On definitions, by Prof. Robb


One major division is between descriptive and prescriptive definitions. Descriptive definitions define a term (or concept) as it’s currently used—if not universally, at least within a given community. Prescriptive definitions, by contrast, say how a term should be used, whether or not this matches current usage.

Two quick examples where the two kinds can come apart: (1) It is sometimes said that gay marriage doesn’t fit “the definition of marriage”. Possibly this is (or was) true descriptively, but so what? Proponents of gay marriage could be saying how the term should be defined, not how it’s actually defined. (2) Given how “planet” was defined by the Ptolemaic astronomers, it was nonsensical to say that the Earth is a planet. But Copernicus could say (and in effect did say): so much the worse for the definition. What the term should mean allows for Earth to be a planet.

What guides our search for a prescriptive definition? We are guided by normative principles—social and ethical principles in example (1), epistemic principles in example (2). Note that when asking for a prescriptive definition, one isn’t asking, “What do I want this term to mean?” One’s desires don’t necessarily reflect the relevant normative principles.

A few years ago, our course asked students to define “revolution” and “humanities.” Let’s make these our main examples for discussion. Are we here aiming for descriptive or prescriptive definitions? My own hunch is that for “revolution,” we’re looking for a descriptive definition, one abstract enough to cover political revolutions, scientific revolutions, personal revolutions, and so on. But in the case of “humanities,” I think we’re really asking how we should use this term: we want a prescriptive definition. Put another way: the search for a definition of “humanities” is aspirational. But others will disagree—this is something you will need to decide as you try to define this term and others.

By the way, prescriptive definitions usually borrow from their descriptive counterparts, so the two kinds aren’t wholly distinct. This is clearly the case in examples (1) and (2) above. A prescriptive definition that was entirely insensitive to actual usage would amount to just changing the subject, which isn’t the idea.


A common complaint about the search for definitions is that it’s “just semantics,” implying, I guess, that it’s trivial, arbitrary, a matter for stipulation, or something like that. This can sometimes be the case, especially when what’s being defined is a technical term. But most of the time, the search for definitions isn’t like this. When we ask about the definition of a term, we’re really asking about the nature of what the term stands for (alternatively, the nature of what it should stand for: see Part 1). The word itself isn’t as important.

One way to see all this is to switch from the “formal mode” (in which we’re talking about words) to the “material mode” (in which we’re talking about things). For example, framing the question of Plato’s Republic in the formal mode, we might ask:

What is the definition of “justice?”

But switching to the material mode, we would ask:

What is justice?

Even if these two questions are equivalent, the second makes it clearer that in the search for definitions, we’re asking about something in the world (justice, in this case)—we’re trying to discover its nature. And the nature of something usually isn’t trivial, arbitrary, or a matter for stipulation.

The point can be easy to miss, because the meaning of a word is up to us: words don’t have their meanings intrinsically. If meaning is up to us, shouldn’t getting definitions be easy, a matter of invention rather than discovery? No, because in most cases what we’re defining is a term that’s already been given a meaning, i.e., already stands for something. And it’s the nature of that thing that we’re really after. (Even in the case of prescriptive definitions, we’re after something external to us, namely the nature of the thing that’s the best referent of the term—best, that is, according to the relevant normative principles: see Part 1.)

Return then to the sorts of question we have asked in Humanities:

What is the definition of “revolution?”

What is the definition of “humanities?”

Putting them this way is fine, but to avoid the “just semantics” complaint, I’d prefer put them in the material mode:

What is a revolution?

What are the humanities?


In Plato’s early “Socratic” dialogues, when Socrates asks for the definition of a term (e.g., “courage,” “piety,” “justice”), his interlocutor often responds with list of things that satisfy the term (sometimes the list has just one member). Call this the “list method.” For example, someone trying to define “courage” might say it means: standing firm in battle, engaging in public speaking, taking organic chemistry.

The list method could be applied to our earlier terms. The humanities are the topics and methods of English, Philosophy, the languages, the arts, and so on. A revolution is the American Revolution, the Russian Revolution, the Copernican Revolution, and so on. (Here and below, I will sometimes move freely between formal and material modes—see Part 2 for this distinction.)

One problem with the list method is the “and so on” part: shouldn’t the list be explicit and complete? But how to complete it? For lots of terms, the list may be indefinitely large. Think, for example, of all of the various ways one might be courageous, or all of the revolutions throughout history. But perhaps this problem isn’t too serious: as long as the items on the list are representative, that could be enough.

