by Tim Sommers
“The world was so recent that many things lacked names, and in order to indicate them it was necessary to point.” –Gabriel Garcia Marquez in One Hundred Years of Solitude
Here’s an apocryphal story that is so good, it should be true. In 1770 Captain Cook became the first European to land in Australia and so the first to encounter a leaping animal with a baby in its pouch. He pointed at it and asked a nearby aboriginal, “What is that?”
“Kan-ga-roo,” our fictional indigenous person responded. And, so, we call them “kangaroos”. Which means, in that local language, “I don’t know what you are saying.”
I always associate that story with Willard Van Orman Quine’s essay, “Ontological Relativity”, because the centerpiece of the essay is what he calls a “radical translation” scenario. We are trying to learn to translate a language that we have no independent knowledge of, a language that, as far as we know, is not related to any other language that we already know how to translate. Suppose a rabbit goes hopping by and a native speaker points at it saying, “Gavagi!” Even assuming we all agree on what pointing is, and we all think that pointing and talking at the same time associates the pointing with the talking, and we are sure that the spatio-temporal area occupied by what we would call a rabbit is what is being pointed at, assuming all that, how do we know whether they mean “There goes a rabbit!” or “Look at those undetached-rabbit-parts!” or “Some rabbitizing is going on over there”. Or, if you do the pointing, how do you know that what they are saying to you is not just, “I don’t know what you are saying”.
I know, I know, only a philosopher would wonder that. But, consider: what we think exists should be revealed by what we say, but what if it’s not? What if what exists is relative to what we say, but we can never be absolutely sure what anyone is saying about what there is? What if ontology, the part of metaphysics that is supposed to be, at a minimum, a catalogue of things that we think exist and are real, is relative to language and language is always indeterminate with respect to ontology. Should we be worried about this?
“Ontological Relativity” turns 50 this year. It is one of the most widely-read pieces in twentieth century philosophy by, perhaps, the most influential American philosopher of the second half of that century. Quine shared with the spirit of his age a naturalism that put logic and science first – and philosophy second. But Quine saw science as an extension of common sense. And he thought there was no bright line between science and philosophy properly done. He thought that logic was important, but not foundational – a tool, not the first, or last, word. He believed that all of our beliefs form a sort of web (“the web of belief”) that makes contact with the world at many places along the periphery, but that is always provisional, revisable so that even our most recalcitrant beliefs (like those associated with math and logic), even beliefs at the very center of the web, are still open to revision. But in “Ontological Relativity” he found a new skeptical worry that not only challenged more traditional views of philosophy, but one that he could never quite figure out to fit with his own views. The argument is challenging and slippery. But fun, too, I hope.
Let’s start over. What’s at issue is what exists. There’s a part of metaphysics – ontology – that deals with what exists. Quine says that the question of ontology is the question of what exists. In theory, then, an ontology could just be a long list of all the things that exist – though we would normally think it had some kind of structure. In fact, for thousands of years before Quine, many philosopher’s thought of ontology as the study not of what exists, but of what must exist – what exists necessarily. What philosophers thought must exist necessarily had slowly been reduced over the years from almost everything, to space and time and God and logic, until, by the time we get to Quine, many philosophers would have said, the only thing that exists necessarily are the laws of logic. (Here’s a cartoonishly simple, but not entirely unhelpful way of seeing the whole history of philosophy: it started as the study of things, became the study of ideas, and, finally, became (in the twentieth-century) the study of language (and logic). But Quine argued that nothing was necessary – not even the laws of logic. And whatever exists is just whatever has something that is true of it. To put it the logical terms that Quine preferred, “To be is to be the value of (a bound) variable”. In ontological relativity, Quine went one more step. We don’t even know what we are saying exists, or what we think exists, at least not absolutely.
Here’s the argument one step at a time.
Language is not a system of labels. Quine calls this “the myth of the museum”. According to the myth, language is a mental museum of exhibits, all of the exhibits are labeled, and switching languages just involves switching the labels around. What’s wrong with this picture?
