MENTAL FILES AND METAFICTIVE UTTERANCES

Metafictive utterances raise a kind of intuitions (intuitions of truthfulness) that pose a problem for a view that combines a referentialist approach to proper names with an antirealist stance on fictional characters. In this article I attempt to provide a solution to this problem within the framework of mental files. According to my position, metafictive utterances literally express an incomplete proposition while pragmatically conveying a complete one, which accounts for the intuitions of truthfulness. The proposition pragmatically conveyed is ‘metarepresentational’, I’ll argue, in the sense that it is about a mental representation or mental file.

"any acceptable account of empty names will either have to accommodate them, or explain them away" (everett 2003, p. 2). now, although both sets of intuitions raise interesting questions and should be accounted for, in this work i will only be concerned with the intuitions of truthfulness. 2 referentialism about proper names holds that the only semantic contribution of a proper name to the semantic content of an utterance in which it occurs is its referent. typically, referentialism is in tandem with a russellian view of propositions, that is, the view that propositions are structured entities constituted by objects, properties and relations arranged in a certain way. on this view, utterances containing a proper name express a proposition partly constituted by the object which is the bearer of the name. now, many philosophers also advocate the view that fictional names are genuinely empty, that is, that there are no fictional characters that are the bearers of fictional names (be that possible objects, non-existent objects or abstract artifacts, among others). these three thesis taken together (referentialism, a russellian view of propositions and antirealism concerning fictional characters) make it hard to account for intuitions of truthfulness concerning utterances (1)-(4). 3 If antirealism is correct, then fictional names are genuinely empty, thus they make no contribution to the proposition expressed by utterances containing them (referentialism), so it seems that those utterances do not express complete truth-conditions (russellian view of propositions), contrary to our intuitions. The challenge is then either to explain the intuitions away or to find a suitable content that allows us to take the intuitions at face value.
the aim of this article is to present a solution to the problem, within the framework of mental files, in line with the second option. In the next section I will sketch the framework of mental files, and I will rephrase the problem within that framework. in the third section, i will discuss recanati's idea of metarepresentation. in the fourth section i will present my own account, according to which metafictive utterances pragmatically convey closed metarepresentational singular thoughts. Finally, i will draw some conclusions.

Mental Files
The mental file metaphor has been elaborated by several philosophers in different ways. One recent and influential account of mental files is that of recanati (2012). 4 According to this view, entertaining a singular thought just is to entertain a thought that deploys a mental file. Mental files are singular, nondescriptive modes of presentation that enable us to gain and store information about objects. Mental files do not refer satisfactionally, however (that is, to whatever object satisfies the information contained in the file), but relationally (they refer to the object to which the subject stands in a certain epistemically rewarding relation to). now, concerning singular terms in natural language recanati says What about singular terms in language? they occur in sentences, and sentences express (and elicit) thoughts. From the interpreter's point of view, to understand a sentence is to entertain a thought. if the sentence contains a singular term referring to an object a, thinking the relevant thought involves deploying a file also referring to a. (recanati, ms, pp. 1-2) The framework maintains referentialism and russellianism for mental files and thoughts: first, analogously to singular terms, the file sole contribution to the content (i.e. the truth-conditions) of the thought in which it occurs is the object to which it refers; secondly, analogously to utterances, singular thoughts express singular russellian propositions partly constituted by the object which is the referent of the file. Now we can restate the problem of metafictive utterances in terms of mental files: according to the mental file framework understanding a sentence that contains a singular term referring to 'a' consists in entertaining a singular thought that deploys a mental file referring to the very same object. Thus, understanding a metafictive utterance consists in entertaining a singular thought that deploys a mental file referring to the same object as the fictional name. Now, if fictional names are genuinely empty then the corresponding files are also empty, thus, the relevant thought, that is, the thought we should entertain in order to understand the utterance, does not express a truthconditional content. Again, this goes against our intuitions.