But there’s a more fundamental problem with the list method, and this is what Socrates liked to point out: Even if the list is complete (or at least representative), it still fails as a definition, for it doesn’t say what all of the things have in common in virtue of which they satisfy the term. When we’re defining “revolution,” we want to know what makes all of these things revolutions, what unifies them so that they all deserve the name. And a mere list doesn’t tell us that. The same goes for all of the examples of humanities, or the examples of courage.

I agree with Socrates: a mere list can’t be a definition. But I’ll say at least this much about the list method: it’s a start. That is, if you’re trying to define a term, it’s useful to have a few clear examples to work with. They can serve as data as you formulate your definition.


If Socrates is right, a definition of a term X (“courage.” “justice.” etc.) shouldn’t be merely a list of things that are X. It should say what all Xs have in common in virtue of which they are Xs. A traditional way of spelling this out is that a definition of X should present necessary and sufficient conditions for being X.

A necessary condition for being X is a condition that something must have to count as X.

A sufficient condition for being X is a condition that by itself is enough for something to count as X.

If I define X as C, then C should be both necessary and sufficient for being X. (Note that C might have several components: in the usual case, the components are each necessary for X, and taken together are sufficient for being X.) The process of discovering necessary and sufficient conditions is sometimes called “conceptual analysis.”

It’s very difficult to work with this distinction, and it trips up lots of people, including philosophy professors. You can find examples and exercises online if you want some practice. Start with something relatively easy: what is the definition of “table” (material mode: what is a table)? Suppose I say:

A table is a piece of furniture with four legs.

This definition fails to present a necessary condition for being a table, because there can be tables with fewer than four legs. (Some have one; some have none, as when a table folds out of a wall.) The definition also fails to present a sufficient condition for being a table, because a chair is a piece of furniture with four legs, but a chair isn’t a table. To fix these problems, I might try:

A table is a piece of furniture designed to hold small objects on its top surface.

This looks better, but we might still wonder if it presents a sufficient condition, for a shelf fits the condition, but isn’t a table. It might not present a necessary condition either, for there can be “artistic” tables that were designed to be viewed, but never used. (One might ask at this point: why not just look up “table” in the dictionary? That’s the topic of Part 5.)

What about “humanities?” A first try could be:

The humanities are those academic disciplines that use qualitative methods.

But this doesn’t present a sufficient condition, for some of the social sciences are qualitative. Nor does it present a necessary condition: some philosophers (studying mathematical decision theory, for example) use quantitative methods, as do the digital humanities.

Taking a cue from a previous faculty member, Prof. Ingram’s, lectures a few years ago, we might try this:

The humanities are those academic disciplines that emphasize intertextual references, the development of empathy, and style.

In fairness to Prof. Ingram, we should note he was presenting these conditions as “tendencies” of the humanities, not a definition, but it’s still instructive to consider it as a definition. One problem here is that this seems not to present a necessary condition, for philosophy is in the humanities, but philosophy doesn’t usually emphasize any of these. Does the definition at least present a sufficient condition? You might try on your own to challenge this. (For example, are there social sciences that emphasize these three features?)

In his Revolutions: A Very Short Introduction (OUP 2014, p. 4), Goldstone defines “revolution” like this:

Revolution is the forcible overthrow of a government through mass mobilization (whether military or civilian or both) in the name of social justice, to create new political institutions.”

It’s worth thinking about whether this presents necessary and sufficient conditions for being a political revolution (which is all Goldstone intends, I believe). But if we were to take this as a definition of “revolution” in its most general sense, it clearly fails to present a necessary condition, for it doesn’t include scientific or artistic revolutions.

These are just initial attempts—I won’t try here to present a successful conceptual analysis of “humanities” or “revolution.” I leave these as exercises.


If we’re trying to define a term, it’s natural to consult the dictionary. Lexicographers are trained scientists, and very good at what they do. So if I want to know what the humanities are, or what a revolution is, why can’t I just look it up?

One immediate problem is that a standard dictionary is trying to capture actual usage, and thus isn’t too helpful in formulating a prescriptive definition (see Part 1). If my aim in defining “humanities” is to say what the humanities should be, a dictionary won’t be too useful. But let’s set this point aside and assume we’re concerned only with descriptive definitions. Even then, a dictionary has some problems.

Here is my dictionary’s first definition for “table”:

“a piece of furniture with a flat top and one or more legs, providing a level surface on which objects may be placed, and that can be used for such purposes as eating, writing, working, or playing games.”

This fails to present a necessary condition: A table can have a non-flat (warped) surface, but it’s still a table, even if a bad table. And as noted in Part 4, a table can have zero legs (think of a table that folds out of a wall, or a “Jetsons” table that hovers). And some antique tables are so fragile that they can’t be used for the purposes listed.

