing the drunk, “He is really obnoxious”. A person sees a stamped letter on a table and says to her companion, “That’s from Spain”. Life is full of examples of such discourse-initial demonstratives being used without demonstrations.

And there are examples of deictic demonstratives that are not discourse-initial and are not accompanied by a demonstration. Consider this again:

(17) A woman came in. She sat down.

According to Stojnić, if ‘she’ is not accompanied by a demonstration, it “is naturally interpreted as bound by and co-varying with ‘A woman’” (p. 42);

even if Mary is jumping up and down in front of the interlocutors, attracting their psychological attention, and being clearly the most salient entity in the speech situation, unless an overt signal establishes her as the referent of ‘she,’ the anaphoric reading will still be the one recovered. (p. 49)

The thought is that (17), uttered without a focal stress on the pronoun, and out of the blue, receives naturally an anaphoric interpretation. (p. 49, n. 22)

This is surely not right. What is “natural” depends on the circumstances, and it is easy to come up with ones where (17) is uttered, out of the blue without overt signal or stress, and yet where Mary is not just the natural but the right interpretation of ‘she’. Suppose that Tom and Harry are seated together, observing Mary’s disruptive behavior with distaste, while waiting for a meeting to start. Harry bends down to his briefcase to find his notes. When he finds them, he looks up, sees no Mary, and so asks, “What happened?” Tom replies, flatly, with (17). His ‘she’ is a recent-memory-demonstrative that refers to Mary.

Finally, in my view, demonstratives, like any other term, can be used meaningfully without being aimed at any audience. This is controversial because it goes against a key Gricean idea: “Meaning entails audience-directed intentions” (Schiffer, 1992, p. 515). Noam Chomsky has an apt response:

Under innumerable quite normal circumstances—research, casual conversation, and so on—language is used properly, sentences have their strict meaning, people mean what they say or write, but there is no intent to bring the audience (not assumed to exist, or assumed not to exist, in some cases) to have certain beliefs or to undertake certain actions. (1975, p. 62)

I agree, and have argued that the basic intentional act in the meaningful use of language is not one of communicating a message but one of expressing a thought (2020, pp. 13–16; 2021b, pp. 62–66). Thus, it is conventional to use demonstratives to express singular thoughts whether aimed at an audience or not. Here are some that are not. The gesture-less utterances, “This is moving”, “He is really obnoxious”, “That’s from Spain”, might each have been made with no audience in mind. Tom thinks about Mary’s performance whilst Harry is elsewhere and mutters to himself, “She is going to disrupt the meeting”. A lepidopterist, alone in the forest, observes a rare butterfly and says, “That I must have”. Briefly, a person can have a singular thought and meaningfully express it using the conventional forms of her language without any thought about an audience.

In sum, there are countless examples of deictic demonstratives without demonstrations: some referring to entities experienced in the past; some to presently experienced entities where no demonstration is needed to communicate; some where the speaker is expressing a thought with no aim to communicate it.

If this is right, the worry for Stojnić that we identified in Sect. 2 becomes more acute. Clearly, the reference of a demonstrative cannot be explained in terms of a demonstration if the demonstrative has no demonstration. So, how is it to be explained? We can look for a model in the sort of explanation that Stojnić needs anyway for demonstrations. I gave a perception-based example of this model:

The reference of a person’s deictic referential demonstrative is fixed in the object in mind by a causal link between the person and the object when it is, or was, the focus of that person’s perception. (2022, p. 997)

On what basis could Stojnić resist some such explanation of demonstratives, given what she needs to explain demonstrations?

6 Stojnić on deictic demonstratives without demonstrations

I turn now to what Stojnić has to say about deictic demonstratives without demonstrations. She makes a couple of preliminary points that seem to play down, at least, the extent of these phenomena. (I), commenting on “She is happy”, Stojnić points out that “beginning a conversation with a pronoun is normally infelicitous, or odd, unless the referent is somehow specified, typically by a pointing gesture” (p. 49). She returns to this later, claiming that felicity requires a “proper setup” (p. 173). Now, first, any utterance at all, even the most meaningful and true, may be infelicitous without a proper setup. For example, in most everyday contexts, it would be infelicitous to inform someone out of the blue, “Jupiter has many moons”. Second, as we have just seen, there are countless setups where the use of a deictic demonstrative without a demonstration is felicitous.

