Jody Azzouni’s Talking About Nothing (henceforth, TAN) is an extraordinarily ambitious book in both its breadth and its depth. To put it in a nutshell, TAN makes a sustained case for a novel, unified approach to our talking and thinking about mathematics, fiction and hallucinations. In the process, TAN touches on a wide variety of topics in areas ranging from logic to the philosophies of mathematics and science, and from metaphysics and epistemology to the philosophies of mind and language and gets to challenge a few central tenets of contemporary analytic philosophy. Since my contribution to this book symposium cannot do justice to its breadth and depth, it will focus mainly on three specific aspects of the book – Azzouni’s proposal with respect to fiction (which I will critically discuss in §2), his argument against pretence theories of fictional discourse (which I will evaluate in §3) and his ‘blitz’ argument against supervenience-based approaches to inter-level metaphysics (which I will question in §4) – in an attempt to explain why I found Azzouni’s proposal interesting but ultimately not convincing.
2. What do we talk about when we talk about nothing?
Virtually all human societies produce and consume fiction in some form or other (storytelling, plays, novels, comics and movies being just a few examples of different forms of fiction) and producers and consumers of fiction often find themselves seemingly thinking or talking about fictional characters such as Odysseus, Mickey Mouse or Norman Bates. This widespread practice seems to raise a philosophical dilemma that I will call the fiction dilemma. On the one hand, we’d like to deny that, say, Odysseus exists; on the other, we’d be inclined to say that sentences such as and are true (while sentences such as and are not).
Odysseus is the main character of the Odyssey,
Odysseus is the father of Telemachus
Odysseus is the main character of Don Quixote
Odysseus is the son of Telemachus
The dilemma arises from the fact that we wouldn’t seem to be able to have it both ways. How could something be truly predicated of the referent of ‘Odysseus’ if there is no such thing as the referent of ‘Odysseus’? When faced with the fiction dilemma, philosophers typically opt for one or the other of its horns. Inflationists claim that, although ‘Odysseus’ does not refer to any actual concrete person, it still refers to a real entity of some sort or other and they appeal to this entity to try to explain the seeming discrepancy between the truth-values of (1) and (2), on the one hand, and those of (3) and (4), on the other. Deflationists, on the other hand, deny that ‘Odysseus’ refers to anything at all and are happy to deny that sentences such as (1) and (2) can be literally true.
Not all philosophers, however, take the fiction dilemma by the horns. Biflationists1 try to go between the horns of the dilemma by striking a third way between inflationism and deflationism. Azzouni is one of them. Although Azzouni agrees with deflationists that fictional entities have no ontological status whatsoever, he nevertheless agrees with inflationists that sentences about those objects can be (literally) true. Azzouni’s key move is to deny a crucial assumption shared by both inflationists and deflationists (and by the vast majority of analytic philosophers) – i.e. the assumption that existence is a condicio sine qua non for reference. In order to do so, Azzouni distinguishes an ontologically committing sense of ‘refer’ (which he labels ‘referr’) from an ‘empty’ one (which he labels ‘refere’) (and does the same for other semantic or intentional notions such as ‘about’). Azzouni claims that while everything that is referredr to must exist, what is merely referrede to does not. So, for example, according to Azzouni, although ‘Odysseus’ (unlike, say, ‘Ithaca’) does not referr to anything, it referse to Odysseus.
But, if Odysseus does not exist, how can sentences such as (1) and (2) be just as true as sentences such as which are exclusively about(r) what exists? Azzouni’s answer is that (1) and (2) just like (5) have what he calls truth-value inducers (henceforth, TVIs) that determine whether or not the sentence is true. However, whereas the TVIs for sentences such as (5) are ‘… the objects in the world of which the truth values of the sentence[s] are in virtue’ (Azzouni 2010: 18), the TVIs for sentences such as (1) and (2) are ‘a blend of relevant objects that exist – us, our language, and our epistemic practices included – and that jointly yield the indispensability of the truth [of those sentences] to our assertoric practices in ordinary life and our sciences’ (Azzouni 2010: 18).
