Reference fiction, and omission

Abstract

In this paper, I argue that sentences that contain ‘omission’ tokens that appear to function as singular terms are meaningful while maintaining the view that omissions are nothing at all or mere absences. I take omissions to be fictional entities and claim that the way in which sentences about fictional characters are true parallels the way in which sentences about omissions are true. I develop a pragmatic account of fictional reference and argue that my fictionalist account of omissions implies a plausible account of the metaphysics of omissions.

1 Introduction

Consider the following sentence:

The omission of the name of Ezra rather militates against the supposition that Ben Sira had the Chronicler’s book before him when he wrote [the first book of Psalms].¹

Call this sentence O. Is O true? At least one condition on O’s being true is that every singular term in O refers. Thus, ‘Ezra’, ‘Ben Sira’, and the term ‘Chronicler’s book’ all must denote some entity or other in order for O to be true.

To what does the term ‘the omission’ refer? ‘The omission’ (in this context) looks like a singular term. The omission of a name, however, would seem to be nothing at all. Does ‘the omission’ in this context refer to nothing? If that is correct, then it appears that ‘the omission’ cannot refer to anything. O, then, is neither true nor false; it is meaningless because one of the basic expressions in O fails to refer.

This problem is not restricted to O alone. One can omit more than just a name. One can omit to buy milk on the way home from work or omit to weed the garden for all of July. Omissions can be mundane (e.g., omitting to set the napkins correctly) or serious (e.g., omitting to remove the kids from the hot car). For all of these omissions, there is some suite of sentences that we can use to describe the situation.² Are all of these sentences meaningless for the reasons that O is meaningless?

In this paper, I argue that sentences that contain ‘omission’ tokens that appear to function as singular terms (hereafter, O-type sentences) are meaningful while maintaining the view that omissions are nothing at all. I take omissions to be fictional entities and claim that the way in which sentences about fictional characters are true parallels the way in which sentences about omissions are true.

The following discussion divides into six sections. In Sect. 1, I outline the semantic puzzle of the meaningfulness of O-type sentences. The conclusion of this puzzle is that no O-type sentence can refer and we therefore cannot speak meaningfully about omissions. I also note several strategies for addressing the puzzle. In Sect. 2, I discuss one strategy to address the puzzle, identification accounts of omissions. Identification accounts identify omissions with some positive ontological entity. While these accounts can easily account for the semantic contribution of ‘omission’ terms to sentences like O, I argue that they are implausible or, at least, no more plausible than alternative strategies. This leaves open the possibility of exploring other strategies for addressing the semantic puzzle outlined in Sect. 1. In Sect. 3, I outline nihilistic accounts of omissions and the corresponding solution to the semantic puzzle that these accounts imply. These accounts presume that omissions are absences (and are thus opposed to identification accounts). Nihilistic accounts of omissions avoid some of the problems that plague identification accounts, and the rest of the paper explores one particular nihilistic account of omissions. In Sect. 4, I argue that the semantic puzzle about omissions parallels discussions about the nature of fictional discourse and a parallel semantic puzzle about fictional names. Thus, resources from the latter domain can be used to solve problems in the former domain. In Sect. 5, I summarize Kripke’s pragmatic theory of fictional discourse, focusing specifically on his Pretense Principle. In Sect. 6, I apply Kripke’s theory to the semantic puzzle of omissions and show that the puzzle disappears if we take omissions to be a kind of fictional entity. I also distinguish my fictionalist account of omissions from similar accounts and discuss some of the virtues of my fictionalist account.

2 The semantic puzzle

In this section, I present the semantic puzzle about the meaningfulness of O-type sentences. I call this puzzle the Failure of Failures to Refer (FFR):

• FFR-1) Omissions are nothing at all. [premise]

• FFR-2) Sentences have a truth-value only if every basic expression in the sentence successfully refers. [premise]

• FFR-3) O-type sentences are truth-evaluable only if ‘omission’ tokens within O- type sentences refer. [instance of 2]

• FFR-4) Semantic tokens, if they refer, refer to something. [premise]

• FFR-5) ‘Omission’ tokens do not refer. [1, 4]

• FFR-6) O-type sentences are not truth-evaluable. [2, 3, 5]

• FFR-7) To speak meaningfully about some topic T, some relevantly broad class of sentences within T must be truth-evaluable. [premise]

• FFR-8) To speak meaningfully about omissions, some relevantly broad class of O-type sentences must be truth-evaluable. [instance of FFR-7]

• FFR-9) One cannot speak meaningfully about omissions. [6, 7, 8]

There are several strategies available to undermine FFR, two of which I consider in this paper (I focus on these two strategies because they primarily focus on the premises that have to do with omissions and O-type sentences). One strategy is to deny FFR-1. This strategy includes views that reductively characterize omissions in terms of entities that belong to some positive ontological category. I call these views identificationist, because they attempt to identify omissions with some other ontological entity (Lewis 2004, p. 282). This, in turn, implies that FFR-1 is false. I discuss the identification strategy in Sect. 2. Another strategy is to accept that omissions are nothing at all (i.e., accept that FFR-1 is true). I call these views nihilist, because they presume that (at least some) omissions are nothing at all. This strategy implies that the move from FFR-1 and FFR-4 to FFR-5 is invalid.³ In Sects. 3–6, I explore a nihilist view of omissions (drawing on resources from discussions of fictional discourse) and the corresponding solution to FFR that the nihilist view implies. The key to this strategy is taking omissions to be fictional entities.

3 The identification strategy

As noted above, identification accounts of omissions reductively characterize omissions in terms of entities that belong to some positive ontological category (e.g., facts, events, etc.). I refer to these accounts as identificationist because they attempt to identify omissions with some entity in a positive ontological category. For example, an identificationist account might claim that omissions are identical to some higher-order states of affairs (‘Those are the only actual states of affairs’) that are related to a set of first-order actual states of affairs. Thus, the omission of Ben Sira’s name would just be the higher-order state of affairs that exhaustively picks out all actual first-order states of affairs as the only actual first-order states of affairs (Armstrong 1997).⁴

One might think that the best way to maintain the meaningfulness of O-type sentences is to pursue the identificationist strategy and deny that omissions (whether sometimes or always) are just absences (thus denying FFR-1). So, one might claim that the omission of Ezra’s name is identical to some fact, event, state of affairs, or abstract object (some unrealized or uninstantiated act-type; more on this below). The overall strategy is to say that an omission is actually something, so that ‘omission’ terms in O-type sentences denote this something.