Let me put it in other words. Metafictive utterances are intuitively truthevaluable, so when we understand a metafictive utterance we should, according to the mental file framework, entertain a thought that is truth-evaluable. But the mental file that corresponds to an empty name is also empty and the file's sole semantic contribution to the proposition expressed by the thought is its referent. Thus, we have the intuition that metafictive utterances are truthevaluable but we cannot account for it in terms of the thoughts elicited by the utterance, as it is expected within the mental file framework. Now, after briefly discussing recanati's view in the third section, i will propose in section 4 a way of amending the mental file framework in order to satisfactorily account for metafictive utterances.

Recanati about empty singular terms and thoughts
in his (ms) recanati discusses some cases of thoughts and utterances containing empty files and singular terms. Though he is not concerned with metafictive utterances but with attitude reports, it is worth considering what he says about it, for it will be helpful for presenting my own proposal. As we have already said, utterances containing empty singular terms elicit singular thoughts (that is, the subject thinks via a mental file) but it seems that they do not express complete truth-conditions, for in the absence of an object the thought does not express a singular proposition. However, it is well known that sometimes uttering a sentence that contains an empty singular term does not prevent one from expressing a truth-evaluable content. the kind of example recanati is interested in is (5) Leverrier thought that the discovery of Vulcan would make him famous.
to explain how we can account for the truth-conditions of an utterance like (5) in terms of the thoughts elicited by it, recanati postulates a derived function of mental files. The primary function of a mental file is to serve as a repository of information about a single individual to which it refers, information that is gained through the epistemically rewarding relation that individuates the file's type. However, Recanati claims, files can play derived functions. One such function is a metarepresentational one: files allow us to think about objects in the world but also allow us to represent how other subjects think about objects in the world. this function is supposed to explain all 'intentional' uses of singular terms. in order to provide such an explanation the notion of indexed file is needed. An indexed file consists of a file and an index: the file is the file that another subject uses in thinking about an object, the index refers to the subject to whom the file belongs. By appealing to indexed files, Recanati accounts for difficult cases involving attitude's ascriptions, negative existentials and Geach cases.
A further distinction is between 'loaded' and 'unloaded' indexed files. A 'loaded' indexed file is an indexed file that is linked to a regular file referring to the same thing, in the mind of the subject. In those cases the file has existential import and it preserves the reference it has in the subject to whom the file belongs. in that case, the subject has two ways of thinking of the object: on the one hand her own file, on the other, a vicarious file indexed to another subject. 'Unloaded' indexed files are indexed files that are not linked to a regular file in the mind of the subject; those indexed files do not have existential import. 'Unloaded' files are fully opaque, that is, they do not refer to anything. There are for recanati only two cases in which an utterance containing a singular term is associated with an 'unloaded' indexed file: First option: the utterance does not express a genuine thought, but only a 'mock thought', as Frege puts it. if i say to my children: 'Santa Claus is coming tonight', i do not express a genuine singular thought. i only pretend to refer to Santa Claus, and to predicate something of him. (the same thing is arguably true if, echoing my children, I tell my wife: 'Santa Claus is coming tonight'. Here the file associated with 'Santa Claus' is indexed to Santa-Claus believers and unloaded, so the whole speech act has to be seen as a form of pretense.) Second option: the utterance expresses a thought that is globally metarepresentational -it is about someone's, e.g. my children's, representations, rather than about what these representations are about. this corresponds to pseudo-singular belief ascriptions. i think negative (and positive) existentials too are meta-representational... (recanati, ms, p. 16) Unfortunately, the derived metarepresentational function of files, in the way Recanati presents it, does not seem adequate to account for metafictive utterances. On the first option, the whole speech act is analysed as a form of pretense. Now, this is plausible for fictive or parafictive utterances, such as 'Santa Claus is coming tonight' (said by a father to his children in order to generate expectations on them), but it doesn't seem to fit well with metafictive utterances in general. Suppose that believing that my son is old enough to know the truth I say to him, "Son, Santa Claus is a fictional character". It would be at least controversial to claim that in saying "Santa Claus is a fictional character" the whole speech act is just a form of pretense, i.e. that i'm just pretending "that there is a character named Santa Claus and pretending to predicate about him that it is a fictional character". To the contrary, my assertion seems to be as serious as it gets. Furthermore, claiming that the whole act is just a form of pretense is at odds with one of the features of metafictive utterances we pointed out before, namely, that metafictive utterances seem to be true in the context of real world.
