Blow My Mind

Here's something I'll be thinking about during my sober moments over the next two weeks.

In seminar the other day, one of my philosophy professors (you might be able to guess who) gave an awfully convincing argument about why existence might not exist; more precisely, why the statement "There exists an X" has indeterminate truth value. I'm going to do my best to reconstruct the argument, and then try to offer some objections/clarifications.

The rough sketch is this: The Quinian argument against the validity of the analytic/synthetic distinction in "Two Dogmas of Empiricism" doesn't work unless one presumes verificationism, and since verificationism is a profound failure, the distinction remains valid. However, Quine does point us towards a weaker conclusion, namely that the analytic/synthetic distinction is vague, and there are borderline cases in which it is unclear if the truth of a statement is a function of its meaning or of facts about the world.

Vagueness in the analytic/synthetic distinction is a linguistic indeterminacy. But it might be the case that in using the quantifier to make existence statements, (that backwards capital "E" before it became Eminem's logo, i.e. "There exists...."), the indeterminacy gets pushed out into ontology.

How so? To make sense of the quantifier, we need a conceptual analysis of the folk concept of "existence." Like other folk concepts, this one is the composed of various folk axioms. For each axiom of existence, we can ask whether it is an analytic feature of the macroscopic concept, or a synthetic feature. If the latter, then evidence that tells against the axiom might cause us to jettison the axiom from our concept, but the concept will be preserved. If the former, then to attempt to refer to the concept without referring to the axiom is to change the subject.

Example: It's somewhat uncontroversial that unmarriedness is an analytic component of the concept of "bachelor." If in conversation, someone says to you, "Jones is a bachelor, but he is married," you are within your rights to tell him that whatever he meant in using the term "bachelor," he was not referring to bachelors. Conversely, suppose you and the same interlocutor observe a cat sitting on a mat, and you say to him, "Lo, that cat is on that mat." He responds, "No, the cat is not on that mat." Here you and your friend have a substantive disagreement. The property of being on a mat is in no sense true of cats by virtue of the meaning of "cat."

Now let's look at the folk concept of existence. There are certainly other axioms associated with this concept, but for the purposes of this argument, we only need to look at two. They are:

1) The accomodationist principle: When we look at form/matter pairs with which we are unfamiliar, we tend not to obstinately insist that the pair is not a thing. We grant it ontological reality, if only for the sake of courtesy.

Example: Ernest Sosa invented the concept of the "snowdiscall" to motivate his own theory of conceptual relativity, which is tangential to the issues in this argument. The snowdiscall is a mass of snow with any shape ranging from perfectly spherical to perfectly disc-shaped, including every shape in between. Snowballs are all also snowdiscalls, but some snowdiscalls are not snowballs. Our culture happens to be a snowball culture; if someone presents us with a disc-shaped mass of snow, we would affirm that it is not a snowball. However, suppose one of us travels to a far-off land in which there is no snowball concept, only snowdiscalls. Would any of us be so stubborn as to deny that the snowdiscall exists? There are some philophers who deny the reality of everything but the simples; for them, snowballs and snowdiscalls are equally unreal. For the rest of us who think that composites do exist, to affirm the existence of snowballs but to deny that of snowdiscalls would be rather obtuse ontological chauvinism. We certainly do not have a concept-term for every possible form/matter pair, but we accomodate upon discovering new ones.

2) The reductionist principle: By the lights of Occam's Razor, we seek to clean our ontology of explanatory redundancies. Composites are included as part of the ontology just in case they provide an explanation of data that their component parts could not provide. Otherwise, we say that while it might be efficacious to speak about certain composites, they are not actually "real."

Example: Neither snowballs nor snowdiscalls explain anything in the world that their constituent particles could not explain. We are inclined to say that while there are contexts in which it is useful or convenient to refer to "snowballs" and "snowdiscalls," what really exists is matter in a region of space-time with a particular configuration. Conversely, organisms might be said to really exist, because a composite organism has consciousness, and thus by Leibniz's law is not identical to its constituent parts. Moreover, certain facts about the world are explained by the existence of organisms that cannot be explained by reference to a configuration of organic matter.

Those are the folk axioms of existence that motivate this argument. I think it's clear why they are in conflict with one another, but I'll try to make it explicit. The conflict is the oldest conflict in ontology: What is. The accomodationist principle (1) drives us to affirm an ontology of plenitude, the reductionist principle (2) drives us to affirm a barren ontology. The two principles motivate disparate and incommensurable notions of existence. This tension in ontology spills over into language; two utterances of the statement, "There exists an x" might turn out to have very different meanings despite being composed of identical words, because a quantification using the first notion of existence is going to have different truth conditions than a quantification using the second notion of existence.

If ontological controversy can generate linguistic controversy, might the opposite not occur? If the argument I have framed is sound to this point, then such an outcome is unavoidable.

