This is an excerpt from ‘4 Things I Learned About Epistemology’
This next part is a little difficult, because it requires a certain amount of accumulated experience to really land. By that I mean you need to have seen others try and fail to define seemingly simple terms. A famous example is “game”. It is surprisingly hard to define a word like that, considering we use it all the time and so must know what it means on some level. If you’ve never seen proposed definitions of “game” get shot down with counterexamples over and over again, you might not believe me and think it’s really easy to define “game”. This is why Michael Huemer doesn’t cover his view of concepts in Understanding Knowledge until showing us the failure of analysis with respect to “knowledge” – philosophers in the 20th century tried very hard to analyze that particular term, and yet there are good counterexamples to every proposed analysis.
The reason I wanted to cover this topic is exactly because I was already familiar with the difficulty of conceptual analysis – with providing a set of conditions that correctly classifies things in all possible circumstances. I already knew that seemed impossible with philosophically interesting terms, so the section of Huemer’s book examining more fundamental assumptions that led to the problem jumped out at me. But if you don’t occupy a position like that on the epistemic landscape, one where you can see the extreme difficulty of analysis, then this isn’t quite as groundbreaking. Nevertheless, I am not going to spend a chunk of time going through a Socratic dialogue, trying and failing to analyze terms like “game”, “cause”, “time”, “good”, “justice”, and so on just to prove the point. I’ll take it as an assumption going forward that correctly classifying things in all possible circumstances – providing a set of necessary and sufficient conditions free from all counterexamples – is damn near impossible with philosophically interesting terms.
The actual shape of our concepts, so to speak, is fuzzy and complex, not neat and sharp and precise. (In this way, our concepts more closely mirror the physical world than, for example, the abstract world of geometry.) Indefinability is a consequence of those fuzzy borders. The secret is that indefinability is not a problem – it doesn’t matter that we can’t successfully define the words we use.
Think about the gradual evolution of language, and about how we acquired language. Did you study a dictionary? No, you heard others using language and mimicked their usage – you figured out what words meant by noticing how others used them. Which came first? Language or dictionaries? We were acquiring and using language long, long before we tried to systematically define every word we use. And how did we generate these dictionaries in the first place? We basically just consulted linguistic intuition, testing it against specific cases – which really just means imagining using it in different contexts and seeing if it sounds right or not.
Of course, it’s entirely natural to think that you can only use a concept once you already know what it means, and knowing what it means requires being able to give a good definition of it. There’s a view of concepts implicit in this way of approaching the issue, which Huemer calls the Lockean view of concepts, after John Locke:
1. Concepts are introspectively observable mental items.
2. Most concepts are constructed out of other concepts.
3. Introspection is reasonably reliable.
4. Therefore, for most concepts, we should be able to see how the concept is constructed from other concepts.
A definition would then simply describe how the given concept is constructed from other concepts. (Understanding Knowledge, p.31)
This way of approaching concepts doesn’t always fail. Think about a relatively simple geometrical concept, like a triangle. We can see that “triangle” is composed of other concepts: three, straight line, closed. If you understand the concept “triangle”, you can introspect and see that it involves these other concepts. In other words, the concept is constructed out of other concepts. So when it comes to philosophically interesting concepts like justice, knowledge, causation, time, and so on, we should be able to go about it in the same way, right?
Well, this is what philosophers have been trying and failing to do. With the exception of some simple concepts, mathematical concepts, and stipulative definitions, analysis has not yet yielded a perfect definition of pretty much anything. In fact, if some future philosopher does correctly analyze knowledge, the analysis would be so convoluted that it wouldn’t be helpful to anyone who didn’t already understand the word! If one wanted to understand the term, telling them the correct definition would probably confuse them more than just giving examples of how to use it.
“There are three things wrong with the Lockean view. First, understanding a word (except in rare cases) is not a matter of knowing a definition. Understanding a word is a matter of having the appropriate dispositions to use the word – being disposed to apply the word to the things that it applies to, and not apply it to the things it doesn’t apply to. Accordingly, the way that we learn words is hardly ever by being given a verbal description of the word’s meaning. We learn almost all of our words by observing others using the words in context. We then attempt to imitate others’ usage – to apply the word in circumstances that seem to us similar to those in which we have previously observed others using the word. Each time we hear the word used, that slightly influences our dispositions. You understand a given word to the extent that you can successfully imitate the accepted usage.” (UK, p.33)
Huemer goes on to argue that concepts are generally not constructed out of other simple concepts, like in the case of “triangle”, and that concepts are not always introspectively observable. Understanding a word is a matter of having the appropriate dispositions to use the word – concepts are dispositional mental states. Our main access to a concept comes “not through directly reflecting on the concept, but through activating the dispositions that constitute our understanding.” We can do this in real life or just in our imaginations. We imagine a situation and are inclined to use certain words and not others. This not only tracks with how we actually learned language to begin with, but also with the common method of checking definitions by consulting our linguistic intuitions:
“If the way we applied concepts were by applying a pre-given definition, then this methodology would be backwards; we would reject intuitions about cases that conflict with our pre-given definitions, rather than the other way around.” (UK, p.33)
We don’t just keep insisting that knowledge is justified true belief after we’ve been given counterexamples. We don’t say, “Actually, they did have knowledge because it satisfied our pre-given definitions.” Note that, without any successful definition, we nevertheless feel justified in rejecting proposed analyses when they conflict with our intuitions. That’s because we understand the concept even in the absence of a successful analysis.