Randy Clarke has done the best (and most recent) work in arguing against identification strategies for omissions (Clarke 2014, chs. 1 and 2). In what follows, I adopt and develop a number of his insights. The goal this section is not to refute every (or even some) identificationist position; rather, the goal is to show that no identificationist account obviously works. Either the identification account requires adopting a number of substantive higher-order commitments or fails to provide a comprehensive account of omissions. Thus, I aim to show that we have no reason to prefer identificationist accounts of omissions to nihilist accounts.

Some claim that omissions are negative facts, where a negative fact is some fact of the form ‘It is not the case that...’ (cf. Bennett 1988, pp. 218–221 and Bennett 1995, pp. 85–92). On this view, an omission is some negative fact that is made true by some bit of behavior. For instance, a proponent of the negative fact view might claim that Ben Sira’s omission of Ezra’s name just is the fact that it is not the case that Ben Sira writes Ezra’s name, which fact is made true by Ezra’s writing of other sorts of names. Jonathan Bennett summarizes the view as follows:

If ‘negative’ or any of its kin is to be seriously employed in our present context [of analyzing omissions], it must be applied to facts about behaviour rather than to acts. At any given moment, one’s behaviour is the subject of countless facts, infinitely many of them negative (1995, pp. 86–87; original emphasis).

The idea, here, is that omissions are just negative facts about behavior, not some special negative entity like a negative act.⁵

One problem is that Bennett does not offer a clear account of what a fact is. I can think of two ways to construe facts, though neither ends up being helpful for the proponent of the negative fact view (what follows is not particular to Bennett’s account; rather, it holds generally for any proponent of the negative facts view of omissions).

On the one hand, a fact might be a state of affairs that obtains.⁶ This view of facts, however, does not make clear what a negative fact would be. Either a negative fact would be a distinctively negative feature of the state of affairs that obtains (e.g., a negative event), or a negative fact would be a negative characterization of some positive aspect of the state of affairs. Both of these views have difficulty accounting for omissions (as I argue below), though there is another problem here. On this view of facts, the negative facts account of omissions reduces to other kinds of identificationist accounts (some of which I discuss below). Far from explaining the negative fact view, this account of facts just transforms the view into a different one.

A different account of facts might take facts to be propositions. Here, a negative fact would be a negative proposition that represents the world as not being some way. For instance, I am not now running, so there is some true negative proposition that represents (in some way) my not running. Omissions, then, would be these negative propositions that are true because of the various things that agents are not now doing. There are two problems with this view, however. Recently, there has been growing skepticism that propositions are entities that are intrinsically representational (cf. Speaks et al. 2014). Thus, the idea that there are negative propositions that represent the world as not being some way is controversial. One might think that the representational component of propositions is not essential to the negative facts view under discussion here. We could take facts (construed as propositions) to be structured entities constituted by objects and properties. This, however, collapses the negative facts view into differents views that take omissions to be either some composite of an object, a property, and a relation of noninstantiation or some composite of an object and a negative property (both views have difficulties, as I argue below).⁷

There are two upshots of this discussion of negative facts. The first is that the view receives no clear statement. Once we clarify core commitments of the negative facts view, we see that the view ends up collapsing into other views or that the view depends on controversial higher-order commitments. The second is that none of these considerations decisively refute the position that omissions are negative facts, though the argument suggests that the view is not obviously correct. Once we unpack the various assumptions built into the negative facts view, it becomes less clear that the view is more plausible than one that assumes that omissions are absences (i.e., assumes the denial of FFR-1). For these reasons, the negative facts view does not seem to enjoy any obvious dialectical advantage over a nihilist account of omissions.

One natural interpretation of omissions is that they are special kinds of events or some complex that contains an event as a constituent. We normally use the term ‘the omission’ in conjunction with some term for an event-type (‘the omission of getting the milk’; ‘the omission of writing Ezra’s name’; ‘the omission of weeding the garden’). It is natural, then, to think of omissions as related to events. We do not want to countenance omissions as merely negative events.⁸ After all, what would an instance of not-writing look like? The underlying problem with negative events is that it is difficult to see how to characterize the negativity of events without lapsing into some of the problems that a nihilist would face. For instance, is a negative event some positive event that lacks something? For that characterization to work, the term ‘lack’ must have some content. And there seems to be no way to reductively characterize ‘lack’ in terms of entities that enjoy positive ontological status (this, in turn, is why positing negative events appears to be just as prima facie implausible as the nihilist view, because these events appear to be partially constituted by an absence).

For this reason, some have proposed that omissions are some complex that contain at least one distinctively negative constituent. Let e name the event of writing the Psalms and n name the event of writing Ezra’s name. Barker and Jago suggest in a recent paper that we can take omissions to be e and n tied together by the relation of noninstantiation (Barker and Jago 2012). Thus, Ben Sira’s omission is the entity composed of e, n, and the relation of noninstantiation.

The problem is that this view entails that all objects are related, either by the relation of instantiation or noninstantiation, to all properties. This seems implausible. Objects have genuine ties to the properties that they possess, but they need not bear any relation to the properties they lack. Positing a relation of noninstantiation seems ad hoc. In addition, it is not clear how this avoids nihilism, as mentioned above. Is the relation of noninstantiation meant to relate objects to properties that they lack? If so, what is the content of ‘lack’? If noninstantiation is simply basic, then the ad hoc charge returns.

Graham Priest proposes a similar view, though he claims that we should think of omissions as negative states of affairs (or components of negative states of affairs). Ben Sira’s omission of writing Ezra’s name is equivalent to some state of affairs that contains the event of Ben Sira’s writing of Ezra’s name and some negative polarity, both of which are tied together by the relation of instantiation. The negative polarity accounts for the negativity of the state of affairs (e.g., Ben Sira’s omission of writing Ezra’s name) (Priest 2000, pp. 317–318; cf. Beall 2000).

