Correction to: Linguistic convention and worldly fact

1 Correction to: Philos Stud https://doi.org/10.1007/s11098-018-1088-5

The original publication of the article contains two formatting errors, the second of which significantly inhibits readability.

First, in the fourth paragraph of Sect. 4, there’s a quotation in which parentheses appear where there should have been square brackets, resulting in a misquote. The sentence containing this quotation should read as follows:

But Warren (2015b, 91) argues that conventionalists have reason to think this standard view gets the order of explanation backward: conventionalist views “are most naturally launched against the background of meta-semantic theories that make dispositions to use sentences central and take any explanatory work done by propositions [and, presumably, by worldly (i.e., nonlinguistic) facts] to be less fundamental”.

Second, the three theses introduced in the ninth paragraph of Sect. 4 are indexed using Arabic numerals that are already in use as indices for previously introduced theses. As a result, the ensuing discussion of the relationships between these new theses and the previously introduced ones is difficult to follow, since it’s often not clear what thesis is being referred to. To avoid this problem, the new theses should have been indexed using Roman numerals. The four affected paragraphs, then, should read as follows:

The question, then, is what sort of story, if any, remains available. And the answer, I claim, is that there’s no story for conventionalists to tell here. After all, if they deny that the fact that all vixens are foxes—i.e., v—plays a role in explaining the truth of S, only three possibilities remain: (i) the truth of S fully explains v; (ii) the truth of S plays a role in explaining but doesn’t fully explain v; or (iii) neither plays a role in explaining the other. And conventionalists, it turns out, can’t accept any of these. I consider them in turn.

In a certain sense, (i) is the most promising—it ensures that there’s a tight explanatory relationship between v and the truth of S, a relationship that can explain how our access to the latter might give us access to the former. The problem with (i), though, is straightforward: it doesn’t actually give conventionalists a way to avoid (C). And this is independent of epistemological considerations. To embrace the claim that the truth of S fully explains the fact that all vixens are foxes, after all, is just to replace (3) with the following:

(3′) All vixens are foxes purely in virtue of S’s being true

And the argument for (C) remains valid when (3) is replaced with (3′). In fact, (1) and (4) aren’t even needed in this new version of the argument: (C) follows just from (2) and (3′). By (3′), v obtains purely in virtue of the truth of S, which means that, if S itself were true purely in virtue of facts about convention, v would thereby also obtain purely in virtue of facts about convention.¹ But (2) tells us that v doesn’t obtain purely in virtue of facts about convention. By (2) and (3′), then, S can’t be true purely in virtue of facts about convention, which is to say that (C) must be true.

So it’s easy to show that, if conventionalists embrace (i), the objection from worldly fact loses none of its force. Showing why conventionalists can’t embrace either (ii) or (iii), though, isn’t so straightforward—neither of them leads to outright inconsistency with conventionalism. The problem is instead an epistemological one: neither (ii) nor (iii) can give conventionalists the resources to explain, in a naturalist-friendly way, our near-perfect reliability about facts like the fact that all vixens are foxes. And this is just to say that any conventionalist view that embraces either claim will thereby fail to do the epistemological work conventionalism is intended to do.

We can see why this is by noting that, if either claim is correct, then conventionalists have no way of explaining why S’s being true by convention is sufficient to guarantee that v obtains. Let’s suppose that either (ii) or (iii) is true: either the truth of S plays some role in explaining v but doesn’t fully explain it, or neither plays any role in explaining the other. In either case, there are some facts that play a role in explaining v but that aren’t explained by the truth of S.² Furthermore, at least one of these facts—call it f—can’t either be a fact about convention or obtain purely in virtue of facts about convention (nor can its role in explaining v be derivative of its role in explaining any fact about convention), since (2) rules out the possibility that v obtains purely in virtue of convention. So, if conventionalists are to explain our near-perfect reliability about facts like v (and thereby meet the Object-Level Belief constraint), they must explain how we can be sure, in cases like this, that some suitable f obtains. Otherwise, they won’t be able to explain our ability to rule out the following possibility: that v fails to obtain as a result of there being no suitable f available. But conventionalism just doesn’t give us the resources to explain, in any naturalist-friendly way, how we have access to any such f.³ And that means that conventionalists who go in for either (ii) or (iii) will thereby be unable to meet the Naturalist-Friendliness constraint.