Let's turn back to principles (1) and (2). The looming question all along has been, are these analytic components of the concept of existence, or synthetic components? Everyone is going to have slightly different intuitions about the answer to that question, as well they should; one way to understand the competing ontological theories in contemporary metaphysics is as ways of answering the question. I have my own intuition, too, which I'll get into in a bit, but I wouldn't deign to suggest that my answer is a manifestly obvious one. In fact, all that I would say is certain about this is that no answer is clearly the right one. There are good reasons for believing (1) is analytic and good reasons for believing it is synthetic; likewise with (2). We have then, a problem of vagueness, which as philosophy students know, is one of the absolute thorniest problems in philosophy.

One way to get a purchase on this vagueness is through supervaluationism; by that I don't mean that supervaluationism is necessarily the correct solution to the problem of vagueness, but rather that its methodology might be enlightening in this case. Supervaluationism is the process of examining all possible precissifications of a vagueness in order to arrive at super-truths. Take the concept of baldness. It's thoroughly vague as to where to draw the boundary between bald and not-bald, but there are some cases, e.g. complete hairlessness and complete hairiness that are bald or not-bald in all possible worlds, i.e. super-bald or super-not-bald.

Supervaluationism, I am willing to concede, is inadequate in the Sorites-like examples to which it is usually applied. Finding out the completely bald and completely hairy are perhaps the only super-bald/not-bald cases while all others are indeterminately bald isn't really all that edifying. (Supervaluationism does preserve the excluded middle, however; for every contentious baldness case, it is either bald or not-bald under all possible precissifications. Hence, p is indeterminate, ~p is indeterminate, but (p or ~p) is super-true. This is why the theory was appealing in the first place.) But supervaluationism can be a very powerful tool for approaching a vagueness, like the one in this argument, that has a finite set of possible valuations.

And here's how it applies:

Precissification 1: The accomodationist principle (A) and the reductionist principle (R) are both synthetic features of the concept of existence. This precissification is where fruitful ontological debate can take place, because both accomodationists and eliminativists/reductionists would be arguing about the same subject matter, not just talking past each other.

Precissification 2: A is analytic, R is therefore synthetic and false. What follows is that complexes do really exist.

Precissification 3: R is analytic, A is synthetic and false. Therefore only the simples and explanitorily significant complexes exist.

Precissification 4: A and R are both analytic. Any "There exists" statement is therefore logically equivalent to statements about round squares. Nothing exists.

Our use of the quantifier, therefore, is indeterminate between these precissifications of the concept of existence. What supervaluation has achieved, however, is preservation of an excluded middle: under all precissifications, X exists or doesn't exist (where X is an alleged existence). Problem: "X exists" is indeterminate. "X does not exist" is indeterminate. We've pushed a linguistic indeterminacy into the world, and now we have a third track in ontology. The universe that we have taken to be utterly independent of our own conceptual schemes looks like, at bottom, an arrangement of Carnapian bundles of language games.

This is not a conclusion I'm happy with. I'd like to figure out how to avoid it. And I have two thoughts about how and why that might be done. First, I'm not entirely convinced that the indeterminacy in the analytic/synthetic distinction really can be pushed into ontology. What we might say is that "existence" has nothing to do with existence, and all that we have done in the foregoing argument is trace out the limits of "existence." But then we're in a bifurcated Kantian realm of phenomena and noumena. I'm also not thrilled by that possibility.

So one other proposal, perhaps a bit more careful. What drives us to conclude that the quantifier itself produces statements of indeterminate truth value is a sense of intractable vagueness regarding the analyticity or syntheticity of the folk axioms of existence. I admit that it is difficult to resolve, but need this vagueness be intractable? The solution---well, my thought about a solution---is apply the model of contextualism, which David Lewis proposed as a remedy for the problem of skepticism in epistemology. I'd need to flesh this proposal out a bit more, but it would go something like this:

Just as the epistemological debate, in which we're willing to bracket talk about knowledge in favor of talk about "knowledge," we can envision an ontological contextualism premised on bracketing talk of existence in favor of "existence." We certainly seem to have accepted that assumption in the course of the argument above. Now that it's clear we're referring to "existence," this much seems at least prima facie plausible: Just as in epistemological contextualism, in which we distinguish the ordinary context in which knowledge statements are true from the philosophical/skeptical context in which knowledge is destroyed, we might say that the ordinary context for existence-statements is an accomodationist context, whereas the philosophical context is reductionist. By fixing the context of existence statements, we can thus avoid any indeterminacy in making them; we will have, in effect, adopted (for non-arbitrary reasons) a particular precissification of the concept of "existence." The objection I can already hear is that we have not in doing this salvaged determinacy for the unrestricted quantifier. All I can think of in reply at the moment is to question the assumption that we ever actually use an unrestricted quantifier. Another, deeper objection is that the philosophical context need not be reductionist and the ordinary context accomodationist; in fact, it could be quite the opposite. In ordinary speech, we might walk into a room and say, "Lo, this room is empty." Someone overhearing us, himself in the philosophical context, might reply, "No, there are particles, complexes, and mereological sums everywhere."

Okay, all done. For now.

[N.B. Maybe it's time to get typepad--ed.]