Let me begin the semantic argument by presenting some familiar instances where we use semantics to constrain ontology. A famous case is Davidson’s event semantics (Davidson, 1967). Recall the basic line of thought here. Consider the following infer- ence:
• John buttered the toast slowly, so John buttered the toast.
This seems to be an inference good in virtue of its form. But if we were to try and portray its form using the standard sort of translations into first order logic, we’d have to formalise ’buttered...slowly’ as an atomic predicate, yielding:
• Buttered-slowly(John), so Buttered(John)
And that’s not an inference good in virtue of its form. However, Davidson noticed that if instead we were to interpret the sentence as a quantifier over events, we could make sense of it. For then we could have something like:
• ∃e. Agent(John, e) & Action(buttering, e) & Way(slowly,e), therefore∃e. Agent(John, e) & Action(buttering, e)
And this inference isgood in virtue of form, in just the same way the inference from "Someone is happy and tall" to "Someone is happy" is. So if we assume the existence of events, in addition to individuals, then we can make sense of the inference, and that supplies us with a (defeasible) reason to believe in events.
Here’s another example. Elbourne, leaning on work by Heim (both cited above), argues that we can best make sense of donkey anaphora if we introduce the idea of
situations. One can think of a situation as a part of a world. For example, part of the world is what’s going on with me: it’s a situation consisting of a man typing, with a grey Norwegian day outside the window.
Situations can help us with anaphora as follows. Without getting into the details (some of which we saw earlier, in the chapter on names) too much, we can note that there’s some independent reason for thinking a sentence like:
(117) If a farmer owns a donkey, he beats it
Has an underlying syntactic logical form something like: (118) If a farmer owns a donkey, the farmer beats the donkey.
Given this, we can make sense of what’s going on if we say that the if-clause functions as a quantifier over situations, and in particular, it quantifies over situations in which there’s a farmer who owns a donkey. The sentence is true provided every such situation can be extended to one in which the consequent holds. So, in this example, we take each antecedent situation, that is each situation consisting of a farmer who owns a donkey, and try to extend to one in which the farmer in the situation beats the donkey in the situation. If we can, then the sentence is true. The advantage of using situations is that they’re small enough that we can secure uniqueness for the definites in the consequent relative to each situation we quantify over, something we couldn’t get with worlds. Again, it seems reasonable to say that the fact that this semantics works, if it works, provides us with a defeasible reason to believe in the existence of situations.
They key thing to note here is that these sorts of arguments seem relatively un- controversial and an established way of doing semantics. Provided a particular meta- physics explains some data better than another, and that metaphysics is not otherwise outlandish (for example: it’s quite plausible that we see events and situations, talk about them using nominalisations, evoke them in physics, and so on), we should feel free to adopt it. Of course, what it means to ’adopt’ it is a massive question, which could be taken in heavier or lighter senses. As I said above, for this argument I’m happy to accept the lightest possible sense: whatever you think semanticists are doing when they do this, that’s what I’m doing.
I think to some extent saying that should be enough for the purposes of this thesis. If there is some data best explained by the positing of a certain metaphysics, then one has reason to adopt that metaphysics. I’ll later argue that assuming the existence of temporal parts, one particular metaphysics of them best captures certain data.
However, the reader might have qualms. For she might think that I’ve already let metaphysics intrude too much even by thinking that a semantics should talk about temporal parts, even granting they exist. Should not semantics be neutral?
Sider, in the passage quoted in the introduction, certainly seems to think so. He continues to say:
The advantage of this approach is that it allows linguists..to be guided by concerns internal to linguistics (Sider, 2011,p123)
One might well agree that the existence or not of temporal parts is not a concern inter- nal to linguistics. The next argument purports to show that it is. The reason for this is that semantic externalism is a concern internal to linguistics and if semantic external- ism is true, then if stage theory is true, referring expressions stand for stages, while if endurantism is true, they stand for enduring objects. And this second conditional has
consequences for formal semantics. The goal of the next section is to argue for the truth of the whole conditional. However, prior to doing so, I need to review some (familiar) material concerning externalism.