Defending Explosive Universal Fictions

We think FMP is false, as not every fiction is closed under conditional elimination.

If you achieve both steps, then, because (2) is a conditional and (1) is its antecedent, f's closure under FMP-LOCAL_f ensures that the consequent is also true-in-f. And since the consequent is that every proposition is true, it follows that f is a universal fiction.

Ricksand (2020) raises three objections to our proposal. Here, we take the opportunity to reply to these concerns, thereby clarifying and expanding on our argument.

Before turning to Ricksand's objections, it is useful to discuss the background dialectic. Doing so will clarify our ecumenical approach and serve as a foundation for our replies.

There are (at least!) two substantive difficulties one faces when following our recipe. The first concerns ensuring that (1) and (2) are part of f's content. Addressing this requires saying something about the broader question of how to make a particular proposition true in a given fiction.

This is a hoary, difficult matter, to which there is no straightforward answer. One naïve idea is that saying makes it so; roughly, if some statement is explicitly made in a fiction (for example, by the fiction's narrator), then the expressed proposition is true in that fiction. Philosophers and literary theorists have roundly (and rightly) rejected this stipulatory account, as, for example, any fiction featuring an unreliable narrator is a counterexample. A second, related notion is intentionalism: if the (or an) author of fiction f intends that p is true-in-f, then p is true-in-f. This approach has also been largely rejected, running into numerous apparent counterexamples (see, for example, Lewis 1978, though see also, Stock 2017).

Another way of being part of f's content is to be imported, that is, a proposition brought into the fiction from the outside. However, what (if any) propositions should be imported is another controversial matter.³ Yet another way is to be implied; that is, if p is a logical consequence of some proposition that is true-in-f, then p is true-in-f. This is especially unhelpful, since not only does it move the bubble in the carpet (since it requires that we already know some of f's content), but it is not clear which notion of logical consequence we should employ.

We mention these to highlight that there is no good general story about how to guarantee that a proposition is true in particular fiction. This makes addressing the first issue extremely difficult, as it is hard to know whether one has succeeded in making (1) and (2) true in f.⁴

The second difficulty concerns ensuring that f is governed by FMP-LOCAL_f. As before, there is a lurking larger problem: namely, settling what (if any) principles constitute the logic of fictional truth. At least at first glance, not all fictions’ content obeys the same logical principles. For example, Priest's Sylvan's Box and Bradbury's A Sound of Thunder are inconsistent, yet their content is not governed by the principle of explosion (otherwise, they would be universal too). Similarly, essentially incomplete fictions like Blade Runner seem to violate the principle that, if (P or Q) is true-in-f, then P is true-in-f or Q is true-in-f.⁵ These issues have led some, such as Routley (1979, 10) to hold that there is no uniform logic of fictional truth.⁶ If Routley is correct, then proving that f's content obeys FMP-LOCAL_f becomes that much harder (and explains why we need to appeal to local, rather than global principles).

In light of these substantive disagreements, in our argument for universal fictions, we wanted to avoid committing to any specific view and make as few controversial assumptions as possible. So, to ensure that (1) and (2) are part of f's content, we suggested including explicit statements in f that expressed the relevant propositions. This is because, while we think that saying does not always make it so, it generally does. That is, if fiction f includes an explicit statement that expresses a proposition p, then, ceteris paribus, p is true-in-f. Consequently, including these explicit statements is a fairly uncontroversial way of getting (1) and (2) to be part of f's content.⁷ To stress, we do not think this is the only way of doing this, just the least controversial way.

Our strategy concerning the second issue was also ecumenical. Specifically, we suggested including an “innocuous” conditional and antecedent as part of f's content. Together, these strongly suggest (though do not strictly entail) that the relevant consequent is also true-in-f. In turn, this means that f's content is closed under FMP-LOCAL_f. And denying this closure would require denying the fictional truth of the consequent, which is “utterly implausible” (2017, 78). For example, in our Monsieur Impossible, we included that (i) if Monsieur Impossible is a member of the King's Musketeers, then he works for the King, and that (ii) Monsieur Impossible is a member of the Musketeers; the intuitive result is that (iii) Monsieur Impossible works for the King is also fictionally true. This strongly suggests that Monsieur Impossible is closed under FMP-LOCAL_MI, as denying this would seem to require denying the (extremely plausible) fictional truth of (iii). As before, we do not think this is the only way to guarantee that FMP-LOCAL_f is true of f, but is, we believe, a fairly uncontroversial method, compatible with a wide variety of views about the logic of fictional truth.