(II), in an important note, Stojnić remarks that “deictic gestures are diverse, and need not exploit hand pointings; they can involve head gestures, eye gaze, etc.” (p. 49, n. 24). A speaker can certainly gesture by jerking her head or eyes. But an eye gaze is not a gesture. I shall not fuss about that because there are three more important points to make about eye gazes. (i) In the many cases where a demonstrative is the immediate result of an eye gaze at an object, treating the gaze as if it were a gesture that determines the referent of the demonstrative would yield the same result as a perception-based view. (ii) A memory-demonstrative is not only not accompanied by a gesture at the referent but also not by an eye gaze at it. (iii) The explanation of singular thoughts is explanatorily prior to the explanation of demonstrative reference (Sect. 2). Whereas gestures, properly so-called, are irrele-vant to the explanation of the thought’s reference, eye gazes (and their analogues with some other sensory modes) are arguably central to that explanation and hence to the explanation of the reference of the demonstrative used to express that thought.

Despite playing down demonstratives without demonstrations, Stojnić accepts that there are some. In the continuation of the important note, she says:

The prediction of the account is therefore not that deixis without a pointing gesture is impossible discourse initially, but that it succeeds only for utterances that employ linguistic mechanisms that independently require construing a particular referent as the center of attention in the current state of the ongoing discourse. (p. 49, n. 24)

What linguistic mechanisms does Stojnić have in mind? For an answer, Stojnić directs us to chapter 12 and, more importantly, to SSL’s early paper with the promising title, “Deixis (Even Without Pointing)” (Stojnić et al., 2013, mislabeled “2018”). What we need is an answer to the basic question that opened Stojnić’s book: “What determines the referent of expressions like ‘that’ on an occasion of use?” (p. 1), particularly of deictic uses without demonstrations. I earlier (Sect. 1) found a sign in Stojnić’s discussion of demonstrations that she does not appreciate the demands on a theory of reference. The signs are vivid in her discussion of these demonstratives. For, as Nowak and Michaelson (2020, pp. 129–134) bring out in their criticism of that early paper, her answer is seriously incomplete. And the incompleteness remains in chapter 12.

An answer to the basic question for any term can be a description theory, explaining reference by certain descriptions speakers associate with the term; or the answer can be an ultimate theory, explaining reference by direct relations to reality (Sect.1). Stojnić’s talk of a “causal and historical explanation” for demonstrations seemed to be the beginning of an ultimate explanation of them. Demonstrations are indeed plausible candidates for an ultimate theory, but so too are deictic demonstratives; indeed, I have argued elsewhere that they are “the most likely candidates” (2022, p. 997). We offered perception-based ultimate theories for them both in Sect. 2. Stojnić’s (apparent) proposal of an ultimate theory of demonstrations prompts the expectation of a similar theory of deictic demonstratives. Yet, as we will now see, what she seems to offer is not that but an implausible description theory.

According to SSL, the reference of these demonstratives is determined by a “coherence relation” that is “explicitly represented in the logical form of situated utterances” (2013, p. 503). Because of that explicit representation, the grammar alone determines reference. The coherence relation that matters to our deixis issue is Summary, which is “a report of what’s visible in the situation” (p. 504). Precisely how Summary is supposed to fix reference is hard to extract from SSL’s discussion (pp. 514–520), particularly given the absence of a fully worked-out example. Nowak and Michaelson extract the following theory:

demonstrative sentences like ‘that is F’ which are not accompanied by a gesture turn out to have the form of existentially quantified sentences: they will be true with regard to every assignment that returns a pair of events such that the second is a summary of the first and which is such that the central object of the second is an F….non-gestural deictic sentences, like [(5)], are to be understood as truth conditionally equivalent to rather complicated descrip-tive sentences. (2020, p. 131)

I am not confident that Nowak and Michaelson have SSL right, but I can do no better.

So, whereas Stojnić (apparently) has an ultimate theory of deictic demonstratives with demonstrations, she (apparently) has a description theory of ones without demonstrations.

The alleged reference fixing role of SummaryFootnote 20 emerges in SSL’s discussion of an example. Looking at a video of Julia Child’s stovetop from above, someone produces the following utterance “without an overt demonstrative gesture”:

(5) That’s an omelet. (2013, p. 505)

Surely (5) could be made true only by an omelet being on that very stovetop. But not according to SSL’s description theory. Suppose that what is on the stovetop is not an omelet but a pancake. Nonetheless, as Nowak and Michaelson point out: “Somewhere in the universe there is almost certainly a pair of events such that one summarizes the other and has an omelet at its center” (2020, p. 132). So (5) would still come out true.