The second option does not fare better. Metafictive utterances are not globally metarepresentational, that is, in saying that Santa Claus is a fictional character with a metafictive force I am not ascribing to anyone the pseudosingular belief that Santa Claus is a fictional character, for it is me who has that belief. When I said to my son 'Santa Claus is a fictional character' I am not ascribing him the pseudo-singular belief that Santa Claus is fictional character, but trying to communicate my belief to him. It is sufficiently clear that there are at least some metafictive utterances that do not have the form of an attitude report nor can be interpreted as (pseudo-singular) belief ascriptions.
A third option could be that the thought expressed by a metafictive utterance is only locally metarepresentational, that is, when i say to my son 'Santa Claus is a fictional character' I deploy a file that does not have existential import but is nevertheless linked to one of my own regular files (that is, it is linked to my Santa Claus' file, which I share, in a sense, with my son). this possibility is precluded by recanati's framework, however; on his account, an indexed file is either 'loaded', in which case it's linked to a regular file in the subject's mind and has existential import, or 'unloaded', in which case it does not have existential import but it's not linked to a regular file in the subject's mind neither.
So it seems that in performing metafictive utterances we are not just echoing how other person thinks about an object (for those cases are always 'loaded', according to recanati), nor we are globally representing someone else's belief (for metafictive utterances are not content-reporting) nor are we just pretending to assert something. thus, it seems that what recanati says about metarepresentation is not enough to account for metafictive utterances. Still, as I'll argue in the next section, a different but related idea of metarepresentation can be developed in order to solve the problem of metafictive utterances.

Open Metarepresentation vs. Closed Metarepresentation
in the previous section i reviewed some of the ideas about metarepresentation elaborated by recanati in order to account for several cases involving empty singular terms and thoughts. i argued, however, that it is not plausible to extend that idea in order to cover metafictive utterances. In this section, I will claim that a different but related idea of metarepresentation could be helpful in solving the problem raised by metafictive utterances within the framework of mental files.
i want to introduce a distinction between open metarepresentation and closed metarepresentation. this distinction parallels that between open and closed quotation. 5 in open quotation, the agent uses the quotation marks to represent how other subject thinks about an object, or what she says about an object. For example: (6) Hey, 'your sister' is coming over. one can use (6) in order to say that Ann is coming over, knowing that Ann is not the addressee's sister, with the intention of ironically echoing how a third partner thinks about Ann. recanati thinks of metarepresentation in analogy with open quotation. the thought involved in the comprehension of an utterance such as (6) is metarepresentational in recanati's sense because the speaker is echoing another subject's way of thinking about Ann. in the following passage recanati makes this analogy clear: Standardly, quotations are opaque: the expression in quotes refers to itself, rather than to its ordinary referent. this, at least, is true of the central class of quotations which i dubbed closed quotations. Indexed files behave differently. While indexed, the file still refers to its ordinary referent, that is, it still refers to the object the simulated file is about. In standard instances of opaque attitude attribution, a singular term in the embedded clause evokes a file in the ascribee's mind and refers to the referent of that file (not to the file itself) [...] Indexed files can still be treated as a quotational device, but the type of quotation at issue has to be open quotation, not closed quotation. Open quotations have an echoic character but, typically, the quoted words keep their ordinary meaning and reference while evoking or echoing the words of some other person or persons. (recanati, ms, p. 15, my emphasis) However, as i have previously argued, open metarepresentation is not enough in order to account for metafictive utterances. It seems that in saying that Santa Claus is a fictional character I am not merely echoing my son's way of thinking about Santa Claus, nor ascribing him the pseudo-singular belief that Santa Claus is a fictional character, nor am I just pretending to assert that Santa Claus is a fictional character.