Ithaca is an island in the Aegean Sea,
At first, Azzouni’s proposal might seem tempting. Why shouldn’t we have our cake and eat it too if we can do so? The problem is, of course, that it is far from clear that we can do so. I must confess that I usually become suspicious when philosophers start appending superscripts to ordinary language expressions. This is not because I doubt that ordinary language is often ambiguous or that trying to identify and resolve such ambiguities is part of the task of philosophy, but because the claim that a certain expression is ambiguous is often itself ambiguous.
The claim that we should distinguish between two senses of ‘about’, for instance, is ambiguous between a descriptive and a prescriptive interpretation. It is not clear whether the claim is that ordinary English speakers (more or less consciously) do in fact distinguish two senses of ‘about’ (i.e. that they use ‘about’ in two different ways, which correspond to Azzouni’s ‘aboutr’ and ‘aboute’) or that they should distinguish two senses of ‘about’ even if in fact they don’t (i.e. even if, in fact, their use of ‘about’ coincides with Azzouni’s use of either ‘about(r/e)’ or ‘aboutr’).
Although Azzouni seems to favour the prescriptive interpretation, I will argue that no matter how we resolve the ambiguity between the descriptive and the prescriptive reading, the claim seems to be highly implausible. Consider the descriptive interpretation first. What evidence do we have to think that ordinary English speakers distinguish between two senses of ‘about’ that roughly correspond to what Azzouni calls ‘aboutr’ and ‘aboute’? As far as I can see (and Azzouni seems to agree), the answer is ‘Next to nil’. In fact, there seems to be evidence that they don’t (and the grip that the fiction dilemma has on us would seem to be a consequence of this). To see why, consider, for example, the following scenario. X, Y and Z are three friends who are talking about the similarities between a movie character, A, and a common friend, B. At some point, X leaves the room for a few minutes, missing part of the conversation. Upon coming back, X hears Y saying something to Z but she is not sure whether Y was talking about A or about B. In these circumstances, it would seem perfectly natural for X to ask Y: The problem, however, is that the ‘about’ in (6) cannot be coherently interpreted as either Azzouni’s ‘aboutr’ or ‘aboute’, because, according to Azzouni, X and Y were either talking aboute A or aboutr B but could neither have been talking aboutr A or B (for A does not exist and therefore cannot be talked aboutr) nor aboute A or B (for B exists and therefore, if X and Y were talking about her, they would be talking aboutr her). This is only one possible example of many that seem to undermine the descriptive interpretation of the claim.
Were you talking about A or B?
The prescriptive interpretation, however, does not seem to fare much better. As we have seen, according to the prescriptive interpretation, even if speakers do not in fact distinguish between those two senses of ‘about’ and related intentional notions, they should do so. The problem with this reading of Azzouni’s proposal is that the proposed revisions to ordinary language would be neither sufficiently broad nor sufficiently deep. The proposed revisions would not seem to be sufficiently broad because the problem would seem to extend much beyond the handful of intentional and semantic notions on which Azzouni chooses to focus. Even if we were to accept that X, Y and Z were talking aboute A and aboutr B and that X should have asked if Y was talking aboute A or aboutr B, this would not seem sufficient to prevent ordinary language from running into trouble. After all, Y might have said that A is more outgoing than B, and X might have argued that the similarities between A and B are not as striking as Y and Z seem to think. But, if A does not exist, how could A be more outgoing than or similar to B in the first place? And, if the best strategy in the case of ‘about’ is to distinguish between two senses of the word, why shouldn’t we also distinguish two senses of ‘between’, ‘be more outgoing than’, etc.? And, if we do, why stop there? How can A be tall, intelligent or married if she doesn’t exist? Shouldn’t we go farther and distinguish between two senses of ‘tall’, ‘intelligent’ or ‘married’ (one which applies to what exist and the other to what does not)? Where should we stop appending superscripts to ordinary words? It is far from clear that we can stop once we have appended them to all words expressing semantic or intentional concepts (as Azzouni seems to assume), because one of the main reasons why we find talking and thinking ‘about’ fictional entities philosophically problematic (if there are no such entities) is a very general reason and extends to non-semantic, non-intentional notions as well. The problem stems form the fact that there can be no relation without relata and that, on the face of it, all of the above notions can be plausibly construed (in some way or other) as relations that lack (at least) one of their relata. To take the most controversial case, ‘A is tall’, for example, may be construed as being true if and only it the exemplification relation holds between A and the property of being tall or, more modestly, if and only if the satisfaction relation holds between A and the open sentence ‘x is tall’. So, it would seem to be far from obvious that we can draw a principled line between the portion of our ordinary language that needs to be revised in order to accommodate our apparent practice of talking and thinking about fictional characters and the one that does not.
