Ordering Your Adjectives

It’s time to make good on my pledge to write about the three most interesting topic suggestions I got during the Grammar Girl book giveaway contest. Actually, I got significantly more than three suggestions that would be interesting to write about, and they’ll be a source of inspiration in months to come. But I only have three books to give away, so with some difficulty I’ve settled on the three topics that I’m going to write about first. (Well, first, second, and third, actually, but you get my generalized meaning of first, don’t you? Of course you do.) I’ll start with a suggestion from Carlos Iriarte, who wrote:

As a non native speaker, I would love to see a discussion about the possessive apostrophe, or the proper order of adjectives.

Carlos, you’ll have to go somewhere else to learn more about apostrophes, but as it happens, adjective ordering is a topic that I’ve been curious about for years. I remember trying to do a research paper on it as a grad student, for a seminar in lexical semantics, and finding the topic so complicated that I repeatedly narrowed the focus of the paper, eventually ended up with a sorry piece of work investigating restrictive and non-restrictive uses of just one single adjective, and got the paper back with the comment, “I would not spend any more time pursuing this line of research.” But a blog post, I think I can handle.

Cumulative (and pseudo-cumulative) adjectives

In many English grammars and usage guides, discussions of adjective ordering go hand-in-hand with advice on when series of adjectives should be punctuated with commas. These discussions usually talk about two kinds of adjective: cumulative and coordinate. With cumulative adjectives, you don’t use commas, but how do you know if you’re dealing with cumulative adjectives? There are two diagnostics, both of them based on the meaning of these adjectives, so I’ll start with the semantics of cumulative adjectives.

Suppose you have cumulative adjectives A₁, A₂, and A₃, and noun N, in the string A₁+A₂+A₃+N. A₃ modifies N; A₂ modifies the nominal A₃+N; A₁ modifies the nominal A₂+[A₃+N]. The best example of this kind of modification involves adjectives such as fake. For an example, we’ll use fake Italian marble. Italian modifies the noun marble, straightforwardly enough. Fake, however, doesn’t modify just marble. It builds on the adjective+noun cluster it’s sitting next to, and modifies the nominal Italian marble. Fake Italian marble could be fake marble from Italy, real marble from someplace other than Italy, or even fake marble from someplace other than Italy. If you want to zero in on the “fake marble” meaning, you’d say Italian fake marble. So in this example, the order of fake and Italian depends on the meaning you’re trying to convey. In other words, you can rearrange them, but you’ll have a different meaning.

This fact is the basis for the first diagnostic for cumulative adjectives: the “rearrangement test”. If you can’t rearrange the adjectives and have a grammatical phrase with the same meaning as the original, the adjectives are cumulative. The second diagnostic is the “and test”: Can you separate the adjectives with and and have a grammatical phrase that means the same thing as the original? If not, the adjectives are cumulative. Fake Italian marble fails the and test. Fake and Italian marble means marble that is definitely fake and definitely Italian. The meaning of real marble that’s not from Italy is no longer possible. So on both counts, fake and Italian are cumulative adjectives in the phrase fake Italian marble.

So far, so good, but with many other adjectives, we won’t get such a clear answer. There are many adjectives that have to go in a fairly fixed order, but for no apparent meaning-related reason. For example, how about little green bag? To a native speaker, ?green little bag sounds weird, so little and green clearly fail the rearrangement test.

This is where some grammar and usage guides go off the rails. They try to extend the reasoning for cases like fake Italian marble to cases like little green bag. They’ll say that little isn’t modifying just bag; it’s cumulatively modifying the chunk green bag. OK, let’s look at that a little closer. Let’s say bag refers to the set of bags. Green bag will refer to a subset of the set of bags, namely, the green ones. Little green bag will refer to a subset of the green bags, namely the ones that are also little. Perfect! Just the meaning we wanted! The trouble is that we can do the same exercise with ?green little bag and end up with the same set: bags that are both little and green. So we’re back to wondering why little has to come before green. The unfortunate students of English as a foreign language, meanwhile, are scratching their heads, trying to discern the big meaning difference between little green bag and green little bag.

Why did the rearrangement test work so well with fake Italian marble, but fail with little green bag? It has to do with the semantics of adjectives like fake and other adjectives. Adjectives like little and green are known as subsective adjectives, which means that, for example, green bag will refer to a subset of the set of bags. Fake, on the other hand, is a non-subsective adjective. (It’s also sometimes called a privative adjective.) In terms of our example, fake marble doesn’t refer to a kind of marble.

Anyway, let’s say that, having forced the same semantic reasoning on little green bag as we did with fake Italian marble, we’re now ready to run little green bag through the and test. Can we say little and green bag? To a native speaker, this will sound pretty awkward … or will it? To me it does, but so does five green and speckled frogs, which evidently sounded fine to the writer of that song. But if we ignore these misgivings, that would be a failure for the and test, which lets us conclude that little and green are cumulative adjectives here.

But even though we’ve managed to cram fake Italian and little green into the same cubbyhole, there’s a slightly altered and test that the grammar guides don’t use, which gives undeniably different results for fake Italian and little green. Instead of keeping the adjectives in attributive position (that is, modifying the noun), put them in predicative position (for example, after is/are) and then put in your ands. This marble is fake and Italian has the same problem as fake and Italian marble, so it fails both and tests. But The bag is little and green sounds fine — in fact, the and is mandatory — and we’re still talking about a bag that is both little and green. So little green bag fails one and test, but passes the other.