Nowak and Michaelson continue their criticism with an example of a vegetarian omelet and a ham omelet being prepared side-by-side. Someone present says,

(12) That’s a lovely-looking omelet.

As Nowak and Michaelson point out, the purely quantificational logical form that SSL assign to (12) lacks anything that will identify one omelet rather than the other as the referent (pp. 132–133).

Consider SSL’s example, “That’s Jupiter”, made as a comment on the view through a telescope. SSL claim that Jupiter is the referent because it is “the central entity in the situation seen through the telescope” (Stojnić et al., 2013, p. 519). Nowak and Michaelson are “skeptical that there is a well-behaved property of being what is centrally imaged in a telescope, (2020, p. 134). Earlier, discussing the Child situation, they point out that SSL “never offer a straightforward answer” to “what makes the omelet the ‘central individual’” (p. 129). Indeed! Another expression that looms large in Stojnić’s discussion of prominence, “center of attention” (e.g., p. 32), raises the same concern. In virtue of what are the omelet and Jupiter the center of attention in their respective situations? Absent answers to these questions about centrality, Stojnić’s description theory is seriously incomplete.

The criticisms I have just summarized bring out both the implausibility and incompleteness of SSL’s theory of demonstratives without demonstrations. So, it is odd that SSL do not address these criticisms in their response (Stojnić et al., 2020) to Nowak and Michaelson. And Stojnić does not even mention Nowak and Michaelson’s paper in her later discussion of these demonstratives (pp. 173–176). So, let me add to the woes of her theory.

The most devastating part of Kripke’s refutation of description theories of proper names is the “Ignorance and Error” argument (Devitt & Sterelny, 1999, pp. 54–59). Kripke points out that most users of the names of Cicero, Catiline, Feynman, and Einstein are too ignorant to give the identifying descriptions of these people required by description theories (1980, pp. 80–83). Furthermore, many users of the names of Peano, Einstein, and Columbus associate descriptions that are false of those people (pp. 84–85). This discussion inspires criticism of description theories of many terms, including of Stojnić’s theory of demonstratives without demonstrations. Since Stojnić has not supplied a worked-out example, it is hard to know precisely what description she would claim determines the reference of any particular token demonstrative. But, whatever that reference-determining description is supposed to be, it would be easy, inspired by Kripke, to come up with cases where the speaker of that demonstrative was too ignorant of the situation, or too wrong about it, to provide the required description. Thus, Fiona’s memory-demonstratives at breakfast (Sect. 5) would still refer even if her hung-over memory of the evening was incomplete and erroneous. Fiona’s, “He was really obnoxious”, would still refer to the drunk, who was actually in a dark suit, abusing a man from the floor, in a French accent, even if Fiona thought that the drunk was in a tuxedo, abusing a woman from the stage, in a Spanish accent; The descriptions that a speaker associates with a demonstrative are as irrelevant to determining its reference as are those that a speaker associates with a name are to determining its.

In sum, a description theory of demonstratives of the sort that Stojnić seems to have in mind is very implausible. This may be why nobody has proposed such a theory before, to my knowledge.Footnote 21

This concludes my criticism of Stojnić’s theory of deictic demonstratives. This criticism has three facets. (I) In Sect. 1, I argued that Stojnić’s (apparent) ultimate theory of reference for demonstrations is incomplete. This incompleteness carries over to her theory of reference for the deictic demonstratives that she takes to be dependent on demonstrations for reference. (II) In this section, I have argued that Stojnić’s (apparent) description theory of demonstratives without demonstrations is also incomplete. The theory does not provide what a theory of reference demands: an explicit statement of precisely what description is alleged to be sufficient to identify the referent. In particular, we need to be told in virtue of what a certain entity is “central”, “the center of attention”, and so on in the situation described. (III) Finally, drawing on Nowak and Michaelson and inspired by Kripke, I have argued that Stojnić’s theory of deictic demonstratives, so far as it goes, is false. I have briefly presented an alternative, perception-based, theory that explains the reference of all deictic demonstratives, whether accompanied by a demonstration or not, without any dependence on a demonstration. And it is an ultimate theory of them all, not a description theory.