Closed quotation behaves differently. Consider the following sentences: (8) Diego is tall. (9) 'Diego' has six letters. (8) and (9) exemplify the well-known difference between use and mention. In the first case, the speaker says something about the referent of the name, while in the second case she says something about the linguistic representation 'Diego'. in closed quotation, the quotation marks are not used for echoing other subject's way of speaking about an object, but for making the expression itself the object of the discourse. the speaker displays an expression in order to talk about the expression itself. Closed metarepresentation parallels closed quotation, although there is no mental analogous to the quotation marks. i claim that an agent can entertain a mental file with a closed metarepresentational function, in the sense that the thought is about the file itself. So sometimes when a subject entertains a singular thought, she deploys a mental file in order to think about the file itself, that is, in order to turn the representation itself into the topic of the thought. In those cases, the mental file serves both as a constituent of the thought and as part of its content: the proposition expressed by a thought containing a mental file with a closed metarepresentational function contains the file itself as a constituent, that is, it has the file itself as its topic or subject matter.
now, if we grant that we can have closed metarepresentational thoughts, I will argue, we can solve the problem of metafictive utterances. The idea I want to explore is that metafictive utterances pragmatically convey closed metarepresentational thoughts: that is, although the literal content of a metafictive utterance is not fully propositional, there is a complete singular proposition pragmatically conveyed, namely, the content of the closed metarepresentational singular thought entertained by the speaker. if this is correct we can honor the intuitions of truthfulness.
First of all, a speaker competent with a given fictional name -let's say, 'Santa Claus'-entertains a singular thought that deploys the file SANTA CLAUS. As I said above, sometimes the file does not play its usual referential function but a derived, (closed) metarepresentational one. in those cases, the thought that the speaker entertains is not intended by her neither as contentreporting (as in the case of thoughts that correspond to parafictive utterances) nor as fulfilling its usual referential function. In turn, the speaker entertains a singular mental representation in order to think about the representation itself. So for example, the speaker might entertain a singular thought that deploys the file SANTA CLAUS, knowing in advance that the file is empty, in order to ascribe the property 'became famous because of the Coca-Cola company' to the file SANTA CLAUS itself.
then, the speaker utters, for example, something like (1). the closed metarepresentational thought entertained by the speaker is not the literal content of the utterance, though. As i said before, i maintain that, literally, metafictive utterances express an incomplete content, for the fictional name is genuinely empty. in turn, i claim that the closed metarepresentational thought is pragmatically conveyed by the utterance. How does the process go? in the first step the speaker entertains a closed metarepresentational thought, in the sense previously explained. then, the speaker utters, for example, something like (1). As i said above, the semantics of (1) is not the same as that of the thought: (1) expresses only an incomplete or 'gappy' proposition, for the name is genuinely empty and thus it does not contribute anything to the truth-conditional content of the utterance. When the speaker utters (1) she intends the hearer to entertain a singular thought deploying the file SANTA CLAUS. Arguably, upon hearing the physical token of the name the hearer will infer that the speaker is thinking via a singular, non-descriptive mode of presentation, and that the speaker wants her to think about the object in the same way. this is due both to linguistic and psychological features of names. 6 So the hearer will open a mental file and label it with the name she took from the speaker. Following the example, upon hearing (1) the audience will produce a mental token of that sentence and will open a mental file SANTA CLAUS (or, if she already possesses a Santa Claus' file, she will activate it). on the other hand, when the speaker utters (1) she has the intention that the hearer infers a closed metarepresentational thought, a thought that is about the file itself and expresses the proposition that SANTA CLAUS became famous because of The Coca Cola Company. So the speaker utters a sentence that expresses only a 'gappy' proposition but has the intention that the hearer form the same closed metarepresentational thought that she has. now, in the successful cases the speaker utters (1) with a metafictive intention and this is mutually manifest for both speaker and hearer. thus, given the hearer's recognition of the speaker's intentions together with further knowledge of the context and the semantic information provided by her comprehension of the sentence, the hearer will typically be able to infer, in the face of an utterance like (1), a thought deploying the mental file SANTA CLAUS with a derived, closed metarepresentational function.