The problem with Priest’s account is that there is no good explanation of what a polarity is. In particular, Priest does not explain of what positive and negative polarities are variants. Take any pair of positive and negative states of affairs; the relation of instantiation, the objects, and the properties between the two are equivalent. In what way does polarity alone mark a distinction?

Priest thinks that these polarities are analogous to the polarities of physical theories, like the spin of electrons. The problem with this analogy is that physical theories offer reductive analyses of physical polarities. We understand, for instance, that spin-up and spin-down are variants of intrinsic angular momentum (Dodd 2007, pp. 390–391 and Clarke 2014, pp. 40–41). Priest’s theory offers no corresponding reductive analysis for metaphysical polarities. Again, positing negative polarities seems ad hoc.

Sometimes, omissions seem to just be descriptions of actions in terms of what those actions are not. For any action, there is some range of positive and negative descriptions that we can use to designate the action. Omissions, on this view, are just descriptions of actions in terms of what those actions are not. Bruce Vermazen offers the example of some action that we can describe as either ‘his twisting of the buttons’ or ‘his omitting of eating too many hors d’oeuvres’ (Vermazen 1985, p. 95). The negative description does not have a negative referent; rather, it has a negative sense (Varzi 2006, p. 5). On this view, omissions are a semantic phenomenon insofar as omissions just are the negative characterizations of some action.

While this analysis might work for some omissions, it will not work for all of them. This is because it is not always the case that positive and negative descriptions of some behavior designate identical actions. Take the following example, adapted from Clarke. Suppose that Walter promises Al that he will bring some bourbon to the party at 8:00 pm. On the walk over to Al’s, Walter forgets about the bourbon, and walks right past the liquor store on his way to Al’s.⁹ Is it the case, in this scenario, that the positive description ‘Walter’s walking to Al’s’ designates the same action as the negative description ‘Walter’s forgetting the bourbon’? The two pick out things with distinct and incompatible properties. Walter might walk quickly or sluggishly to Al’s, though his forgetting the bourbon does not have a speed. Similarly, Walter’s walk occurs along a certain route, though his forgetting does not seem to occur anywhere (at the very least, his forgetting might be isolated to a particular moment that certainly does not co-locate with the walk) (Clarke 2014, pp. 26, 37–38).¹⁰ This sort of concern applies to the Ben Sira case. Ben Sira’s writing the Psalms occurs at a certain location and takes place across a certain stretch of time. The relevant negative descriptions appear to pick out something that lacks these spatial and temporal properties. For these reasons, there are two potential flaws for this kind of identification strategy. Either the strategy does not work, because no negative description of some behavior refers to the same thing as a corresponding positive description. Or, the strategy is not fully comprehensive, because there are some cases of omissions (such as Walter’s forgetting of the bourbon) that the strategy does not correctly diagnose.

Identification strategies attempt to resolve the puzzle about O-type sentences by denying the claim that omissions are nothing at all (i.e., FFR-1). In general, one recurring problem is that while particular identification strategies might work for some cases of omissions, these strategies seem to fail for other cases of omissions. In particular, no identification strategy discussed above seems adequate to the task of identifying something to which Ben Sira’s omission is identical. So even if some identification strategy is correct for some omissions, it will not extend far enough and thus fail to solve the puzzle raised above.

A related problem is that adopting an identification strategy requires adopting a number of substantive ontological commitments. These commitments do not make the various identification strategies incoherent, but they do render these strategies costly. At the very least, the arguments offered above show that no identification view obviously offers a fully comprehensive, reductive characterization of omissions. This gives us reason to consider a nihilist account of omissions.

These arguments, of course, are too quick to convince proponents of any particular identificationist strategy. The function of these arguments is to show that identificationist accounts of omissions are not obviously more plausible than nihilist accounts.¹¹ While nihilism has its own difficulties (some of which I address below) these difficulties are not any more serious than those faced by identificationists. The purpose, then, of this first section is to argue that our default position with respect to the metaphysics of omissions need not be any kind of identification account.

The difficulties of several identification strategies suggests that we have reason to accept (or at least no conclusive reason to deny) the claim that omissions are nothing at all and thereby accepting the truth of FFR-1. While this seems plausible (or at least acceptable) in light of the previous considerations, it has the unfortunate effect of leaving us with no easy route out of the puzzle about the meaningfulness of O-type sentences.

4 The nihilism strategy

A different strategy for dealing with FFR denies that omissions can be reductively characterized in terms of other entities. That is, omissions are absences or nothing at all (Shepherd 2014). Call these accounts nihilist, because they take omissions to be the absence of something. On these accounts, the omission of Ben Sira’s name is simply nothing at all. It’s an absence, or a hole in the world.

Nihilist accounts divide into two kinds, pure and impure. Impure nihilist accounts claim that only some omissions are absences, while others are identical to some entity with positive ontological status. For instance, Randy Clarke argues that sometimes, when we talk about a person’s not doing something we refer to an action, like when the description ‘not moving’ picks out someone’s holding still while playing a game. Here, Clarke claims, the phrase ‘not moving’ refers to the action of holding still (Clarke 2014, p. 21).¹² Thus, Clarke advances an impure nihilist account of omissions, because he holds that some omissions are absences while others are identical to some action. Pure nihilist accounts, however, deny that some omissions are identical to actions or any other entity with positive ontological status.

In this paper, I offer a pure nihilist account of omissions. Once my account of omissions is on the table, I will offer some reasons for thinking that pure nihilist accounts of omission enjoy certain advantages that impure nihilist accounts do not.

The pure nihilist argument I make implies that the move from FFR-1 and -4 to FFR-5 is invalid. Briefly, truth-evaluability does not require reference; rather, truth-evaluability requires either reference or reference under some pretense. ‘Omission’ tokens do refer under some pretense, hence O-type sentences fulfill one of the necessary conditions on truth-evaluability. This argument, then, undermines the move from FFR-1 and -4 to FFR-5.

5 Fiction and omission

The first step in the argument claims that the semantic function of ‘omission’ tokens is similar to the semantic function of fictional names. Roughly, this is because omissions and fictional characters are ‘non-entities’ of a distinctive sort.¹³ In this section, I make the parallels between fictional characters and omissions explicit. Thus, if we can formulate a plausible theory of the semantics of fictional names, this will help us to undermine FFR.