Things get even worse when we move to more important terms. Consider “morality.” Philosophers have struggled to define this. Is the moral action the one that produces the most benefit? Is it the one that respects rights and duties? Should God play a role in a definition of morality? These are good questions. But here’s my dictionary’s first definition:

“principles concerning the distinction between right and wrong or good and bad behavior.”

This isn’t helpful. It lacks the theoretical precision and informativeness we want out of a definition. And it looks circular as well, since it uses the terms “right,” “good,” and so on, which are pretty closely tied to the term we’re trying to define. It’s worth noting, however, that this definition arguably does present necessary and sufficient conditions for morality. What this shows is that we want more than just necessary and sufficient conditions in a definition: we also want a definition to be useful and informative. And dictionaries often fail in this respect.

For “humanities,” my dictionary has this:

“learning or literature concerned with human culture, especially literature, history, art, music, and philosophy.”

This looks a lot like a list (Part 3), and thus has the limitations of that method. (It also includes history in the list, but history is, at least at Davidson, classified as a social science.) And anyway, being concerned with human culture is neither necessary nor sufficient for being in the humanities. Not necessary: many branches of philosophy aren’t about human culture. Not sufficient: most social sciences are about human culture.

As for “revolution,” my dictionary has this (I’ve edited it some):

a forcible overthrow of a government or social order in favor of a new system, [or any] dramatic and wide-reaching change in the way something works or is organized or in people’s ideas about it.

This isn’t bad, as the second part covers scientific and artistic revolutions. But Goldstone (quoted earlier) wouldn’t like this: he’d say it fails to present a sufficient condition, and I’m inclined to agree. Not any overthrow of government or wide-reaching change counts as a revolution.

None of this is to say that dictionaries shouldn’t be used—they are useful for getting the basic idea of what something means. But if your goal is a deeper, theoretical understanding of something, dictionaries are limited.


Sometimes the search for necessary and sufficient conditions (Part 4) fails. Many of Plato’s early dialogues end without a successful definition of the central term, and the question is what to do then. We could just admit that we know very little about what’s being defined. (This seems to have been Socrates’ lesson.) A more drastic reaction is to say that conceptual analysis is impossible. Some philosophers have in fact thought that informative necessary and sufficient conditions cannot be obtained for any interesting terms. I think this is too pessimistic. But for the purposes of this last part, let’s assume that we can’t give necessary and sufficient conditions for terms such as “humanities” or “revolution.” Then what?

One idea is to move to what’s called a “cluster definition.” Here’s the basic idea: instead of worrying about necessary and sufficient conditions for X, we first just think about all the conditions that are relevant to being X and list them. (This isn’t the list method from Part 3; what we’re listing are not examples of Xs, but conditions or criteria relevant to being X.) Suppose for example we’re trying to define “art.” What is it for something to be a work of art? Here’s a sample list of relevant conditions from a book by philosopher Denis Dutton. (To construct this list, I clipped short phrases from a review of Dutton’s book linked here.)

Art is a source of experiential pleasure.
The making of art involves skill and virtuosity.
Works of art have a recognizable style.
Art involves novelty and creativity.
Art is subjected to critical judgment.
Works of art are representational.
Art is separate from ordinary life.
Art expresses individuality.
Art is shot through with emotion.
Art challenges the intellect.
Works of art have an identity by virtue of being part of a tradition or institution.
Art provides an imaginative experience.

Now you might want to quibble with some of these items. But remember that we’re not saying any one of these is necessary for being art: we’ve abandoned that search. Rather, we are saying:

Something that satisfies all of these is definitely art; so the conditions are collectively sufficient.

Something that satisfies none of these is definitely not art; so while no single condition is necessary, something must have at least one of these to count as art.

Something that satisfies some of these is, to that extent, art.

And that’s all there is to defining “art.” (We could add a bit more sophistication by weighting the criteria, making some more important than others, but that’s optional.) That’s what a cluster definition is.

So might a cluster definition be the right option for “humanities” and “revolution?” Maybe—this is something for you to think about. But for what it’s worth, I’m skeptical that we can give necessary and sufficient conditions for “humanities” (and this is so whether we’re after a descriptive or prescriptive definition: see Part 1). So for the humanities, we must settle for a cluster definition. In that case, we should construct a list of conditions relevant to something’s being part of the humanities—Prof. Ingram’s list of “tendencies” in the humanities (see Part 4) is a good start. As for “revolution,” I’m inclined to think we probably can give necessary and sufficient conditions, so that we won’t have to settle for a cluster definition.