My theory comes with a similar perception-based theory of demonstrations (Sect. 2). What basis could Stojnić have for resisting some such theory of demonstratives and demonstrations? Some such theory is a plausible completion of her (apparent) ultimate theory of demonstrations (Sect. 2). And, given that these demonstratives and demonstrations are both devices for expressing the singularity of singular thoughts, we should expect that the theory that explains deictic demonstratives will indeed be similar to the one that explains demonstrations. The worry first identified in Sect. 1 is a serious threat to Stojnić’s view of deictic demonstratives.

On my view, demonstratives and demonstrations are distinct conventional devices for expressing a singular thought about an object in which the thought is grounded. It is common for them to be used together for that purpose “in apposition” (Sect. 2). When they are, and all goes well, the demonstrative and the demonstration will each semantically refer, in its own right, to the one object.Footnote 22 But all may not go well, as shown by the popular examples of Kaplan’s Carnap-Agnew (1979, p. 396) and Reimer’s keys (1991, p. 190). I have discussed these examples elsewhere (2022, pp. 1001–1003; 2021b, pp. 286–287) and will say no more here beyond restating my verdict on the Kaplan case. I concluded that Kaplan’s demonstrative straightforwardly semantically referred to Carnap’s picture, but his demonstration failed to refer because that picture was not, as the convention requires, in the area to which Kaplan pointed. Still, Kaplan’s demonstration did speaker-refer to that picture because it was expressing a thought grounded in that picture. So, I might have continued, what he literally said, which must take account of the meanings of both these terms in apposition, is only “half true”. Stojnić has similar examples (p. 50; Stojnić et al. 2013, p. 508) which I would treat similarly.

My view of demonstratives is traditional in taking demonstratives to be ambiguous. They not only have the deictic reading that we have been discussing but also bound variable and anaphoric readings, and even an attributive reading (note 19). Stojnić emphatically rejects this ambiguity (p.40). She surely sees this rejection as a strength of her view. For, it is common to embrace Grice’s Modified Occam’s Razor, “senses are not to be multiplied beyond necessity” (1989, p. 47), and Ruhl’s “Monosemantic Bias”, “the hypothesis that a form…has a single meaning, and that all the complicating factors that make it appear polysemic have their sources in contextual contributions to meaning” (1989, p. 8). I have argued against the Razor and the Bias: we should hypothesize a linguistic meaning whenever doing so best explains regularities in usage (2013b, pp. 297–300; 2021b, pp. 143–169). So, we should posit several meanings for demonstratives.

Finally, consider the claim that the grammar, including Summary, alone determines the referent of a deictic demonstrative unaccompanied by a demonstration. Stojnić’s description theory is too incomplete to support this remarkable claim: we have not been told what makes a certain entity “central” in a situation. And it is hard to see how any completion could support the claim: a description that does not itself involve a demonstrative may not uniquely describe the referent.

There is no conventional linguistic rule for a demonstrative according to which it refers to Child’s omelet, the drunk at the party, or to any other entity in particular. Rather, the rule for deictic demonstratives, as for demonstrations, is a way of linking the demonstrative to an object in a context. I have aired a perception-based theory of that way. But let us follow the example of our discussion of demonstrations and describe that way vaguely and theory-neutrally: it is the way of using a demonstrative to refer to a (meaning compatible) object that is directly linked to the demonstrative somehow or other in the context. The conventional meaning of a demonstrative, like that of a demonstration, demands saturation by whatever object in the context is linked in that way. And that is the meaning of the demonstrative expressed in the logical form of the utterance. So, the logical form alone does not identify a particular referent; it directs us outside the language to find the referent.

This concludes my criticism of Stojnić’s radical view that context-sensitivity resolution is “determined entirely by grammar”. The view is wrong for deictic demonstratives as well as for demonstrations (Sect. 3) and ‘I’ (Sect. 4).