If the proposed revisions do not seem to be sufficiently broad, they do not seem to be sufficiently deep either. All the revisions would seem to do is to merely brush the philosophical problem under the terminological carpet. Azzouni starts by introducing the notion of what it is for something (such as a sentence or a thought) to be aboutr something else. From what he tells us, what he means by ‘aboutr’ seem to coincide by and large with what philosophers usually mean by ‘about’. However, Azzouni tells us that (at least in some cases) if that ‘something else’ (the something our thoughts or sentences are supposedly aboutr) happens not to exist, then our sentences and thoughts can still be ‘about’ something even if (I would be tempted to add ‘in fact’) they are aboutr nothing at all. They can do so by being aboute that ‘something’. But how is this proposed revision supposed to provide us with anything more than a whitewash? Whether or not we introduce ‘aboute’ in our language, the problem is still that, unlike our thoughts and sentences aboutr Ithaca, our thoughts and sentences aboute Odysseus seem to be aboutr nothing at all (at least if we concede, as both Azzouni and deflationists do, that Odysseus does not exist). The fact that we can define a technical notion that is satisfied by fiat by certain sentences and thoughts and that we can label that notion ‘aboute’ does not seem to provide any genuine relief from the worry that those sentences are no more about(r) Odysseus than they are about(r) Ithaca, Mickey Mouse, or the highest prime because, ultimately, they are about(r) nothing at all.
It would seem that, just as we can’t define our way out of, say, the Liar’s paradox by distinguishing two senses of ‘untrue’ (one that applies to false sentences and the other one that applies to Liar-like sentences), so we can’t define our way out of the puzzle posed by talking and thinking about fictional entities by distinguishing two senses of ‘about’ either. This is because, in both cases, one of the two senses of the word would seem to have little or nothing in common with the other beside the fact that they happen to be both designated by words that are spelled and pronounced (almost) in the same way. Azzouni seems to leave little doubt about this when talking about ‘referr’ and ‘refere’ in the following passage and many other similar passages throughout TAN:
After reading this passage, one might wonder why we should call referencee ‘referencee’ as opposed to, say, ‘deferencee’ or ‘preferencee’, for referencee seems to have little more in common with our ordinary notion of reference (roughly what Azzouni labels ‘referencer’) than the name.