I claimed that the speaker, in performing a metafictive utterance, has the intention to convey a closed metarepresentational thought; i claimed also that the hearer is in turn able to pragmatically infer the required thought. in order to make this work, it is important that the hearer is able to capture the speaker's intentions. to grant that, the only thing we need to assume is that competent speakers are able to differentiate fictive, parafictive and metafictive utterances in virtue of features of context, and this is not a very controversial assumption. In effect, that the speaker has the intention to convey a closed metarepresentational thought (that is, that the utterance is a metafictive one) might be the most relevant interpretation of her utterance. By way of illustration, consider an utterance of (1). First of all, we can plausibly assume that the hearer is able to reliably dismiss a fictive interpretation of (1) just by judging whether the conventional requisites for being a legitimate producer 7 in the Santa Claus' stories are fulfilled. What about a parafictive interpretation? Suppose further that in the context it is common knowledge for speaker and hearer that in the story Santa Claus did not become famous because of the Coca Cola Company._ This fact precludes a parafictive interpretation on the part of the hearer. the hearer will assume that the speaker, if she intends (1) as a parafictive utterance, will try to stick to the facts of the story and will not try to mislead the audience. So if the speaker explicitly states something that is false in the story and it is common knowledge that it is false in the story, then she surely does not intend the utterance as parafictive. Thus, a metafictive interpretation seems to be the only option left (besides, hearer's further knowledge of context, including her knowledge of the speaker's inferential behavior and dispositions to act, might make a metafictive interpretation the most relevant one). Now, as metafictive utterances exhibit a closed metarepresentational character, the hearer will infer a metarepresentational proposition about the file SANTA CLAUS. This is just an example. Different circumstances might make it manifest for the hearer that the speaker has the intention to convey a closed metarepresentational thought. the important point is that to grant regular speakers the ability to capture the metafictive force of an utterance is not a very demanding condition.
So the proposal is that in uttering (1) the speaker pragmatically conveys a singular metarepresentational proposition. In a first step, the hearer entertains a singular thought that deploys the mental file SANTA CLAUS. In a second step, the hearer goes (pragmatically) from this empty thought to a proposition that includes the file itself as a constituent, after inferring that the file fulfills in this occasion a closed metarepresentational function. i say that the proposition pragmatically inferred is singular because it includes the file, which is a type of object. i say that the proposition is metarepresentational because it is about a file, which is a kind of mental representation. I say that this proposition is pragmatically conveyed because it is not the literal content of the utterance by means of which it is communicated.
now, i claimed that it is the closed metarepresentational thought pragmatically conveyed the one that is the target of the intuitions of truthfulness. When a speaker utters (1) we have the intuition that the utterance is true/false because we pragmatically infer a closed metarepresentational thought that expresses a proposition about the file itself.
A possible objection suggests itself though. it is not plausible, one might think, that what a speaker is saying in uttering (1) is that her own file became famous because of the Coca Cola Company. thus, even if my proposal successfully identifies a singular proposition as the target of the intuitions of truthfulness, it is not the proposition intuitively grasped by any competent speaker. in the next section, i tackle this and others objections.

Objections
I have claimed that in metafictive utterances such as (1) the speaker entertains a thought that is about the file SANTA CLAUS itself, while the hearer, if communication is successful, manages to form an analogous thought in her own mind. Now, this idea raises two problems. The first one has to do with the intersubjective identification of files: if my picture of metafictive utterances is granted, speaker and hearer deploy thoughts each one about her own file, so how is it that they are thinking about the same thing, viz. Santa Claus? the second problem is the following: it is not plausible that what the speaker is saying in uttering (1) is that her own file became famous because of the Coca Cola Company. one might worry that the proposition we ended up with is not the right one, for it seems to be false that my own file became famous because of the Coca Cola Company, while it seems to be true that Santa Claus became famous because of the Coca Cola Company. Let me begin with the first problem.