One of the semantic functions of proper names is to refer. Thus, at least one semantic function of ‘Saul Kripke’ is to refer to Saul Kripke (hereafter, I assume a broadly Millian view of names where the semantic function of proper names just is to refer to some entity or other). Sentences containing proper names are truth-evaluable only if the proper name refers to some unique bearer rigidly.

If proper names have this semantic function, then fictional names pose a problem. This is because fictional characters do not exist; fictional characters are nothing at all. Hence, fictional names do not refer to anything and thereby fail to fulfill their functions. This, however, is counterintuitive. To see this, consider sentences about fictional characters, some of which are true and others of which are false. Here are just a few examples: ‘Olga Zilanov’s husband is a well-known public figure in Russia’ (see Nabokov’s Glory); ‘Dick Diver is not an American’ (see Fitzgerald’s Tender is the Night); ‘Salambo is a Carthaginian’ (see Flaubert’s Salambo). The first and third sentences are true, while the second is false.

For this to be the case, minimally every singular term in sentences about fictional characters must refer. Some of these expressions will be fictional names. Therefore, fictional names appear to refer (cf. van Inwagen 1977, p. 300).

This line of thought closely parallels the earlier line of thought about omissions. The parallel becomes obvious when we amend the earlier FFR argument to fit the fiction argument. Call the amended argument the Failure of Fictions to Refer (FFR*):

• FFR*-1) Fictions are nothing at all. [premise]

• FFR*-2) Sentences have a truth-value only if every basic expression in the sentence successfully refers. [premise]

• FFR*-3) Sentences about fictions (hereafter, F-type sentences) are truth-evaluable only if fictional names within F-type sentences refer. [instance of 2]

• FFR*-4) Semantic tokens, if they refer, refer to something. [premise]

• FFR*-5) Fictional names do not refer. [1, 4]

• FFR*-6) F-type sentences are not truth-evaluable. [2, 3, 5]

• FFR*-7) To speak meaningfully about some topic T, some relevantly broad class of sentences within T must be truth-evaluable. [premise]

• FFR*-8) To speak meaningfully about fictions, some relevantly broad class of F- type sentences must be truth-evaluable. [premise]

• FFR*-9) One cannot speak meaningfully about fiction. [6, 7, 8]

• FFR* is (mutatis mutandis) identical to FFR. Given the parallel, we can justifiably conclude that any solution to FFR* will most likely be applicable to FFR. In the next section, I will consider one solution to FFR*. In the final section, I apply this solution to FFR and offer some concluding thoughts.

6 Fiction and reference

Kripke develops a theory of fictional discourse in Reference & Existence that, for the most part, tracks the main nerve of FFR*. In this section, I utilize aspects of his theory to provide a solution for FFR*.

Kripke’s solution to FFR* states that FFR*-1 and -4 do not conjointly entail FFR*-5.¹⁴ In brief, Kripke’s view is as follows. Fictional names do not simply refer; rather, they refer under a pretense. In virtue of this, F-type sentences express propositions only relative to this pretense. The pretense specifies a context in which certain prescriptions govern the way in which participants under the pretense ought to imagine things. Participation in the pretense and cooperation with the relevant prescriptions creates an object that becomes the referent of some fictional name. Reference under a pretense relates these imagined objects to fictional names. Since fictional names refer within certain pretenses, F-type sentences fulfill one condition on truth-evaluability. Each of these points deserves further discussion.

Pretense functions as follows. Within the context of telling a story, the author establishes a pretense of reference that the audience accepts when interacting with the story. When, for instance, Doyle uses the name ‘Sherlock Holmes’, the audience accepts the pretense that the name refers to some unique object rigidly. This, in short, is Kripke’s Pretense Principle. The Pretense Principle states two things: (a) fictional names refer to objects only under a pretense, and; (b) F-type sentences express propositions only relative to some pretense (Kripke 2013, p. 23). The pretense, here, has two functions. The first is to specify a context within which individuals implicitly agree to embed all relevant sentences under a sentential operator that restricts the scope of quantification to the story alone. The sentential operator will be something like ‘In A’s story...’ or ‘According to story F...’. To take an example, when we interact with stories about Sherlock Holmes and say ‘Sherlock Holmes is a great detective’, we really claim that ‘According to Doyle’s story, Sherlock Holmes is a great detective’. In ordinary discourse, we rarely explicitly mention the context denoted by the sentential operator because the context is assumed.

F-type sentences only express propositions relative to the contexts in which speakers of F-type sentences agree (implicitly or explicitly) to participate in the pretense. This is because the constituents of the propositions that F-type sentences express will only exist within the contexts that the pretense specifies. That is, the fictional entities to which fictional names refer exist only in virtue of individuals accepting the relevant pretense and cooperating with the prescriptions that dictate how one ought to imagine the referents of various F-type sentences (Kripke 2013, pp. 148–49).¹⁵

This brings us to the second function of pretense, namely the prescriptions that govern the ways in which audience members must imagine some fictional character to be. When individuals accept a pretense, they agree not only to embed relevant sentences under a special sentential operator, but they also agree to imagine the referents of certain singular terms within these sentences as being a particular way. With respect to story telling, there are various conventions that govern what we ought to imagine when someone tells a story. Thus, when Fitzgerald writes:

A mile from the sea, where pines give way to dusty poplars, is an isolated railroad stop, whence one June morning in 1925 a Victoria brought a woman and her daughter down to Gausse’s Hotel. The mother’s face was of a fading prettiness that would soon be patted with broken veins; her expression was both tranquil and aware in a pleasant way (Tender is the Night, p. 3).

we are to imagine two people in a certain location at a certain time that have a particular degree of physical attractiveness, etc. In virtue of following the prescriptions, the audience and the author create fictional entities. These fictional entities then serve as the constituents of the propositions that F-type sentences express.

One might think that this analysis of pretense concedes that fictional characters do exist. After all, fictional characters (on this view) are objects that both serve as constituents of propositions and make some semantic contribution to the truth-value of F-type sentences. In what sense, then, do fictional characters not reside on the same ontological plane as you, me, and every other actual person?