Referencer is a … relation between terms … and items in the world. … Referencee, on the other hand, isn’t a relation at all. A term that referse bears no referentialr relation to anything, and consequently it isn’t aboutr anything either. (Azzouni 2010: 220)
Some of Azzouni’s remarks, however, might indicate that relationship between referencer and reference is closer than passages such as the one above suggest. Since I’m not sure my interpretation of these remarks is correct, however, let me quote in full one of the most explicit passages in this direction. Azzouni writes:
In the above passage, Azzouni seems to be suggesting that semantics (or more precisely the ‘semantics module’ in our heads) is indifferent to the distinction between referencer and referencee because, when ‘the semantics module’ is given ‘Odysseus’ or ‘Ithaca’ as an input, its role is not that of returning the object (if any) referred to by ‘Odysseus’ or ‘Ithaca’ as an output (for neither is in our heads) but only that of returning some representation of that object. So, for example suppose that (2) (i.e. ‘Odysseus is the father of Telemachus’) or (4) (i.e. ‘Ithaca is an island in the Aegean Sea’) were to be inputted in the semantics module, all the semantics module would need to do is to spit out as an output the corresponding sentences in what Azzouni calls ‘the speaker’s internal neurologically based language of thought’. In other words, we can roughly think of the semantics module as a translator that translates sentences from natural languages and, for example, the ‘language of thought’ and vice versa. According to Azzouni, all this translator needs to operate are rules such as:
We can think of the semantics module as assuming there are objects, properties, more general semantic values, and so on; we can think that those values are passed on to other modules. But this is a use/mention conflation: the value themselves aren’t passed on, regardless of whether or not there are such. Only a mechanism of co-identifying(r/e) the terms in the representations of different modules can actually occur. (It’s a use/mention error to think that values – actual objects, say – are themselves moved from one module to another.) But in a context where we are trying to evaluate exactly what ontological presuppositions are needed to get the whole thing to work, speaking of the semantics module so assuming things is just inviting philosophical confusion. It’s better to focus directly on the representations of [the speaker’s internal neurologically based language of thought], and recognize the semantic module value-ascription system as operating via the manipulation of those representations, concomitantly with the stipulation of equivalencies and identifications between representations (in the semantics module), terms in those representations, and representations (and terms) in other modules. For any of this, referencee is every bit as good as referencer. (Azzouni 2010: 238, emphasis added)
‘Odysseus’ in English refers(r/e) to the same object referred(r/e) to by ‘(8vdfd)’ in the language of thought,
‘Ithaca’ in English refers(r/e) to the same object referred(r/e) to by ‘^(ervn8844$$)(^(%wecwe%))’ in the language of thought,
‘Odysseus is the father of Telemachus’ is true in English if and only if ‘((wef∼)(8vdfd))*((&(t799))(&(&(z94y4$)))%(#(eq4)*(Hf4wv8w)))’ is true in the language of thought.
‘Ithaca is an island in the Aegean Sea’ is true in English if and only if ‘(30478563)+(^(ervn8844$$)(^(%wecwe%))*(#(vrwion309sss)*(YdcesiTT))’ is true in the language of thought.2
As far as I can see, this proposal has two fundamental problems. The first problem is that it is not clear what it would mean for ‘Odysseus’ and ‘(8vdfd)’ to refer(r/e) to the same object since, on the one hand, we are told that ‘Odysseus’ fails to referr to anything and, on the other, we are told that referencee is not a relation and therefore, a fortiori it is not a relation both ‘Odysseus’ and ‘(8vdfd)’ can bear to the same object. So, if neither ‘Odysseus’ nor ‘(8vdfd)’ refersr to anything and if reference is not a relation, it is not clear what it means to say that those two terms co-refer(r/e). Moreover, even assuming that there were a satisfactory account of what it means for those two terms to co-refer(r/e), it would be odd if the same account were also to apply to ‘Ithaca’ and ‘^(ervn8844$$)(^(%wecwe%))’, for one would expect that, if ‘Ithaca’ and ‘^(ervn8844$$)(^(%wecwe%))’ are in fact co-referring terms, ultimately it is because they both referr to the same real-world entity (i.e. the island of Ithaca). So, it is difficult to see how the proposal under consideration would dispel the suspicion that the notions of referencer and referencee have little or nothing in common.
The second problem is that, even if we could make sense of the notion that ‘Odysseus’ and ‘(8vdfd)’ co-refer(r/e) without referringr to anything and despite the fact that referencee is not a relation, principles such as (7)–(10) seem to just pass the semantic buck around and, if it turns out that the buck stops nowhere (if at no point we reach principles that are usually represented as ‘ “Ithaca” in English refers to Ithaca’ or ‘ “Ithaca is an island in the Aegean Sea” is true in English if and only if Ithaca is an island in the Aegean Sea’)), then we would seem to have failed to establish any connection between language or thought, on one side, and the world, on the other.