This use-based way of understanding the meaning of terms is of course going to be fuzzier and more complex than a rigidly defined, geometrically precise method. Usage is messy. There will pretty much always be edge cases – things that don’t seem to quite fit on either side of a sharp line. On this view of concepts, that’s to be expected. Good concepts will admit borderline cases, too.
One note on edge cases: While it’s true that there will always be borderline cases, that shouldn’t be abused. This fact doesn’t give us free rein to define anything terribly since, hey, there will always be counterexamples and edge cases! Taxonomies are more or less useful depending on how comfortably things fit into the categories within the taxonomy. There shouldn’t be mostly borderline cases. They should be few and far between. A perfectly good concept will still have some edge cases, but you obviously don’t want mostly counterexamples and edge cases.
The radical-sounding conclusion of all this is that understanding is constituted by knowing how to use a word, not being able to provide a perfect definition. You can understand a concept without being able to provide a set of conditions that correctly classifies things in all circumstances. It’s misguided and kind of pointless to get hung up on this activity of endless conceptual analysis. As Dennett points out here, for a lot of philosophers, that just is philosophy. And it’s a total dead end.
“Indefinability of words is perfectly normal and does not pose a problem, since understanding is not constituted by knowledge of definitions. The best way to convey a word’s meaning is through examples.” (UK, p.37)
This brings us to a particular passage I came across in Randal Rauser’s book, The Doubter’s Creed: How to Be A Christian When You Don’t Believe It’s True. (I interviewed Rauser about the main subject of the book as well, if you’re interested.) In attempting to define “naturalism”, Rauser explains the method of ostensive definition by negation:
“If one gives an ostensive definition of something, one defines that thing indirectly by pointing to examples that satisfy the definition. For example, you could define ‘ball’ by providing a formal definition like ‘a ball is a spherical object’ (though American football fans might disagree!). But you could also define ball indirectly by pointing at things that satisfy the definition: a basketball, a bowling ball, a ball bearing, and so on. Second, the idea of defining something by negation involves defining it indirectly by saying what it is not. For example, you might begin to define courage by saying it isn’t cowardice, it isn’t weakness, and it isn’t laziness.” (The Doubter’s Creed, p.26)
Rauser brings this up in the context of defining naturalism, which often involves ostensive definition by negation. Naturalists often describe their view by giving examples of supernatural, religious, or things otherwise deemed “spooky” that they don’t believe in. Giving examples, including negative examples, is a good way to convey a term’s meaning. Showing how the word is used – activating the dispositions that constitute understanding – is more important than being able to give a dictionary-worthy definition. (That is, if understanding is the aim.) This isn’t just the case with naturalism; it’s true for most if not all philosophically interesting concepts. Understanding a concept is not a matter of knowing a definition.
You can actually make a name for yourself just through proving the point that it is hard, if not impossible, to successfully define a philosophically interesting term. If you pop out of the bushes and demand to know the necessary and sufficient conditions free from all counterexamples of any word, you will probably stump a lot of people, especially if you already know those people have edge cases at the forefront of their minds.
The Wittgensteinian view of concepts best explains not only the failure of analysis, but how it’s possible that everyone knows how to competently use terms even though it is so hard to successfully analyze them. It coheres nicely with several facts about the way we acquire language, the utility of providing examples instead of definitions to best understand a term, and our standard method of testing definitions against linguistic intuition. I can’t provide an impeccable definition of knowledge that will satisfy counterexample mongers, and yet we all have no issue using the term. In fact, I’m perfectly comfortable rejecting proposed analyses of knowledge! I reject the justified true belief analysis, since it seems wrong to say that I have knowledge of the time when I’m looking at a broken clock. Not only can I competently use terms that I can’t analyze, I can reject proposed analyses as insufficient. That’s because I understand what the word means despite the fact that I can’t define it.
. . .
One note from Huemer on the Wittgensteinian view of concepts and the contrasting Lockean view:
“I think what I have to say about concepts is like some stuff that Wittgenstein said, but I don’t actually care how well it matches Wittgenstein’s views. I also don’t care, by the way, whether the ‘Lockean theory’ matches Locke’s views. You have to add in caveats like this whenever you mention a major philosophical figure, because there are always people who have devoted their lives to studying that figure and who, if you let them, will give you all sorts of arguments that the famous philosopher has been completely misunderstood and never really said the things they’re famous for saying.” (UK, p.33)
Emerson Green 