Recall that when individuals accept the relevant pretense, they agree to embed all of the relevant expressions under a special sentential operator that restricts quantification to the story alone. Quantification under this sort of operator is not ontologically committing. Thus, ‘In Doyle’s story, $\exists$xPx’ does not entail ‘$\exists$xPx’. To see this, consider an analogous case, adapted from Gideon Rosen. Suppose we are talking about Leibniz, and someone says: ‘In Leibniz’s later writings on metaphysics, tables are just colonies of souls’. This sentence can be stated, believed, and be true without entailing that the sentence ‘Tables are just colonies of souls’ be stated, believed, or even true (Rosen 1990, p. 331; cf. Williamson 2013, p. 153). The pretense that we adopt when discussing Leibniz’s metaphysics is similar to the pretense that we adopt when interacting with and discussing fiction. In all, pretense provides resources for talking ‘as if’ things were a certain way without committing us to things actually being that way.

This might seem unsatisfying. After all, if there is no fictional character, then what serves as the referent of fictional names (Salmon 1998, pp. 297–302)? Answering this question requires further discussion of the creative process that occurs when individuals cooperate with the prescriptions that dictate the way in which we ought to imagine different aspects of fiction.

First, however, let me clarify a minor ambiguity. The account of pretense that I offer allows us to say, truly, things like ‘Sherlock Holmes is a great detective’ or ‘Salambo is a Carthaginian’ without committing ourselves to there being these persons with these properties. The referent of this sort of name is a certain abstract object that an author creates through telling stories. The abstract object is a constituent in the proposition that the relevant F-type sentence expresses. This is a special kind of abstract object because the object exists contingently and depends for its existence on the world being a certain way (i.e., containing people that tells stories that conform to cultural conventions and specify certain prescriptions for imagining, etc.) (cf. Kripke 2013, pp. 69–70).

This does not imply that fictional characters exist. The abstract objects are not identical to Sherlock Holmes or Salambo or any other character. Salambo, for instance, grows weary and becomes depressed. Abstract objects do not have emotions, so no abstract object is identical to Salambo. Instead, we imagine that a certain abstract object has the qualities that the author ascribes to a certain character in a work of fiction. In general, we imagine that a fictional name refers to some object that the name rigidly designates and that this object largely resembles the character described in the novel. The name actually refers to some abstract object, but we pretend that that object has certain properties that the fictional character is said to possess.

One might worry that this still does not address the objection. Fictional characters seem to exist in virtue of some other things (pretense, abstract objects, relevant prescriptions, etc.). Why is it not the case that the fictional character just is the abstract object? To answer this objection, let’s consider the Salambo example again. Suppose we have Flaubert’s story about Salambo. Taking ‘M’ to be the fictional operator that I mention above, the story (let’s suppose) claims that:

$$\begin{aligned} \hbox {Mx} (\hbox {Sx}) \end{aligned}$$

Here, ‘S’ = ‘is Salambo’. On the view of fictional discourse that I present here, we can decompose this sentence into the following:

$$ \begin{aligned} \exists \hbox {yz}(\hbox {Ay}\, \& \,\hbox {Pz}\, \& \,\hbox {Iyz}) \end{aligned}$$

Here, ‘A’ = ‘is an abstract object’, ‘P’ = ‘is a person’, and ‘I’ picks out the two-place imagining relation. The (rough) translation reads, informally: ‘There is an abstract object that some person stands in the appropriate imagining relation to such that the person imagines the abstract object to be Salambo’. This abstract object becomes the constituent in propositions about Salambo and the abstract object (together with the imagining activity) makes some semantic contribution to sentences about Salambo. But, importantly, the abstract object is not equivalent to the fictional character, Salambo. We simply imagine that these two are equivalent (this imagining is part of the function of the pretense).

Someone might wonder whether the imagining makes it the case that the abstract object is the fictional character. Note that when we imagine things to be a certain way this is not equivalent to entertaining some counterfactual state of affairs as actual. When we imagine, we are not just picturing some way the world might have been (cf. Clark 1980, p. 347). This is because it is possible to imagine objects that do not exist in any counterfactual scenario. For instance, I just imagined that a unicorn works in the office next to mine. In doing so, I do not represent a counterfactual scenario to myself, in part because unicorns necessarily do not exist.¹⁶ Thus, the range of what one can imagine is wider than the range of counterfactual scenarios that one can represent to oneself. This argument also supports the point that the content of imagination does not entail ontological commitment. When I imagine that there is a unicorn nearby, I am not ontologically committed to there being any unicorns. Similarly, when I imagine that a certain abstract object is a fictional character, I am not ontologically committed to that abstract object being that fictional character.¹⁷

The foregoing account of pretense, then, implies two further claims. First, authors create abstract objects in virtue of telling a story that demands that individuals adopt a certain pretense when interacting with the story. These abstract objects are the referents of fictional names and the constituents of propositions that F-type sentences express. Second, in virtue of participating in the author’s pretense, we imagine that these abstract objects are characters of some sort and that these characters have certain properties (namely, those that the author describes the character as having). This account preserves two things that any theory of fictional discourse ought to preserve. First, it preserves the truth-evaluability of F-type sentences relative to the relevant pretense. ‘Salambo is a Carthaginian’ is true because under the relevant pretense we imagine that there is some person, Salambo, who is born in Carthage. Second, it preserves some plausible analysis of negative existential statements. ‘Salambo does not exist’ still comes out true because there is no object identical to Salambo (here, we also presume that the negative existential statement is not evaluated under the appropriate pretense). The imagining that occurs under a pretense does not create an object that we quantify over; rather, the imagining allows us to think ‘as if’ there is some object that satisfies the content of the imagining.

In resolving this ambiguity, we also begin to see what the creative process is that both creates the relevant abstract objects and prescribes individuals under a pretense to imagine these objects being a particular way.