To use the classic example, whether or not my utterance of the sentence is true depends on whether or not ‘Hesperus’ and ‘Phosphorus’ (as used by me and my linguistic community) refer(r) to the same celestial body and not on whether or not each of them is associated with other mental representations of that celestial body I might have (and my ‘semantics module’ might associate with it). To rehash a well-known point, if unbeknownst to all of us, all the members of my linguistic community including me use ‘Hesperus’ and ‘Phosphorus’ to refer to the same celestial body, then my utterance of (11) would be true even if I and all other members of my linguistic community wrongly believe it to be false (and even if all of my mental representations of Hesperus differ from each of my mental representations of Phosphorus and, according to my ‘semantics module’, ‘Hesperus’ and ‘Phosphorus’ fail to ‘co-refer(r/e)’).
Hesperus is Phosphorus
One may worry that my argument from depth illegitimately relies on the assumption that the ordinary use of terms such as ‘about’ and ‘refers’ coincides with Azzouni’s use of ‘aboutr’ and ‘refersr’, while Azzouni himself would seem to think that the ordinary use coincides with the way he uses ‘about(r/e)’ (which, being ambiguous between ‘aboutr’ and ‘aboute’, is ‘ontologically neutral’). Now, the question of whether the ordinary use of ‘about’ coincides with Azzouni’s use of ‘aboutr’ (as I suspect) or with his use of ‘about(r/e)’ (as Azzouni seems to think) would seem to be by and large an empirical question, but, as far as I can see, the evidence does not seem to favour Azzouni’s interpretation. On the one hand, evidence that might superficially seem to favour Azzouni’s interpretation appears to be neutral when examined more closely. For example, the fact that ordinary English speakers are willing to assert sentences such as ‘We were talking about Mickey Mouse’ (while denying that there are any talking mice) seems to neither confirm nor disconfirm the interpretation of ‘about’ as ‘about(r/e)’, because it seems to be perfectly compatible with the interpretation of ‘about’ as ‘aboutr’ (conjoined with some form of inflationism or deflationism about fictional discourse). On the other hand, there seem to be reasons to be suspicious of interpreting ‘about’ as ‘about(r/e)’ (as opposed to ‘aboutr’). If the ordinary use of ‘about’ did in fact coincide with Azzouni’s use of ‘about(r/e)’, why would the fiction dilemma have any grip on us in the first place? And, if it were in fact the case that the ordinary use of ‘about’ is already ontologically neutral, why would we need to introduce the notions of ‘aboutr’ and ‘aboute’ in order to free ourselves from the grip of the fiction dilemma? (That is, why wouldn’t it be sufficient to attend to the ordinary meaning/use of ‘about’? And why would anyone need to write a book to convince us that that’s the case?)
I do not necessarily deny that there might be good answers to these questions. For example, one might argue that the fiction puzzle is an artefact of philosophy and that, as part of our philosophical training, we get to unlearn the ordinary meaning of words such as ‘about’ and we start using them in technical ways. Although I would not necessarily deny that some philosophical puzzles might be artefacts of the specialized way philosophers use ordinary terms, the fiction puzzle does not seem to be one of them, for there seems to be a genuine tension between acknowledging that there is no jolly old man who delivers presents on a sleigh pulled by flying reindeer on Christmas Eve and wanting to say that Santa has a white beard (and not a black handlebar moustache), a tension that cannot be resolved as easily as Azzouni seems to think.
