I made omelets for breakfast, and served us each one. “Mine is bigger!” you said. “Yeah, since you’re going to the gym, I figured you could use some extra protein.”
When you looked at both omelets, you saw that “mine is bigger,” and you believed “mine is bigger,” and it was true:
|Mine is bigger||True||True|
If “mine is bigger” is true, then shouldn’t I believe it too? But it isn’t true… not if I believe it. Only when you believe it.
There’s a broader problem here: statements whose meaning, and truth, depend on something other than the statement itself. In this case, it is who is saying, or believing, the statement. Linguists have cataloged a huge variety of related issues.1 “The dog is a Samoyed” is another example: it is true or false depending on which dog is the dog, and that depends on the context.
The great thing about “239” is that it definitely, eternally, unambiguously refers to the same object, namely the number 239.2 This makes it much easier to evaluate the truth of statements like “239 is prime,” because they always mean the same thing.
So there seems an obvious fix for “the dog is a Samoyed.” Let’s outlaw context-dependent beliefs by decree. You don’t believe the sentence, you believe what it means, which specifies which dog. You can believe “dog1514670 is a Samoyed.” That’s legal because dog1514670 is in the National Dog Registry. There’s only one dog with that ID number, so it refers to the same one in every context, and “dog1514670 is a Samoyed” always means the same thing.
Now you can reason about the Samoyed by applying a universal truth: Samoyeds shed a lot. dog1514670 is a Samoyed, so dog1514670 sheds a lot. If that is true of Samantha, the correspondence theory of truth is vindicated.
At breakfast, we both saw the same thing, so we should both believe the same thing. If there were a National Omelet Registry, we could go look the two of them up on its web site, and could believe “omelet681650346 is bigger than omelet798267196.”
Unfortunately, there is no National Omelet Registry.
What sort of world would rationalism be true of? It would need a Cosmic Object Registry to issue a unique ID number for every object in the universe.3 What physical process could get our brains to believe propositions involving globally unique object IDs? Unfortunately, we cannot consult a Cosmic Object Registry via astral connection. And omelets do not come with radio frequency microchips.
Your omelet and mine were visibly different. If every object was perceptually unique, we could believe sentences like “regarding the omelet that, uniquely in the universe, looks like so-and-so: it is bigger than the one that looks like such-and-such.”4 But objects are frequently indistinguishable in practice, which means that forming beliefs based on unique descriptions is infeasible. If you see a swarm of gnats swirling over a creek, you cannot mentally represent this in terms of distinctive visible characteristics of the individual insects.
Linguists and philosophers originally believed that context-dependency was restricted to a handful of red-flag words like “mine,” “the,” and “today.” They devised technical fixes meant to address each. But gradually it became clear that context-dependency is pervasive.
For reasonableness, context-dependency isn’t a problem; it’s the solution to many practical problems. Which eggplant am I cutting up for moussaka? The question answers itself: it is the one I am cutting up. There’s only one of those, and it is right in front of me. There’s no ambiguity, and I don’t need to look it up in the Cosmic Object Registry to reason about it. In this and other ways, context-dependence makes reasonableness more efficient than rationality. (This trades off against rationality’s power, precision, and accuracy in situations in which it is useful.)
In the meta-rational understanding, context-dependency is not a collection of pathological exceptions to meaning’s typical good behavior. Virtually all sentences, beliefs, or representations about the eggplant-sized world are context-dependent. This is unavoidable for multiple reasons, including the practical impossibility of referring to specific objects other than relative to oneself. Once you accept that, many problems faced by rationalism evaporate. Context-dependency is not a source of problems we should eliminate; it provides resources that rationality depends on to work—even when a rational system has found a universal truth that is true regardless of context.
- 1. The relevant Stanford Encyclopedia of Philosophy article is “Indexicality.”
- 2. Unambiguously—except for the number base you are using, social conventions about numerals, and so on. For contrast with physical reality, let’s pretend mathematics is free from the issues that cause rationalism problems, even if it’s not altogether true.
- 3. See Philip Agre’s Computation and Human Experience for extensive discussion of the object identity problem. (Page 244 discusses the “cosmic registry of forks.”) Logical positivism treated this as the question of how proper names refer to individuals; see the Stanford Encyclopedia of Philosophy article “Names.” In 1970s-’80s artificial intelligence, “knowledge representation systems” explicitly assigned unique ID numbers to every specific object they encountered. Present-day rationalisms seem to depend on this possibility implicitly, but because nowadays no one seriously attempts to explain how rationalism would work in practice, the necessity does not become obvious. It is tacitly assumed that beliefs could refer to specific individuals somehow, without considering what mechanism might achieve that.
- 4. This is the “Description Theory of Names.” Again see the Stanford Encyclopedia of Philosophy article on “Names,” for arguments for and against this and other theories.