We can clarify this point further. To do this, consider the example of a game. Suppose some students divide into two groups for the purpose of playing a game. One group consists of ‘survivors’ while the other group consists of ‘zombies’. The game largely resembles the game of tag, with ‘zombies’ playing the role of chasers and ‘survivors’ playing the role of runners. Let’s say that Mary is a zombie and Tom and David are survivors. At a certain juncture, Tom points out Mary to David and says: “Watch out! That zombie will hurt you.” Within the context of the game, this is true because Tom and David, just in virtue of playing the game and intending to play the game correctly, agree to follow certain prescriptions internal to the game. One of these prescriptions is to imagine Mary as a zombie and, more generally, to imagine that there are zombies that have certain intentions, such as harming the survivors. When individuals play the game, they are not ontologically committed to there being some zombie kind. The players imagine there is some kind (zombie) to which certain individuals (e.g., Mary) belong. Statements about zombies are implicitly embedded under a game operator that restricts quantification to the game alone. Tom’s statement is true because the pretense that people adopt when playing the game specifies that players ought to imagine things being a certain way. Tom and David ought to imagine Mary as a zombie. Tom’s statement, then, is true because what Tom expresses elicits a certain pattern of imagining from other participants in the game and these imaginings fit with the prescriptions specified by the relevant pretense.

Fiction seems to operate along the same lines as the game. Individuals that interact with the fiction agree to follow certain prescriptions that dictate what one ought to imagine. These imaginings do not carry any ontological commitment along with them. Pretense enables Tom to talk as if there are zombies in the same way that pretense enables us to talk as if there is some person named ‘Sherlock Holmes’ or ‘Salambo’.

The prescriptions of pretense in fiction mark an important distinction between pretense and mere pretending. The former conforms to well-established cultural conventions, and it is in virtue of these conventions that authors are able to construct, and audiences able to refer to, fictional characters. Mere pretending does not have this effect because there are no prescriptions that govern the imaginings that occur in mere pretending. Thus, pretense can generate truth-evaluable F-type sentences because pretense conforms to conventional standards of story telling. Mere pretending has no such standards, and thus cannot generate truth-evaluable F-type sentences.

The fact that some sort of cultural convention grounds the function of pretense might answer certain worries that one might have about a pretense account of fictional discourse. Jason Stanley argues that one problem with pretense accounts is that there is no systematic connection between F-type sentences and their truth-conditions. Thus, pretense accounts cannot explain our ability to grasp novel sentences evaluated within a pretense (Stanley 2001, p. 41). On the view developed here, there are resources to circumvent this systematicity concern. Because pretense functions according to established conventions, such conventions might underwrite one’s ability to internalize rules that specify functions from pretense to content. So, we might internalize the rules governing different pretenses in the same way that we internalize rules governing the use of idiomatic expressions or rules in a game.

This view of pretense also accounts for our predicating properties of fictional characters. When we say ‘Salambo is a Carthaginian’, we might think that this can only be true if there is some ontological subject that can serve as the bearer of properties. However, with the account of pretense in hand, we see that predicating properties is no problem. Predication statements are embedded under the same sentential operator as all other F-type sentences that restricts quantification to the context of pretense alone (cf. Kripke 2011, p. 65).¹⁸

This account clarifies why fictional names refer under a pretense and why F-type sentences express propositions under a pretense. Without the pretense, there is no abstract object that is imagined to be a certain way. Thus, without the pretense, there is no referent and no constituent to fill out the proposition. Within the pretense, we have both a referent and a constituent. Crucially, the pretense allows us to quantify over fictional characters without bringing in any untoward existential commitments.¹⁹

Returning to FFR*, we can see how this account of fiction undermines the inference from FFR*-1 and -4 to FFR*-5. Recall that the relevant portion of the argument runs as follows:

• FFR*-1) Fictions are nothing at all. [premise]

• FFR*-4) Semantic tokens, if they refer, refer to something. [premise]

• FFR*-5) Fictional names cannot refer. [1, 4]

The argument from this section shows that fictional characters can be nothing at all, though fictional names still refer to something. Fictional names refer to abstract objects that serve as the referent of names and the constituents of propositions that are the bearers of truth-values for F-type sentences.²⁰ Fictional names refer in virtue of occurring in sentences that are implicitly embedded under a sentential operator that restricts quantification to some pretense that individuals adopt when interacting with fiction.

7 Omission and reference

In this final section, I transpose the solution to FFR* onto FFR. Since the treatment of FFR* targeted the inference from FFR*-1 and -4 to FFR*-5, we should similarly look to adapt the corresponding section of FFR. Recall that that segment of FFR runs as follows:

• FFR-1) Omissions are nothing at all. [premise]

• FFR-4) Semantic tokens, if they refer, refer to something. [premise]

• FFR-5) ‘Omission’ tokens cannot refer. [1, 4]

We can undermine this inference if we take omissions to be a kind of fictional entity that operate in similar ways to fictional characters. This strategy leaves FFR-1 on the table and thus presumes a nihilistic account of omissions.

This is not just an ad hoc proposal. Given the argument of Sect.  2, we have good reason to explore the possibility of a nihilist theory of omissions. Thus, we have reasons, independent of FFR, to take omissions to be absences. Additionally, the striking parallel between omissions and fictional characters give us good reason to explore the possibility of omissions as a kind of fictional entity. For these reasons, the fictionalist theory of omissions is well motivated.²¹

I suggest that we take omissions to be theoretically useful fictions analogous to the concept of the center of gravity in physics (cf. Dennett 1992). Just as there is no such thing as a center of gravity, there is no thing that is some omission. To return to our original example, there is nothing that is Ben Sira’s omission of Ezra’s name. However, it is useful for us to talk about Ben Sira’s omission in the same way that it is useful for us to talk about the center of gravity.

The pretense principle developed in the previous section can supply the conventions that we need to talk ‘as if’ there are omissions. Changing the relevant terms, the pretense principle in this domain states that: (a) ‘omission’ tokens refer to objects only under a pretense, and; (b) O-type sentences express propositions only relative to some pretense.