3. In defence of pretence
After considering Azzouni’s proposal concerning fictional discourse, I will now turn to his key argument against pretence theories (which, I should note, is crucial to Azzouni’s case for his own view). Before doing so, however, let me sketch very roughly the multi-level version of pretence theory of fictional discourse that I favour. According to multi-level pretence theory, consumers and producers of fiction engage in different levels of pretence. Here, I will distinguish two levels of pretence – the fictional pretence and the meta-fictional one. At the meta-fictional level, readers of the Sherlock Holmes stories pretend that, by writing A Study in Scarlet, Conan Doyle created something (i.e. a fictional character). They also pretend that the name ‘Sherlock Holmes’ refers to that character and that the character appears along with other characters in a number of novels and short stories written by Conan Doyle. Finally, they also pretend that this character is depicted in a certain way in those stories and that the character also appears in a number of other works of fiction by other authors (although the depiction of the character in those other works is not always entirely consistent with his depiction in the original stories). Within this first-order meta-fictional pretence, readers of the Sherlock Holmes stories engage in a second-order pretence, the fictional pretence. At the fictional level, they pretend that the character created by Conan Doyle and referred to by ‘Sherlock Holmes’ is, in fact, a brilliant detective who lives in London, smokes a pipe, plays the violin and does all the things he is depicted as doing in the Conan Doyle stories.3
With this rough sketch in hand, we are now better equipped to look at to Azzouni’s argument against pretence theories and in particular to its key premise, which Azzouni calls the External Discourse Demand (EDD). In the following passage, Azzouni provides us with a clear and concise characterization of EDD:
According to Azzouni (2010: 119), for example, the (fiction-external) statement may contribute to the confirmation of the empirical hypothesis But why is EDD supposed to constitute a problem for pretence theorists? Azzouni’s argument can be schematically reconstructed as follows: As far as I can see, the problem with this argument is that, although (iii) is true and (ii) seems plausible, (i) (i.e. EDD itself) would seem to be false. Even if (12) were true, it would not contribute in any way to the confirmation of (13). Let me explain why. Suppose that, in the 1930s, American movie producers started producing predominantly movies that engaged their viewers in the pretence that their characters lived in New York and that, as part of this trend, someone produced a cinematic adaptation of The Hound of the Baskervilles entitled The Hound of Brooklyn Heights that engaged its viewers in the pretence that Sherlock Holmes lived in New York City (not London). If in the 1930s New Yorkers watched more movies than any other geographic group in the USA and if they were more likely to go to watch a movie if (the movie engaged them in the pretence that) its main characters lived in New York, then a historian of cinema may take this trend as providing evidence in favour of the empirical hypothesis (13), but, on this version of the story, the alleged truth of (12) would play no role whatsoever in confirming (13). If anything, what would contribute to confirming (13) would be sentences such as or A pretence theorists, however, has no problem with either (14) or (15) being literally true.
Statements like ‘Hamlet appears in Hamlet and Rosenkrantz and Guildenstern Are Dead’, ‘Emma Woodhouse was created by Jane Austen’, and the many statements like them, state facts that empirically confirm statements that aren’t about fictions in any sense. That is, fiction-external statements, when coupled with other statements that aren’t about fictions, will imply (for example) biographical and psychological facts about the creators of those fictions. The popularity (or unpopularity) of such fictional objects, how depictions of them evolve over time, what other uses they are put to by subsequent writers and readers, and so on, will empirically confirm sociological generalizations of various sorts. These statements – the ones that are empirically confirmed and that aren’t about fictions – are truths and falsities simpliciter, both in the sense that such statements are ordinarily asserted as statements we believe and in the sense that they, in turn, are used both in further inferences to and (as confirming or disconfirming) evidence for still other statements. (Azzouni 2010: 118)
Sherlock Holmes is depicted in The Hound of the Baskervilles as living in London; but he is depicted in the movie The Hound of Brooklyn Heights as living in New York4
American film producers in the 1930s were sensitive to the demographic fact that New Yorkers watched more movies than any other geographic group in the USA.
Sentences such as (12) can contribute to confirming sentences such as (13) (this is what Azzouni calls ‘EDD’),
Sentences such as (12) can contribute to confirming sentences such as (13) only if they can be literally true,
If the pretence theory is true, then sentences such as (12) cannot be literally true.
∴ The pretence theory is not true.