The problem of accounting for the intersubjective identification of files is very complex and is far beyond the goals of this article, but let me say at least something. The idea that a mental file can be shared by different subjects is not in principle very controversial. if singular terms in the language can be shared and mental files are the mental counterparts of singular terms, it does not seem crazy to believe that mental files can also be shared. There are at least two different ways of elaborating this idea. On the one hand, Stacie Friend (2011, ms) presents a Perry/Evans inspired informational account of how different individuals can share a notion (in Friend's account, a notion is pretty much like a mental file: it is a cognitive particular which is about a single individual, and it's used to bundle information about it). According to this account, when different individuals think about Santa Claus they deploy different tokens of the same notion type. What makes it the case that two tokens are tokens of the same notion is that they are embedded in the same notion-network. non-referring notion-networks originate in an author's freely created representation, i.e., a notion, associated to freely created information to identify and track characters in stories. non-referring notion-networks are individuated, according to Friend (ms, section 6) by their dominant source of information. in order to develop this idea, Friend appeals to evans' distinction between producers and consumers. Producers are all those subjects that can legitimately introduce information about x in a network about x. in the case of non-referring networks such as those originated in myths and works of fiction which individuals are producers hinges on heavily context-dependent social rules and conventions. This is compatible with my account of metafictive utterances. Thus, when I say to my son 'Santa Claus is a fictional character' what grants that communication is successful, even though i have in mind a proposition that is about my own file and my son has in mind a proposition that is about his own file is that our files are embedded in the same notionnetwork, that is, the information in the files dominantly derives from legitimate producers.
On the other hand, Sainsbury (2005) and Salis (2013) defend a different account based on the idea of name-using practices. A name-using practice is the activity performed in a certain linguistic community of using a name that is about one and the same thing or that purports to be about one and the same thing. According to Sainsbury's approach, name-using practices are individuated by baptisms and baptisms can be empty. two tokens are tokens of the same name if they participate in the same name-using practice, that is, if they have the same origin. in addition, proper names are associated with mental files, which can be identified in terms of the names to which they are associated. now, this account is also compatible with my own. Singular terms elicit mental files, which are their mental counterparts. On this account, I have a Santa Claus' file which is associated to the name 'Santa Claus', a name that participates in a given name-using practice. When i say to my son 'Santa Claus is a fictional character', he forms his own Santa Claus' file, upon hearing my token of the name Santa Claus. My son's Santa Claus file is the same as mine (that is, is a different token of the same type) because it was originated upon hearing a token of a name that participates in the same name-using practice that the token of the name upon which i myself formed my own Santa Claus' file. So the important point is that the problem of intersubjective identification does not seem pressing for my account, for there are two main strategies for accounting for it and both are compatible with my account.
The second problem was that the content I identified does not seem to be the right one, for it seems to be clearly false, for example, that my own file became famous because of the Coca Cola Company, while (1) is intuitively true. The problem is that mental files are mental particulars (to which only I can have access), and it doesn't seem right to claim that when i say something like (1) i'm talking about my own private representations, to the contrary, i must be talking about a representation that is public and shared by other members of the community. the answer to this problem is related to the answer to the first one. Although mental files are mental particulars, so that my own Santa Claus' file is different from everybody else's, they are all tokens of the same type. So when I entertain a thought about my file SANTA CLAUS i manage to entertain a thought about a representation that is public and shared at least by my audience, be that because it is part of the same notion-network, as in Friend's approach, or because it participates in the same name-using practice, as in Sainsbury's (choose your preferred theory). thus, when i entertain a thought about SANTA CLAUS I'm having thoughts about a public representation type, although i can only do that by deploying a particular token of that representation in my head and making it the object of my thinking.

Conclusion
In this article I attempted to solve, within the framework of mental files, a problem that concerns any theory that combines referentialism and a russellian view of propositions with an antirealist stance regarding fictional characters. i claimed that postulating a closed metarepresentational function for mental files can help us solve the problem. With this in mind, I argued that metafictive utterances pragmatically convey a closed metarepresentational thought, that is, a thought where the file that corresponds to the singular term in the sentence becomes the topic of discourse. the proposition pragmatically conveyed, i contended, is the one that accounts for the intuitions of truthfulness. Finally, i tackle some possible objections concerning the intersubjective identification of files and the general plausibility of the account.