Pretense in this domain will have the same function as pretense in the domain of fictional discourse. With respect to omissions, the idea is that we adopt some pretense within which we implicitly embed O-type sentences under a sentential operator that restricts quantification to a limited domain that does not entail ontological commitment. Here, the relevant domain would be some particular theory. The pretense also specifies that we imagine omissions to have certain properties. As discussed in Sect. 5, this does not mean that omissions are identical to or identified with certain abstract objects. The omissions result from some agent imagining an abstract object to have certain properties. This does not mean that the omission is the combination of the abstract object and the imagining activity; rather, the combination of the abstract object and the imagining activity enables us to talk ‘as if’ there are omissions.

This is similar to some accounts of modal fictionalism. Gideon Rosen, for instance, claims that fictionalists about possible worlds can embed statements about possible worlds under an operator that specifies a particular theory. So, for some sentence like ‘There could have been blue swans’, Rosen claims that the fictionalist embeds this under some operator that the fictionalist adopts as a pretense, like ‘According to theory X...’. Thus, the modal ficitonalist can claim ‘According to the modal realist, there is at least one world that contains blue swans’ without committing to the existence of concrete, spatio-temporally isolated possible worlds (Rosen 1990, p. 336).

The fictionalist about omissions can do the same thing. We can implicitly embed sentences about omissions under the sentential operator ‘According to theory X...’. The theory, here, specifies the prescriptions that dictate the way in which we ought to imagine omissions to be and supplies the conventions that enable us to talk ‘as if’ there are omissions that resemble the entities that the theory posits. Thus, talk about omissions operates along the same lines as talk about fictional characters. The author of a theory creates a certain abstract object and, in virtue of creating the theory, specifies certain prescriptions that determine the properties that we ought to imagine that abstract object to have.²²

For example, we might explain Ben Sira’s omission along fictionalist lines as follows. When we talk about ‘the omission’, we implicitly restrict our discourse to some theory that specifies what we ought to imagine the omission to be. So, when we talk about ‘Ben Sira’s omission’, we use this as a paraphrase for ‘According to theory X, Ben Sira’s omission is such and such’. This does not entail that we take omissions to be something which have such and such a character, and thus we avoid any messy ontological commitments.

Above, I mentioned that fictional talk is constrained by certain norms that govern fiction-making activity. These norms make it the case that there is a systematic connection between sentences embedded under the relevant pretense and the truth-conditions of those sentences. These norms, I suggest, are certain social, cultural, interpersonal, or intrapersonal conventions, obligations, and/or expectations. In the case of omissions, these norms constitute some of the prescriptions that dictate and constrain how we talk about omissions. For instance, Ben Sira’s omission of Ezra’s name only counts as an omission because there is some expectation, convention, or norm that dictates that Ben Sira ought to write down Ezra’s name. Thus, norms call our attention to a certain bit of behavior that we characterize as an omission only because we imagine some abstract object (e.g., a counterfactual state of affairs) to have certain properties. Without the guidance of norms, there seems to be nothing that governs our imagined ascription of properties to the relevant abstract object.²³ The implication of this is that norms are essential to omissions. This means that in order for some person, P, to omit to A, it must be the case that there is some norm that both applies to P and dictates that P either ought to or ought not A. Where there are no norms, there is no broader system of conventions that guide our fiction-making activities with respect to omissions.

Someone might object that the fact that omissions require norms in some way is too strong. Norms are certainly important to the fictionalist account of omission, but why include the stronger claim that these norms are essential? Another problem is that there seem to be obvious counterexamples to the claim that norms are essential to omissions. For instance, Clarke (2014, p. 29) advances a weaker view, stating that omissions often result from norm violation or norm compliance, but that this is not a necessary condition on omitting. That is, omitting to A does not require that one ought to or ought not to A in some sense. He gives the example of someone who omits to write needless words (Clarke 2014, p. 32). Clarke argues that this individual successfully omits to write needless words despite the fact that there is no relevant norm related to this omission.

I think there are two reasons for fictionalists about omissions to adopt the stronger view that norms are essential to omissions. The first is that the conventions that guide and constrain our omissions talk guarantee a systematic connection between pretense-embedded O-type sentences and the truth conditions of these sentences. And this allows the fictionalist about omissions to avoid concerns about the systematicity of fictionalist discourse noted in Sect. 5.

The second reason to adopt the stronger view is that the stronger view provides theoretical machinery to address certain puzzles about the metaphysics of omissions. Consider the case of Ben Sira again. There are two questions that we can raise with respect to this case. First, why does Ben Sira omit Ezra’s name and not something else (e.g., another name)? Second, why does Ben Sira and not someone else omit Ezra’s name?

With respect to the first question, the puzzling issue seems to be this. When Ben Sira omits Ezra’s name, he also fails to do a number of other things. He does not write any other name, he does not take a walk, and he does not think of his favorite childhood moment. In fact, there seem to be an infinite number of other things that Ben Sira does not do when he omits Ezra’s name. So why focus on Ben Sira’s omission of Ezra’s name? One could say that we pick out Ben Sira’s omission of Ezra’s name because there is some norm that calls for Ben Sira to write Ezra’s name. On this view, norms make salient the omissions that we care about from the infinite number of omissions that occur every moment.

This view, however, seems implausible. This is because it seems odd to say that at every moment there are an infinite number of omissions that ‘occur’ (in some sense). The stronger view seems more plausible here, namely that Ben Sira’s norm violation is part of what constitutes his omission of Ezra’s name. And this norm violation is part of what makes it the case that Ben Sira omits to write Ezra’s name rather than some other name that he did not write down.