American movie producers in the 1930s produced predominantly movies that engaged their viewers in the pretence that the movie’s characters lived in New York City,
Some American movie producer in the 1930s produced a movie, The Hound of Brooklyn Heights, which engaged its viewers in the pretence that Sherlock Holmes lived in New York City.
The point can be seen even more clearly by considering a somewhat analogous case. Suppose that a psychiatrist is evaluating whether Henry, a child who has displayed some signs of antisocial and violent behaviour, should be diagnosed with Anti-Social Personality Disorder (ASPD). At some point, the psychiatrist realizes that Henry, who is intent in removing the heads from some dolls, is in fact pretending to be beheading some of his classmates. When the psychiatrist asks Henry what he is doing, the child answers that he is killing his classmates. However, upon further questioning, it is clear that Henry does not actually believe that the proposition expressed by is true; he is only pretending that it is true. Nevertheless, the psychiatrist may take the fact that the child engages in that sort of pretence as evidence that confirms the empirical hypothesis What is crucial to note is that what provides empirical support for the hypothesis that Henry has ASPD is not (16), which expresses the content of Henry’s pretence and which, by anyone’s admission, is clearly false, but As far as I can see, this case is analogous to the case Azzouni considers. In both cases, the fact that an individual or a group of individuals engage in a pretence with a certain content contributes to confirming a certain empirical hypothesis, however, in both cases, what contributes to confirming the empirical hypothesis is not the proposition that forms the content of the pretence – the proposition that such and such is the case – but the proposition that the individuals in question pretend that such and such is the case.
Henry is beheading his classmates
Henry has ASPD.
Henry is pretending that he is beheading some of his classmates.
In this section, I have argued that Azzouni has failed to establish the truth of EDD. Since EDD represents a cornerstone in Azzouni’s case against deflationary approaches to fictional discourse and for his own proposal, my arguments, if sound, would seem to seriously undermine his case.
4. Supervenience rescued
In this last section, I will consider Azzouni’s ‘blitz’ argument against supervenience approaches to inter-level metaphysics in Chapter 4 of TAN. If I understand it correctly, Azzouni’s argument is roughly that:
(iv) Sentences such as (12) can be literally true (from (i) and (ii) above),
(v) If sentences such as (12) are true, their truth fails to supervene on ‘the regularities in the underlying sciences (e.g. psychology, and ultimately biology and physics)’ (TAN: 186),
(vi) If the truth of sentences such as (12) fails to supervene on the regularities in the underlying sciences, then the relations between different theories cannot be explained in terms of supervenience.
∴ The relations between different theories cannot be explained in terms of supervenience.
I have already explained in the previous section why I think Azzouni has failed to establish (iv). However, I think that, even if (iv) were true, the above argument would still fail to be sound because (v) would seem to be false by Azzouni’s own lights. According to Azzouni, (12) is true and it is true in virtue of there being TVIs that make it true, where, following Azzouni the TVIs for (12) should be characterized as ‘a blend of relevant objects that exist – us, our language and our epistemic practices included – and that jointly yield the indispensability of the truth [of (12)] to our assertoric practices in ordinary life and our sciences’ (Azzouni 2010, 18). Although I am not entirely sure I understand what the TVIs for (12) would be (other than the kinds of pretence practices pretence theorists postulate), it seems that two things are true of them. The first is that whatever they are, they are the kinds of things that fall within the scope of science (this would seem to be true of three elements of the blend Azzouni mentions – i.e. us, our language and our epistemic practices); the second is that the truth of (12) supervenes on the existence of the objects that are its TVIs (any world at which the TVIs for (12) exist is a world at which (12) is true). But, if both of these assumptions about the TVIs for (12) are true, then (v) would seem to be false as the truth of (12) would supervene on facts that fall within the scope of the underlying sciences.
In TAN, Azzouni develops and defends a novel, unified approach to our thinking and talking about mathematics, fiction and hallucinations. Although I found Azzouni’s proposal very interesting, in this paper, I have tried to explain why, ultimately, I was not convinced.5