Well, language is a public, social. There’s no such thing as a private, untranslatable language. (This is Wittgenstein.) Suppose everyone had a box with a “beetle” in it, but no one could ever, no matter what, look inside another person’s box – only their own. For all I know, I have a turtle in my box and you have a rock and we both just call them “beetles”. Maybe, I have nothing in my box and whatever’s in your box changes all the time. What’s in the box can play no real role in language. For the same reason, and in the same way, what’s in your head or your mind, the exhibits in the museum, cannot be what meanings are. When it comes to language, all we have to go on in assigning meaning or reference (the things your words pick out) is your behavior.
Now, the overall system of labels or words, like the objects posited by any theory, is always “underdetermined”. For any finite set of observations or experiments, there are always many “empirically equivalent” rival theories. For any finite set of language behavior, any finite set of words or labels or terms of a theory, there’s always a way to rearrange them without changing their reference. (See, rabbits vs. rabbit parts above.) And there are ways of changing around bits of the language to come up rival translations that do equally well as interpretations of observed behavior.
And, here’s where it gets really crazy, you can’t ever just fix the terms by brute assignment. I can’t ever pin down exactly what you mean by what you say, because the only way I can try is by using the language that we share. So, language is “indeterminate” all the way down. I can’t pin down what you mean by what you say and I can’t even pin down what I mean by what I say.
Why can’t you tell me what you mean by just pointing and saying by this, I mean that? Well, again, see rabbits vs. rabbit parts above. Okay, but why can’t I at least do that for myself. That is, why can’t I fix the meaning at least for me, in my own head, just by mentally pointing at what I mean? Well, in order this to be a language, I can’t just assume I infallibly know the contents of whatever I say. I still must start by fixing the reference or meaning of the words. I can’t fix the reference from within the language that we share. That’s what step one showed. So, I have to fix it by some kind of mental pointing or, better, having some metalanguage that assigns meanings to all the terms in the object language. Now, from outside and above, from my metalanguage or pointing, I have fixed the first-order language. But the metalanguage is still a language, I am going to need a metametalanguage to fix its terms. The pointing is still pointing. I am still going to have to point at what I was pointing at. And so on. I need another level and another and then I have an infinite regress of metalanguages (or pointings) all frantically trying to fix the reference of the language just below it while getting fixed by the language (or pointing) above it.
One more time. Meaning is a public, behavioral phenomena. You can’t fix the meaning of your own words any easier than you can fix mine. It might seem like you can. But it seems like you could have a private language. But you can’t. Think of it like this. Suppose whenever you encounter something new you label it (museum-style, by ostension (that is, by mentally pointing at it)). After a while, you have a huge list of words. The first problem is that every word in your language is a beetle in a box to everyone else. But are you even capable of keeping it fixed? You have to map the language onto your behavior to make it a language even to you. But for every possible way you could map it, there are many other possible mappings. Why can’t you just specify your preferred mapping? Because you need a metalanguage to do that. You need to be able to say ‘This maps onto that’ from now on. But you need a metametalanguage to fix the terms of your metalanguage. And a metametametalanuge for that.
Quine already believed that nothing is necessary. And if nothing is necessary, then everything is just a theory. All entities are theoretical entities. But now Quine seemed to have shown that there isn’t any way of being sure of what exists according to any particular theory or language. Just as Einstein overturned the Newtonian account of absolute space and time, Quine overturned the myth of absolute meaning.
Just as “there is no absolute position or velocity; there are just the relations of coordinate systems to one another, and ultimately of things to one another… [So] I think that the parallel question regarding [language] calls for a parallel answer, a relational theory of what the objects of theories are. What makes sense is to say not what the objects of a theory are, absolutely speaking, but how one theory of objects is interpretable or reinterpretable in another.”
I bet you’ll wonder about this the next time you see a rabbit. Or a time-slice of four-dimensionally extended rabbitiness. Or whatever. You know what I mean. Or do you?