We would be the first to admit that neither of these solutions are indisputable (what philosophical arguments are?). Nor are these strategies the only way to resolve these two issues. But they are likely the best one can do without providing deep and controversial answers to the two lurking general questions outlined above. If one has a completely worked out story about the necessary and sufficient conditions for fictional truth as well as an account of the logic of fictional truth, then you could probably do better. Yet that was not our goal. We wanted to sketch a recipe for generating universal fictions that was as theory neutral as possible.

With this background discussion out of the way, we turn to Ricksand's objections.

In reply, we offered cases where the “innocuous” conditional and antecedent are inconsistent, though the consequent is suitably mundane. For example, our Clara's Crazy Caper (2017, 78) included the conditional (i) if exactly three and not exactly three carrots are consumed, then some carrots have been consumed, as well as the contradiction (ii) exactly three and not exactly three carrots are consumed. As before, it is strongly intuitive that, given this setup, this makes (iii) some carrots have been consumed is true-in-CCC. Thus, the same argument applies: either the objector grants that the consequence is part of the fiction's content, in which case they must accept that the content is governed by the unrestricted FMP-LOCAL_CCC, or they have to take on the “utterly implausible” consequence that (iii) is not true-in-CCC. At minimum, this places the burden of proof on the objector.

Here, Ricksand objects that, “it is not clear why this example… would pose a problem to an objector.” If the objection “consists of categorically denying that instances of FMP-LOCAL can be fictionally true when [the antecedent] is inconsistent, it hardly amounts to a counterargument to provide another example … where [the antecedent] is inconsistent, since this is the very kind of case the objector will not accept” (2020, 236).

In reply, first note that the issue is not whether an instance of FMP-LOCAL is fictionally true; what matters is whether it is true of, not in the relevant fiction. Rather, Ricksand's objector must “categorically deny” that any instance of the fictional truth of both a conditional and the relevant inconsistent antecedent entails the fictional truth of the consequent. This categorical denial looks extremely difficult to maintain. For example, suppose that (i) ((P&P) → ((P&P) & Q)), (ii) (P&P), and (iii) Q are all true-in-f. Denying that (iii) is true-in-f because (i) and (ii) feature a contradiction looks ridiculous. Yet, this is what Ricksand's objector is committed to. At minimum, this objector bears the burden of proof of explaining why we should accept this strongly counterintuitive result. And, until this is forthcoming, we have enough to warrant thinking that f is closed under FMP-LOCAL_f.

Ricksand's second objection concerns our discussion of an alternative universal fiction recipe.⁸ Per this alternative, one can produce a universal fiction f* by telling a story that explicitly includes some statement like, “everything is true.”⁹ This is meant to entail that every proposition is true-in-f*.

We have significant worries about this alternative recipe (2017, 74–75). We illustrated our worries via an analogy: that “everyone is treacherous” is true-in-Threepenny-Opera does not entail that, for example, “Obama is treacherous” is true-in-Threepenny. This is because, plausibly, the “everyone” quantifier only ranges over characters in the story, and not every individual is part of Threepenny's cast. Similarly, it is plausible that “everything is true” being true-in-f* does not entail that absolutely every proposition is true-in-f*. This is because, plausibly, the “everything” quantifier only ranges over those propositions that are in fact true-in-f*, and not every proposition is part of f*’s content. So, we think it is best to “sidestep [this] route and offer a different pathway to universal fictions” (2017, 75).

Ricksand's second objection is that our worries about the alternative recipe apply equally to our own. In brief, why think that the quantifier in the principle of explosion has a universal range, while the “everything” in these other stories is restricted?