The second question has a similar puzzling implication. Ben Sira omits to write Ezra’s name, but everybody else on Earth also fails to write down Ezra’s name. So why do we say that Ben Sira omits to write Ezra’s name? Again, one could say that everyone does omit to write Ezra’s name, but certain norms make salient Ben Sira’s omission over the others. This view seems implausible for the same reason as the view above, namely that it seems implausible to say that everyone always omits what everyone else omits. The stronger view, again, seems more plausible. The reason why Ben Sira (and no one else) omits to write down Ezra’s name is that there is some norm that calls for Ben Sira (and no one else) to write down Ezra’s name (cf. Henne et al. 2016).²⁴

Thus, the stronger view that I advocate here, namely that norms are essential to omissions, seems better-suited to answer certain puzzles about omissions than weaker views like Clarke’s. Additionally, Clarke’s main argument against the stronger view that I advocate relies on a case that does not seem to be a counterexample to my stronger view. Recall that Clarke claimed that someone might omit to write needless words without thereby following or violating a norm. This, in turn, was meant to establish the claim that violating or complying with a norm is not a necessary condition on omitting. In the example, however, one could say that there is an operative norm in the sense that there might be an expectation (whether social, interpersonal, or intrapersonal) that the individual in question avoid writing needless words. And this expectation, as a kind of norm, would partially constitute the individual’s omission of writing needless words.²⁵

Let’s turn now from the metaphysical dimension of the fictionalist view to some of the semantic dimensions. The pretense principle developed in this paper can supply the resources that we need to account for the reference of ‘omission’ tokens within O-type sentences and the constituents of the propositions that O-type sentences express. For some ‘omission’ token within an O-type sentence, we can say that the referent of the ‘omission’ term is some abstract object created by the author of a particular theory. We ought to imagine this object as having certain properties (e.g., ‘being an event’ or ‘being a component of a negatively polarized state of affairs’, etc.), and this object serves as the constituent of the proposition that the O-type sentence expresses.

One feature of this theory is that agents create abstract objects in the process of making fictions. Thus, any time someone talks about an omission, they either create an abstract object or refer to some abstract object created by some previous ‘author’-individual. Some might find this feature of the view untenable, because abstract objects are not the sorts of things that are created. This aspect of the account, however, should not be troubling. There seem to be other contexts (similar to omissions talk) where we create abstract objects. Consider, for instance, the creation of new act-types that fulfill certain functional roles within the context of social practices that enable agents to generate and utilize novel reasons in their practical reasoning (cf. Gauthier 1994; Thompson 2009, part III). For example, within the context of playing baseball, we create the new act-type denoted by the term ‘stealing a base’ that does not exist prior to the invention of the game context. This act-type is plausibly taken to be an abstract object, so the creation of abstract objects seems to be nothing fantastic. The creation of abstract objects that we imagine to be omissions seems to be an extension of this more general practice of introducing novel act-types into our ontology. If we utilize norms to construct omissions, then omissions are partially constituted by the operative demands and expectations that structure social space.

Thus, the fictionalist theory of omissions resembles many features of the aforementioned account of fictional discourse. O-type sentences refer under a pretense and express propositions under a pretense. Sentences like ‘The omission of Ezra’s name’ outside of the relevant context will be meaningless, though within the appropriate pretense these sentences will have genuine truth-values. In this way, we can undermine FFR in the exact same way as FFR*.²⁶

In Sect. 3, I mentioned that some nihilist accounts of omissions are pure while others are impure. The fictionalist account that I offer here is pure, meaning that every omission is an absence. Thus, all omissions are nothing at all. An impure nihilist account makes a weaker claim, namely that only some omissions are nothing at all. Other omissions, however, are such that when we refer to those omissions we thereby refer to an action as well. Clarke (2014) advances an impure nihilist account, which he explains as follows:

About a few cases, however, act-identification seems correct. ‘The child’s not moving’ does seem to designate her act of holding still [in a game of hide and seek]. The not moving is just as difficult as the holding still is. There is no implausibility in locating the omission where and when the action is...Some of what we regard as omissions and refrainings, then, are actions of a familiar sort, albeit such actions described in terms of things they aren’t. These omissions are things, but not negative entities; the negativity lies entirely in the expressions used to designate them (2014: 27; emphasis added).

Thus, sometimes when we refer to omissions we also refer to an action. And, insofar as Clarke seems to correctly assess the identification of not moving with holding still, this consideration seems to weigh decisively in favor of an impure nihilist account of omissions. Why, then, would one want to adopt a pure nihilist account?

One reason is that the pure nihilist account offers a unified theory of omissions. On the fictionalist view that I offer, all omissions are absences. On an impure nihilist account (like Clarke’s), omissions are either absences or identical to actions. If one thinks that unity is a theoretical virtue that offers a pro tanto reason to favor a more unified theory over a disjunctive theory, then we have a (defeasible) reason to prefer pure nihilism to impure nihilism.

If, however, pure nihilism does not accurately describe the cases, then we should opt for impure nihilism despite the latter’s inherent disjunctivism. And Clarke does seem to correctly identify the equivalence of holding still and not moving. How might a fictionalist assess the case? One response is to just bite the bullet and say that the not moving is just as much a fictional object as Ben Sira’s omission of Ezra’s name. That is, if there were no norms (conventions, expectations, obligations, etc.) that dictate that an agent not move, then there would be no such way to describe an instance of holding still as not moving because there would be no such thing as an omission of moving. Another response is to say that not moving (in the context of hide and seek where someone is holding still) is not an omission, but a refraining (Clarke even hints at this possibility in the above quotation, though he does not provide a principled distinction between refraining and omitting). One could then stipulate a distinction between refraining and omitting and claim that refrainings can be identical to actions, though omissions cannot. This response is coherent, though successful only by fiat. The upshot of this discussion, however, is that pure nihilist accounts of omissions enjoy a theoretical virtue that impure accounts do not (namely, unity) and can furnish different assessments of the cases that seem to motivate an impure nihilist account. These considerations seem to give a slight dialectical advantage to pure nihilist accounts of omissions.

I will address one final concern. Someone might wonder how to fit the fictionalist account of omissions to some theory of moral responsibility for omissions. If omissions are really nothing at all, then what is it we are responsible for when we omit something? This, I take it, is a normative consideration that is settled by our practices of assigning responsibility ascriptions. These normative considerations are isolated from the metaphysical and semantic considerations offered here. Thus, one can generate a theory of responsibility for omissions without adopting a substantive theory of what omissions are (and vice versa). For this reason, I can set aside the issue of what it means to be responsible for an omission.²⁷

8 Conclusion

In conclusion, this paper argues that we can solve the puzzle raised by FFR if we adopt a fictionalist view of omissions. The view requires us to interpret omissions as a kind of fictional entity similar to fictional characters in stories. This paper also suggests that there are rich parallels between discussions of fiction and omissions. Given the increasing interest in omissions as a topic within the metaphysics of agency, we would do well to draw on resources developed in discussions of the metaphysics of fiction.