So here’s where we are so far: Non-subsective adjectives really do modify nouns (or adjective+noun clusters) in a cumulative way, and the name cumulative adjective fits. Their ordering is determined by the way their semantics work. More typical subsective adjectives don’t modify nouns in a cumulative way, and calling them cumulative adjectives is at best unhelpful and at worst downright confusing. So what determines the order you put them in, if not their meaning?

Linguists still do not have a definitive answer. What they do have is a hierarchy of adjective classes that (for whatever reason) occurs in a more-or-less fixed order in English. The order is fixed in other languages, too, though not all, and it’s not quite the same order across languages. Here is a composite hierarchy I’ve assembled from those given in several sources I’ve looked at:

evaluation

size

shape

condition

human propensity

age

color

origin

material

attributive noun

I’ll call adjectives like these subsective fixed-order adjectives, but if you’re really used to thinking of them as cumulative adjectives, maybe the term pseudo-cumulative adjectives will go down better. The ordering hierarchy for these adjectives can be violated if there’s an adjective you want to put first for emphasis (I want the GREEN little bag, not the RED one!: COLOR>SIZE), or maybe to grab people’s attention (brown-eyed handsome man: COLOR>EVALUATION), or for unclear reasons (big bad wolf: SIZE>EVALUATION), but for the most part, following respecting the hierarchy will produce more native-like adjective orderings than choosing randomly.

Coordinate adjectives

Grammar and usage guides define coordinate adjectives as the opposite of cumulatives in every way. They’re separated by commas. They can be separated by and. They can be rearranged. They each modify the noun separately, not cumulatively. These four properties are valid; the question is, when do adjectives behave this way? The answer is that when you have more than one adjective in the same class from that hierarchy, they become coordinate adjectives. For example, there are four evaluative adjectives in the nominal phrase sexist, egotistical, lying, hypocritical bigot, from the 1980 movie 9 to 5. These adjectives could easily be rearranged…

egotistical, lying, hypocritical, sexist bigot
sexist, lying, egotistical, hypocritical bigot
lying, hypocritical, sexist, egotistical bigot
hypocritical, sexist, egotistical, lying, bigot
…

… but in the movie they never are. Multiple characters, including the boss himself, refer to the boss in this way, and part of the humor arises from the treatment of these coordinate adjectives as if they were arbitrary fixed-order adjectives, or as if sexist, egotistical, lying, hypocritical bigot were a common, recognized and lexicalized compound noun. The order was probably chosen in order to emphasize the rhyme between egotistical and hypocritical.

Here’s an example of mixed pseudo-cumlative and coordinate adjectives, from Weird Al Yankovic’s “Generic Blues” (from the same album as the song I wrote about here):

I’m just a no-good, scum-sucking, nose-picking, boot-licking, sniveling, groveling, worthless hunk of slime.
Nothing but a low-down, beer-bellied, bone-headed, pigeon-toed, turkey-necked, weasel-faced, worthless hunk of slime

These adjectives could be rearranged, although with some constraints. The negatively evaluative no-good and low-down probably need to stay where they are, at the beginnings of their respective strings of adjectives. In other words, they’re pseudo-cumulative with respect to the other adjectives. Worthless, if it goes anywhere else, would go with no-good and low-down; why Yankovic put it at the end I can only guess. Maybe for him, worthless hunk of slime was already a set phrase before he incorporated it into these lyrics. But the rest could be reordered among themselves; in other words, they are coordinate adjectives with respect to each other. Yankovic himself chose to cluster all the “human propensity” present-participial adjectives, and further separate them into direct-object-incorporating ones (scum-sucking, nose-picking, boot-licking) and intransitives (sniveling, groveling); and do the same with the “shape/physical appearance” past-participial adjectives involving body parts, with those involving animals further clustered. According to the hierarchy above, the human propensity adjectives would precede the physical appearance ones if they appeared in the same noun phrase, but that constraint could probably be pushed a bit.

The 9 to 5 and “Generic Blues” coordinate adjectives also pass the and test, although to get a result less subject to speaker variation, I’d go with the predicate-adjective version: The boss is sexist and egotistical and …, and This hunk of slime is low-down and beer-bellied and …. In attributive position, separating adjectives with and usually only sounds natural when they are (or sound like) they’re contradictory; for example, a complex legal and economic issue. Unfortunately, even the predicative version of the and test doesn’t work so well for adjectives like nose-picking and boot-licking, as speakers would probably say, “He picks his nose” or “He’s a boot-licker.” Oh, well, no test is perfect.

To sum up: If there’s definitely a meaning difference between different adjective orderings, let that determine how you order them, and don’t use commas. If you can’t find a meaning difference, don’t go trying to force there to be one. Instead, go by the adjective-ordering hierarchy, and don’t use commas. If more than one adjective has the same kind of meaning in the hierarchy, then use commas, or ands or buts if the adjectives have contrastive meanings.

Carlos, thanks for the suggestion, and I have a copy of Grammar Girl’s 101 Words Every High School Graduate Needs to Know to send you … as soon as you can give me a postal address.