As well as apparently undermining our recipe, Ricksand suggests that this demonstrates the “triviality” of FMP. This is because we “concede that [FMP does] not obtain with necessity in all fictions, and that it is only a local version of FMP which allows for the construction of a universal fiction, since the principles necessary for rendering a fiction universal must be presented explicitly in order to obtain. However, by conceding that no version of FMP obtains without explicit statements to that effect they also inadvertently undermine their own criticism” (Ricksand 2020, 236).

First, the issue is to have FMP-LOCAL be true of, not true in the relevant fiction—we want the content to be closed under the principle, not for the principle to be part of the content. Second, it is hard to see how FMP is “trivial” if it is true of some but not all fictions. Third, we do not concede that no version of FMP obtains without some explicit statement that it does so. Our argument for FMP-LOCAL_f involves no such a statement, and, frankly, we would not even know what sort of explicit claim one could make that would be of any use. “FMP-LOCAL_f governs f's content” being part of f's primary, explicit content certainly entails that “FMP-LOCAL_f governs f's content” is true-in-f. But, it does not entail that FMP-LOCAL_f in fact governs f's content, which is what matters here.¹⁰

Setting these aside, one reason for thinking the “Everything is true” quantifier is restricted, while our explosion quantifier is unrestricted is that there is no way to secure the relevant range for the former without appealing to something like authorial stipulation, a point that one advocate of this recipe readily admits (Deutsch 1985, 209n16). In contrast, the explosive quantifier is presumed to be absolutely unrestricted, because, when discussing explosion, logicians nearly universally accept that it is so.¹¹ So, we do not need to do anything to make it unrestricted—it comes ready-made that way—but the alternative recipe must rely upon stipulation. And given the substantive debate about the limitations of authorial stipulation, we think this is problematic. This is not to say that this other method must fail—indeed, we never said anything like that! Rather, we think it is better to avoid getting tangled in these “thorny issues” about authorial intent (2017, 75) in favor of our approach.

A second difference is that one can think of the principle of explosions as a derivation or inference rule (for example, that, from a true contradiction, one can derive/infer any proposition). This “generative” understanding fits with our recipe for generating universal fictions: assuming that the relevant fiction's content includes a contradiction and the principle of explosion, one can then use this combination to derive/infer every proposition within the fiction.¹² Further, it does not seem subject to the restricted quantification range worry. However, there is no way to understand the “Everything is true” claim nonquantificationally. Thus, any recipe that relies upon it is subject to the worry. Since our approach potentially avoids it, this looks like another point in its favor.

These indicate that there is a substantive difference between our recipe and the “everything is true” alternative, which suffices to undercut Ricksand's second objection.

First, a point of clarification: no one should accept that including in fiction f the statement, “q is fictionally true” is enough to make q true-in-f. That “q is fictionally true” is true-in-f means that, in f, q is true in some fiction, which may or may not be f. Meanwhile, q's being true-in-f means that, in f, q. And, obviously, the former does not entail the latter.¹³ Presumably, Ricksand meant that including in fiction f the statement that “q is true” is enough to make q true-in-f. So, we direct our reply to this suitably modified version of the dilemma.

We are happy to reject the first horn. We do not think that explicitly stating that q (in the relevant text) is sufficient to make q true-in-f. (Though it is worth stressing again that our recipe requires that FMP-LOCAL_f be true of and not in fiction f.)¹⁴

However, rejecting the idea that (explicitly) saying makes it so does not, contra Ricksand, entail that “some propositions cannot be fictionally true only by virtue of being explicitly stated” (2020, 237). It does entail that some attempts to make some proposition(s) fictionally true by means of explicit statement fail. But it does not entail that, for some particular proposition, every attempt to make it fictionally true via explicit inclusion must fail. This non-entailment is enough to undermine the dilemma's second horn. So, Ricksand's dilemma is no threat either.¹⁵

The general upshot is that Ricksand's objections do not threaten our recipe for universal fictions. Regardless, thinking about them was helpful for clarifying a number of points. For this reason, we thank him for engaging with our work. Finally, to conclude, we would like to mirror our previous ending: even if we have not convinced everyone that universal fictions are possible, we think that highlighting what moves must be made to accept or reject this possibility is interesting in its own right.