7 A word about coherence

Stojnić claims: “Successive contributions to the discourse must be linked into a coherent whole by a recognizable flow of interpretive relationships” (p. 61). It is a consequence of the approach to language argued for in Overlooking Conventions (2021b), and illustrated in the present paper, that Stojnić is wrong about this. Coherence, like salience, has its place in the theory of communication but no place in the theory of meaning. I have argued this about salience (2013b, p. 294, n. 12; 2021b, p. 134). Here, briefly, is how I would argue it about coherence.Footnote 23

The coherence that matters to communication is primarily a property of thinking. One thought coheres with its predecessor if the two are linked in some appropriately rational way. Here are some truisms. (a) Coherence comes in degrees, from highly rational thinking all the way down to mere “association of ideas”. (b) People differ in the coherence of their thinking. (c) The coherence of a person’s thinking varies from time to time; it tends to get worse after a few drinks. (d) Any thought a person may have has a meaning (content) however well or badly the thought coheres with its predecessor. Indeed, its degree of coherence with its predecessor is a function of their meanings. So, a thought’s meaning is independent of its coherence.

A person’s language typically includes a conventional way of expressing any thought that she may have, however incoherent (a way that may demand saturations in context). The resulting utterance will inherit its literal linguistic/semantic meaning from the meaning of the thought it expresses. For, that’s what it is to express a thought conventionally. (Of course, a person may express the thought unconventionally, resulting in an utterance that has a speaker meaning that differs from any literal meaning it may have.) Since the meaning of the thought is independent of its coherence, so too is the meaning of the utterance expressing that thought. So, coherence has no place in a theory of meaning.

Nonetheless, coherence is very relevant to a hearer’s process of understanding an utterance. That understanding involves using multiple clues to figure out, given the context, which meaning of an ambiguous term is likely, what saturations are likely to have occurred, what the utterance might mean non-literally, and so on. The likelihood of any interpretation being correct depends on whether it implies an appropriate degree of coherence to the speaker’s thinking in the context. So, a hearer should interpret David Lewis’ utterances so that they come out highly coherent, even after a drink or two; and a hearer should have much lower expectations of Donald Trump’s utterances. So, coherence has a place in the theory of communication.

Stojnić discusses the following discourse:

(40) John took the train from Paris to Istanbul. He likes spinach.

She claims “that the requirement that a discourse must be coherent is strikingly evident in the interpretive effort (40) elicits. Given apparently unrelated facts about John in (40), we search for a connection” (p. 62). This is right about the interpretive effort, but that effort is not evidence that discourse must be coherent. It is evidence of the role of coherence in linguistic understanding.

8 Conclusion

Una Stojnić urges the radical view that the meaning of context-sensitive language is not “partially determined by non-linguistic features of utterance situation”, as traditionally thought, but rather “is determined entirely by grammar—by rules of language that have largely been missed” (p. vii). The missed rules are ones of discourse coherence (p. 5). I have argued against this radical view as it applies to demonstrations (Sect. 3), ‘I’ (Sect. 4), and demonstratives (Sect. 6). I have made three other main criticisms.

Her discussion starts, appropriately, with the basic question for demonstratives: “What determines the referent of expressions like ‘that’ on an occasion of use…?” (p. 1). Her answer is that reference is to “the most prominent candidate” at the demon-strative’s stage of the discourse. If the demonstrative is a deictic one accompanied by a gesture, then its reference is fixed by that demonstration. This raises the basic question for demonstrations: What determines their referent? Stojnić (in company with Stone and Lepore) talks briefly of “causal and historical explanation” (Stojnić et al., 2013, p. 145) but gives no details. So, we are left with an incomplete theory of reference for demonstrations. That is one of my criticisms (Sect. 1).

This charge of incompleteness carries over, of course, to the theory of reference of any demonstrative that depends on a demonstration for reference. What about Stojnić’s theory of deictic demonstratives that are not accompanied by a demonstration? I argued that her (apparent) description theory of these is also incomplete. The theory does not provide what a theory of reference demands: an explicit statement of precisely what description is alleged to be sufficient to identify the referent. Lastly, I argued that Stojnić’s theory of deictic demonstratives, so far as it goes, is false. I presented an alternative, a version of traditional perception-based theories of reference, a tradition that Stojnić strangely overlooks. My version explains the reference of all deictic demonstratives, whether accompanied by a demonstration or not, without any dependence on a demonstration. It comes with a similar perception-based theory of demonstrations (Sect. 2). What basis could Stojnić have for resisting some such theory of demon-stratives and demonstrations? For, some such theory is a plausible completion of her theory of demonstrations. That concluded my criticism of Stojnić’s theory of demonstratives.

My final criticism was that coherence has a place in a theory of understanding but no place in a theory of meaning (Sect. 7).